- Email: [email protected]
Cost competitiveness of major airlines: an international comparison
Transpn Res.-A, Vol. 32, No. 6, pp. 407±422, 1998 # 1998 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0965-8564/98 $19.00+0.00
Pergamon PII: S0965-8564(98)00007-X
COST COMPETITIVENESS OF MAJOR AIRLINES: AN INTERNATIONAL COMPARISON TAE HOON OUM* and CHUNYAN YU
Faculty of Commerce and Business Administration, The University of British Columbia, Vancouver, B.C., V6T 1Z2 Canada (Received 15 January 1997; in revised form 2 January 1998) AbstractÐThis paper compares unit cost competitiveness of the world's 22 major airlines over the 1986±93 period. First, a unit cost index for aggregate output is computed via a multilateral index procedure. A translog variable cost function is estimated and used to decompose the unit cost dierentials into potential sources: input prices, network and output attributes, and eciency. The results of the unit cost decomposition are used to construct a cost competitiveness indicator after removing the eects of network and output attributes. Our results for 1993 are: (a) Asian carriers (except Japan Airlines and All Nippon Airways) were generally more cost competitive than the major U.S. carriers, mostly due to their substantially lower input prices; (b) Japan Airlines and All Nippon Airways were over 50% less cost competitive than American Airlines mainly because of their high input prices; (c) major European carriers were 7% (British Airways)±42% (Scandinavian Airlines Systems) less cost competitive than American Airlines, because of higher input prices and lower eciency; (d) among the U.S. carriers, American Airlines, United Airlines and Delta were similar in cost competitiveness, while Northwest and Continental enjoyed, respectively, 5 and 12% cost competitiveness over American Airlines; (e) exchange rate ¯uctuation has had considerable eects on the cost competitive position of Japan Airlines and Lufthansa. # 1998 Elsevier Science Ltd. All rights reserved Keywords: airlines, cost competitiveness, productive eciency, decomposition 1. INTRODUCTION
Since the mid-1980s, the world airline industry has experienced a period of signi®cant structural, institutional and regulatory changes. Many countries have deregulated their domestic airline industries. A series of very liberalized bilateral agreements have been signed between the U.S. and countries such as the Netherlands, Belgium, Austria, Switzerland, Finland, and Denmark. Open skies continental blocs are being formed both in Europe and in North America. A movement to create a liberalized continental bloc is under way in Asia Paci®c. As a result, more airlines are being exposed to the pressures of the market-place as the widely-based deregulation and liberalization process advances. The increased competition and recent recession have led to severe and widespread losses in the international airline industry, and forced carriers to undertake major restructuring in order to improve eciency and reduce costs. As the competition intensi®es, the ultimate ability of a carrier to survive and prosper in the globalizing airline industry depends greatly on its cost competitiveness. What constitutes cost competitiveness of an airline? In the simplest terms, an airline is cost competitive if its unit cost (average cost) is lower than that of its competitors on a sustainable basis. An airline may have a lower unit cost than its competitors because it is more ecient, pays less for its inputs, or both. That is, airlines' cost dierentials are determined by dierences in factor prices (including exchange rates) and productive eciency. Knowledge about existing levels and sources of cost dierentials are essential for analyzing public policies and carrier strategies designed to enhance airlines' competitive position. It is important to measure these dierentials and to ascertain their sources. The policies/strategies one would pursue depend on the relative importance of sources of the cost dierentials. A number of studies have focused on airline productivity and eciency. These include Caves et al. (1987), Gillen et al. (1985, 1990), Bauer (1990), Encaoua (1991), Good et al. (1993, 1995), *Author for correspondence. Fax: 001 604 822 8320; e-mail:[email protected]
407
408
Tae Hoon Oum and Chunyan Yu
Ray and Hu (1993), Distexhe and Perelman (1993) and Tofallis (1996). Few studies, however, have examined the issue of airline cost competitiveness. Windle (1991) attempted to attribute the differences in unit costs between carriers to various sources, namely output scale, input prices, operating characteristics, non-optimal capital investment, technical eciency, government ownership and ®rm-speci®c eects. His study was mainly based on the 1983 annual data for 14 U.S. and 27 non-U.S. airlines. Good and Rhodes (1991) examined three aspects of competitiveness and profitability of airlines in the Paci®c region: input prices, output prices and productive eciency, using a panel of 37 airlines over the 1976±86 period. A recent study by Baltagi et al. (1995) applied a translog variable cost function to a panel data of 24 U.S. airlines over the period of 1971±86 to analyze cost changes in the U.S. airline industry in the pre- and post-deregulation era and to identify the cost changes due to technical change, economies of scale and density, and input prices. To our knowledge, no previous study has focused on measuring and comparing cost competitiveness of the world's major airlines. This paper investigates issues related to cost competitiveness of airlines by analyzing the performance of 22 major international airlines during the 1986±93 period. Our analysis focuses on identifying the sources of airlines' cost competitiveness. Speci®cally, a translog variable cost function is estimated, and its results are used to decompose the observed unit cost dierentials among airlines into potential sources: attributes of the ®rm (network characteristics and output mix), input prices, and eciency. The results of the unit cost decomposition are used to assess and compare the cost competitiveness across the sample airlines. In addition, it is believed that exchange rate ¯uctuation has had non-negligible impact on some carriers' cost, therefore, the eect of exchange rate ¯uctuation on factor prices, consequently on unit cost, is also examined via two brief case studies. The following section describes the methodology used for cost function estimation and decomposition. Section 3 then provides a description of the data base. The results of cost function estimation and unit cost decomposition are presented in Section 4, while Section 5 examines airline cost competitiveness. The eect of exchange rates on cost competitiveness is examined in Section 6. The summary and concluding remarks are given in the ®nal section. 2. METHODOLOGIES
This section describes the methodologies for estimating a neoclassical cost function and for decomposing the unit cost dierentials between airlines into potential sources. Since capital input is not always in equilibrium in the airline industry, we follow Caves et al. (1984) and Gillen et al. (1990), and employ the following translog variable cost function to re¯ect the short run cost minimization process (with the usual restrictions on symmetry and linear homogeneity of cost in input prices):* ln VC a0 t at Tt by ln Y i i ln Ri i bi ln Wi bk ln uK 1 1 1 be ln E c ln Z dyy ln y2 i j dij ln Wi ln Wj dkk ln uK2 2 2 2 1 1 2 2 dee ln E dzz ln Z i eyi ln Y ln Wi eyk ln Y ln uK eye ln Y ln E 2 2 eyz ln Y ln Z i fki ln uK ln Wi fke ln uK ln Z i gei ln E ln Wi i gzi ln Z ln Wi
1
where VC is the cost of variable inputs; Y is the aggregate output index; W is a vector of input prices which consists of labour, fuel, and purchased services and materials inputs; K is capital stock which is treated as ®xed in the short run; u is utilization of capital stock (in this case, weight load factor); Ri 's are the revenue shares of freight and mail, non-scheduled services, and incidental servicesy; Z is average stage length; and E is an eciency index computed in Oum and Yu (1995). *Readers who are not familiar with the neoclassical translog cost functions and applications to transportation industries, please refer to, for example, Caves et al. (1984) or Gillen et al. (1990). y The revenue share variables (re¯ecting output mix), Ri 's, are incorporated only in the ®rst-order terms in order to keep the cost function simple.
Cost competitiveness of airlines
409
at 's are coecients associated with year dummy variables (Tt 's) in order to capture the eects of shifts in technical eciency over time. a0 , b's, 's, c, d's, e's, f's, g's are coecients of the translog variable cost function to be estimated.* Note that the variable cost function (1) includes a variable E for eciency. E is the index of residual total factor productivity estimated by Oum and Yu (1995). It re¯ects the overall eciency level of the airlines after removing the eects of factors beyond managerial control, including average stage length and output mix. By including E in the cost function estimation, we recognize the fact that some airlines fail to be on the production frontier, that is, some ®rms are more ecient than others. Once we recognize this, failure to include an eciency indicator may lead to misspeci®cation of the model, and thus bias parameter estimates of the cost function. Therefore, we essentially use a two-step procedure to estimate the cost function. In the ®rst stage, an eciency index is estimated, and in the second stage, the estimated eciency index is used as an explanatory variable in the cost function estimation. In this way, we can explicitly examine eciency eects on airline cost. The following cost minimizing variable input cost share equations can be derived by applying Shephard's lemma to the variable cost function (1): Si
@ ln VC bi j dij ln Wj eyi ln Y fki ln uK gei ln E gzi ln Z @ ln Wi
2
To improve the eciency of estimation, it is customary to estimate the translog variable cost function (1) jointly with the variable input cost share eqn (2). In order to improve the eciency of estimation further, Oum and Zhang (1991, 1995) proposed to add the following expression for the shadow value of capital stock: ÿ Ck @ ln VC ÿ ÿ bk dkk ln uK eyk ln Y i fki ln Wi fke ln E fkz ln Z VC @ ln uK
3
where Ck is the depreciated capital cost which is approximated by the total capital cost multiplied by utilization rate. Equation (3) is basically the ®rst order condition for short-run total cost minimization which endogenizes the capacity utilization. Following Oum and Zhang (1991), we estimate the translog variable cost function (1), the cost share eqnsy (2), and the shadow price of capital input eqn (3) jointly as a system of multivariate equations using a Maximum Likelihood method. Drawing on the properties of a translog variable cost function, Caves and Christensen (1988) and Fuss and Waverman (1992) showed that the (total) unit cost dierential (including capital costs) between any two observations, 1 and 0, can be decomposed into various sources using the following formula: h i c1 ÿ c0 S 1=2 d 1y Cv d 0y Cy ÿ 1 Y 1 ÿ Y 0 + size S 1=2 d 1k Cv d 0k Cv K 1 ÿ K 0 1 1 ÿ S K ÿ K 0 ÿ Y 1 ÿ Y 0 S 1=2 d 1r Cv d 0r Cv R1 ÿ R 0 output mix + 4 S 1=2 d 1w Cv d 0w Cv W 1 ÿ W 0 input prices 1 ÿ S W1k W0k S 1=2 d 1z Cv d 0z Cv Z 1 ÿ Z 0 operating characteristics S 1=2 d 1t Cv d 0t Cv t 1 ÿ t 0 time effects S 1=2 d 1e Cv d 0e Cv E 1 ÿ E 0 efficiency *Please note that subscript is given to each regression coecient for ease of relating the coecient to the variable. y To avoid singularity of variance±covariance matrix, the materials cost share equation was dropped from the estimation. It is well known that the maximum likelihood estimates are invariant to the choice of the share equation dropped.
410
Tae Hoon Oum and Chunyan Yu
where S denotes the average share of variable cost in the total cost for observations 1 and 0, and dxiCv denotes the partial derivative of the variable cost for observation i with respect to variable x. For ease of presentation, this paper uses American Airlines (AA) as the benchmark ®rm against which to compare other airlines. 3. THE DATA
Our sample data consists of annual observations on 22 major international airlines over the 1986±93 period. The primary data are compiled mainly from the Digest of Statistics published by the International Civil Aviation Organization (ICAO), in particular, the annual series of Trac, Fleet & Personnel and Financial Data. Additional data were obtained directly from airline companies, The Airline Monitor, IATA publications, Statistics Canada publications, and a number of other sources. Airline annual reports were used to supplement, cross-check, and correct errors in the ICAO data. The airline companies were contacted for clari®cation when the various sources of data could not be reconciled. Estimation of the variable cost function (1) and decomposition of the unit cost into sources using eqn (4) require detailed data on outputs, network and output attributes, amount and utilization of capital stock, and input prices. 3.1. Outputs Five categories of airline outputs are considered: scheduled passenger service (measured in revenue-tonne-kilometres or RTK), scheduled freight service (measured in RTK), mail service (measured in RTK), non-scheduled services (measured in RTK), and incidental services (non-airline businesses). Incidental services include a wide variety of non-airline businesses such as catering services, ground handling, aircraft maintenance and reservation services for other airlines, sales of technology, consulting services, and hotel business. These are non-core activities of an airline, but use up part of the inputs reported in the data sources which are used by most researchers. Many previous researchers have ignored the presence of non-airline businesses, by failing to include the incidental outputs. Our data show that revenues from non-airline businesses account for up to 30% of total operating revenues for some airlines, with an average of 8% for the airlines included in this study. Therefore, omission of the incidental output, without excluding the inputs used to produce them, would bias performance measures in favour of the airlines who do little non-core businesses. A quantity index is constructed for the incidental output in order to include the incidental services in our analysis. The index is computed by de¯ating incidental revenues by a general price index constructed using the Purchasing Power Parity (PPP) index for GDP obtained from the Penn World Table (Summers and Heston, 1991) and the U.S. GDP de¯ator. The PPP index adjusts for changes in market exchange rates and changes in real price levels of various countries relative to the U.S., and the US GDP de¯ator ensures that the quantity index is comparable over time. A multilateral output index is formed by aggregating the ®ve categories of outputs using the following multilateral index procedure proposed by Caves et al. (1982).
ln Yk
X Rik R i i
2
ln
Yik Y~ i
5
where Yk is the aggregate output index for observation k; Yik is the output i for observation k; R ik is the revenue share of output i for observation k; R i is the arithmetic mean of the revenue share of output i over all observations in the sample; and Y~ i is the geometric mean of output i over all observations. 3.2. Inputs Five categories of inputs are considered: labour, fuel, ¯ight equipment, ground property and equipment (GPE), and materials. The price of labour input is measured by the average
Cost competitiveness of airlines
411
compensation per employee (including bene®ts). Fuel price is obtained by dividing total fuel cost by gallons of fuel consumed.* For ¯ight equipment, a ¯eet quantity index is constructed by aggregating dierent types of aircraft using the translog multilateral index procedure proposed by Caves et al. (1982). The leasing price seriesy for these aircraft types are used as the weights for aggregation. The annual cost for each aircraft type is estimated by the product of the lease price and the number of airplanes. Total annualized aircraft cost is then computed as the sum across all categories of aircraft. The real stock of ground properties and equipment (GPE) is estimated using the perpetual inventory method. The annual cost of using GPE is computed by multiplying the real GPE stock by a GPE service price. The GPE service price is constructed using the method proposed by Christensen and Jorgenson (1969) which accounts for interest, depreciation, corporate income and property taxes, and capital gains or losses. Since the GPE costs are small relative to the costs of ¯ight equipment, these two categories of capital inputs are further aggregated into a single capital stock series using the translog multilateral index procedure. The price of capital input is then computed by dividing total capital cost by the aggregate capital input quantity index. The last category of inputs is materials. The materials input consists of all other inputs not included in any of the input categories discussed above. As such, materials cost is the catch-all cost category, and thus includes numerous items such as airport fees, sales commissions, passenger meals, employee travel, consultants, non-labour repair and maintenance expenses, stationery, and other purchased goods and services. The materials cost is computed by subtracting labour, fuel and capital input costs from the total operating cost reported in ICAO's Financial Data. Similar to the incidental output, it is necessary to construct a materials price index in order to include the materials input in our analysis. Since the materials cost also includes uncountable items and activities, the general price index used in constructing incidental output quantity index is used as a proxy for the materials input price. It is worthwhile to note that many studies on airline productivity have excluded the materials input, presumably because the materials input is dicult to measure. Since the materials cost accounts for 35±50% of the total cost depending on airlines, the exclusion of such an important input will obviously bias the empirical ®ndings on the relative eciency of airlines. Such a bias would be further aggravated as many airlines now outsource a variety of services, which should be accounted for in the materials cost. 3.3. Characteristics of airlines An airline's unit cost is in¯uenced by variations in the network, the operating and market conditions, and the regulatory and institutional environment, over which airlines have limited control. Therefore, it is necessary to examine the eects of these variables on the observed unit cost in order to make proper inferences about cost competitiveness. For this purpose the following variables are compiled: average stage length, average weight load factor, and revenue shares of freight and mail, non-scheduled services, and incidental services.{ In addition, the residual TFP index computed in Oum and Yu (1995) is used as an indicator of overall eciency, and is considered as an important determinant of carrier cost. The 1993 key statistics for the 22 sample airlines are reported in Table 1. The list is organized by continent and by revenue size. It shows, among other things, that Qantas, Singapore Airlines, Cathay Paci®c, Japan Airlines (JAL) and KLM Royal Dutch Airlines (KLM) are long-haul carriers, and Korean Air has heavy emphasis on freight services and non-airline businesses. Overall, many Asian carriers and some European carriers generate a relatively high proportion of their revenues from freight services as compared to the U.S. carriers. European carriers (except British Airways) have heavy emphasis on non-airline businesses as well. *Although the ICAO Financial Data reports fuel expense data, it does not report fuel price or quantity. Many airlines have provided the data on quantity series of fuel consumption upon our request. Fuel consumption for some U.S. carriers are also collected from The Airline Monitor. The fuel quantity data for Canadian carriers are collected from Statistics Canada 51-002 and 51-206. As was done in Windle (1991) a fuel quantity regression model was used to estimate fuel consumption for those airlines whose fuel consumption data are not available to us. The regression included available tonne-kilometres, aircraft kilometres, load factor, aircraft hours, aircraft departure, aircraft type and year eects. y The aircraft leasing price data were kindly supplied to us by Avmark Inc. { Although a number of points served is an important characteristic of an airline network, it is not included here because we are unable to obtain a consistent time series data especially for the non-U.S. carriers. It is noted, however, that some of the previous studies involving non-U.S. carriers, such as Good and Rhodes (1991), Good et al. (1993), and Distexhe and Perelman (1993), have not included this variable either.
412
Tae Hoon Oum and Chunyan Yu Table 1. Descriptive statistics of the sample airlines, 1993 Revenue No. of Average wage Stage length Load % Pax % Fre % Inc Rev/expb (mill U.S.$) employee ('000 U.S.$) (km) factor (%) reva(%) reva(%) reva(%)
North America American United Delta Northwest US Air Continental Air Canada Canadian
14,737 14,354 12,375 8448 6624 5086 2099 1947
91,773 78,105 69,537 42,439 45,986 36,191 18,184 15,208
56 62 70 63 61 39 36 36
1566 1631 1206 1395 866 1360 1459 1630
60 67 62 67 59 63 64 68
88 88 92 87 92 90 85 85
3 5 4 7 1 3 9 8
8 6 2 4 5 6 3 3
1.04 1.02 0.98 1.04 0.98 0.99 0.97 0.94
Australasia Japan Airlines All Nippon Singapore Airlines Korean Air Cathay Qantas Thai
8591 7226 3482 3342 2957 2688 2311
22,008 13,870 14,664 15,398 13,483 15,475 19,247
107 94 25 31 52 39 19
2396 1034 4214 1617 2990 4257 1524
65 62 71 65 70 70 67
79 89 77 58 75 77 73
13 5 19 20 17 10 12
5 5 3 18 5 6 14
0.97 1.01 1.09 1.09 1.10 1.06 1.09
Europe Lufthansa British Air Air France KLM SAS Swissair Iberia
9532 8694 6826 4321 3623 3540 2745
46,338 47,705 43,258 26,859 19,439 19,788 25,676
57 35 47 52 63 67 48
1071 1712 1637 1856 702 1243 1196
66 71 68 67 63 60 67
65 89 65 64 77 66 85
13 7 14 18 5 10 7
18 0.2 20 17 17 23 7
0.999 1.09 0.92 1.06 0.99 1.01 0.96
a b
Revenue shares for passenger, freight and incidental. Ratio of operating revenue over operating expenses based on ICAO data. 4. VARIABLE COST FUNCTION AND DECOMPOSITION OF THE UNIT COSTS
In this section, we present the estimated variable cost function, and use it to decompose the unit cost dierentials into sources. The results are used to identify the eects on unit cost of input prices, output mix, average stage length, and eciency, and to evaluate and compare cost competitiveness in the following section. 4.1. Cost function estimation The maximum likelihood parameter estimates, t-statistics and other summary statistics on the variable cost function estimation [eqn (1)] are reported in Table 2. The ®rst-order coecients for input prices indicate that at the mean data, labour and fuel inputs account for 32 and 15%, respectively, of the total variable cost. This leaves the materials input to account for 53% of the total variable cost. The ®rst-order coecient for the capital input variable is negative, implying a positive shadow value of capital input.* The stage length has a statistically signi®cant negative coecient, implying that variable cost decreases with stage length. The coecient for the eciency variable is negative and statistically signi®cant, indicating that ecient ®rms are likely to have considerably lower costs. The coecient for the time shift dummy for year 1993 indicates that the eciency of using variable input has improved by 3.3% between 1986 and 1993 due to industrywide technical progress. The negative coecients for %Incidental and %Non-Scheduled indicate that, other things being equal, carriers with a higher concentration on non-airline (incidental) businesses and non-scheduled services are expected to have lower variable costs. On the other hand, carriers with higher concentration on freight services may have a cost disadvantage.y *Many empirical studies on airline variable cost function report a positive ®rst-coecient for the capital stock variable, which implies a negative value of the shadow price of capital. See, for example, Gillen et al. (1985, 1990), and Caves et al. (1987) which was used extensively by Windle (1991). y At ®rst glance this appears to be contrary to our intuition that freight services require less input than passenger services. However, this result is plausible for the following reasons. Since cargo yields per RTK are far lower than passenger yield per RTK (average yield for our sample is U.S.$1.03 per RTK for passenger vs U.S$0.33 per RTK for freight), cargo output receives very low weight in aggregating outputs. Therefore, the amount of increase in output index caused by cargo output is relatively small as compared to the amount of increase in input cost caused by cargo output. If this is the case, % freight variable would have a positive coecient as in our case.
Cost competitiveness of airlines
413
Table 2. Variable cost function estimates Variable Constant (a0 ) Output (Y) Labour (W1 ) Fuel (Wf ) Capitala (uK) Stage length (Z) % Freight.(Rf ) % Non-sch.(Rn ) % Incidentl.(Ri ) Eciencyb (E) E.output (E Y) E.labour (E Wl ) E.fuel (E Wf ) E.capital (E uK) Eciency (E E) Labouroutput (W1 Y) Fueloutput(Wf Y) Capitaloutput (uK Y) Stageoutput (Z Y) Number of observations Log-likelihood function
Coefficient 8.109 1.123 0.318 0.153 ÿ0.106 ÿ0.305 0.055 ÿ0.007 ÿ0.040 ÿ1.138 0.050 0.090 0.100 ÿ0.063 0.723 ÿ0.054 0.059 0.017 0.028
T-value 1316.2 235.96 70.08 89.38 67.48 37.61 11.80 6.95 25.83 64.63 1.02 2.47 5.92 3.72 3.50 4.59 9.93 1.54 2.30
Variable Outputoutput (Y Y) Labourfuel (W Wf ) Labourcapital (W1 uK) Labourstage (W1 Z) Labourlabour (W1 W1 ) Fuelcapital (Wf uK) Fuelstage (Wf Z) Fuelfuel (Wf Wf ) Capitalstage (uK Z) Capitalcapital (uK uK) Stagestage (Z Z) 1987 (T87 ) 1988 (T88 ) 1989 (T89 ) 1990 (T90 ) 1991(T91 ) 1992 (T92 ) 1993 (T93 )
Coefficient
T-value
ÿ0.013 ÿ0.038 0.036 ÿ0.088 0.188 ÿ0.059 0.008 0.088 ÿ0.028 ÿ0.012 0.036 ÿ0.003 ÿ0.014 ÿ0.009 ÿ0.010 ÿ0.021 ÿ0.029 ÿ0.033
0.70 5.14 4.85 8.07 7.65 12.12 2.00 15.19 7.34 1.21 2.09 0.42 2.16 1.42 1.34 3.25 4.34 4.87
178 1572.28
All variables except time dummies are in natural log and normalized at mean; in addition to the cost shares equations, the equation for the ratio of depreciated capital cost to variable cost, which is equal to negative partial derivative dk Cv , is included in the regression. a Capital is capital stock multiplied by weight load factor. b Residual TFP index from Oum and Yu (1995) is used as Eciency.
4.2. Unit cost decomposition The 1993 unit cost (cost per unit of aggregate output) dierentials between each airline and American Airlines (AA) are decomposed into dierent sources using eqn (4), and the results for 1993 are summarized in Table 3. Column (1) lists the observed unit cost dierence expressed in percentage dierence of each airline's unit cost relative to American Airlines (AA). For example, in 1993, Delta's unit cost was 13.5% higher than AA's while Singapore Airline's unit cost was 30.6% lower than AA's. Asian carriers, except Japan Airlines (JAL) and All Nippon Airways (ANA), have substantially lower unit costs than the major U.S. carriers while most of the U.S. carriers, in turn, have substantially lower unit costs than the European carriers. KLM is an exception in that its unit cost is about 10% lower than Delta's. Columns (2)±(7) in Table 3 report on the decomposition of the unit cost dierences, that is, the contribution of each source to the observed unit cost dierence. Each of the entries listed under `Sources of dierence' is measured as the percentage dierence in total unit cost that would result if the only dierence between that particular airline and AA were that particular source. For example, for United, in column (4) under `labour price', 3.3 indicates that if the only dierence between United and AA in 1993 were the price of labour, then United's unit cost would have been 3.3% higher than AA's. Columns (2) and (3) show the eects of stage length and output mix (scheduled passenger, freight, non-scheduled, and incidental businesses). Variations in stage length alone account for a substantial portion of the observed unit cost dierences, especially for the carriers at extreme ends of the scale. For example, if other things are equal, Singapore Airlines' and Qantas' unit cost would be 23 and 25% lower than AA's, respectively, because of their long stage length. SAS and US Air, on the other hand, would have 25 and 18% higher unit cost than AA, respectively, again if other things were equal. Output mix, in general, has only limited eects on the observed unit cost. The only noticeable exception is British Airways (BA). BA has very little incidental and nonscheduled services, thus its unit cost would be 11% higher than AA's if other things were equal. Columns (4)±(6) of Table 3 are the percentage dierences in unit costs between each airline and AA attributable to, respectively, the dierences in labour price, other input prices, and all input prices together. It is noted that Asian carriers, with the exception of JAL and ANA, enjoyed signi®cant cost advantage relative to AA due to their lower labour and non-labour input prices. For
414
Tae Hoon Oum and Chunyan Yu Table 3. Unit cost decomposition and cost competitiveness, 1993 (% above and below AA's unit cost) Observed unit cost difference (1)
Sources of difference Firm characteristics
Input prices
Efficiency
Stage Output mix Labour Other inputs All inputs (2) (3) (4) (5) (6)
(7)
Cost competitiveness (8)=(6)+(7)
North America America United Delta Northwest Continental US Air Air Canada Canadian
0.0 ÿ1.7 13.5 ÿ3.7 ÿ10.7 40.9 12.7 3.6
0.0 ÿ1.2 7.7 3.3 3.8 17.6 1.9 ÿ1.1
0.0 0.2 3.1 2.1 ÿ2.6 ÿ4.2 3.1 1.5
0.0 3.3 7.7 3.9 ÿ10.7 3.0 ÿ12.4 ÿ12.3
0.0 0.4 ÿ0.3 0.7 ÿ0.7 ÿ0.4 0.9 3.7
0.0 3.7 7.4 4.6 ÿ11.4 2.6 ÿ11.5 ÿ8.6
0.0 ÿ3.8 ÿ5.6 ÿ9.9 ÿ0.7 17.4 19.9 13.7
0.0 ÿ0.1 1.8 ÿ5.3 ÿ12.1 19.9 8.5 5.0
Australasia Japan Airlines All Nippon Singapore Airlines Korean Air Cathay Qantas Thai
50.1 114.8 ÿ30.6 ÿ25.2 ÿ18.3 ÿ24.6 ÿ20.8
ÿ13.0 13.0 ÿ23.0 ÿ0.8 ÿ17.4 ÿ24.9 0.6
3.0 3.9 6.8 0.1 4.4 0.5 ÿ0.1
22.6 18.8 ÿ16.9 ÿ14.7 ÿ2.1 ÿ8.9 ÿ22.4
15.9 21.5 ÿ3.3 ÿ9.0 ÿ4.2 ÿ0.1 ÿ29.8
38.4 40.4 ÿ20.3 ÿ23.8 ÿ6.4 ÿ8.9 ÿ52.1
14.3 23.1 3.9 0.8 2.6 11.6 42.9
52.7 63.5 ÿ16.3 ÿ22.9 ÿ3.8 2.7 ÿ9.3
19.3 29.2 21.9 3.3 81.5 46.4 36.9
ÿ1.2 11.2 ÿ2.4 ÿ6.2 25.0 5.4 7.6
ÿ0.8 ÿ2.0 11.0 1.7 ÿ4.7 ÿ2.2 2.8
ÿ5.2 0.6 ÿ12.8 0.6 4.4 9.4 ÿ4.9
14.0 16.3 9.9 15.5 21.0 26.1 10.0
8.8 16.8 ÿ2.9 16.0 25.4 35.5 5.1
12.4 3.8 10.2 ÿ5.3 17.0 2.8 16.4
21.2 20.6 7.3 10.7 42.4 38.3 21.5
Europe Air France Lufthansa British Airways KLMa SAS Swissaira Iberia a
1992.
example, over 90% of Korean Air's unit cost advantage over AA came from its lower input prices. On the other hand, JAL's high labour price caused a 23% unit cost disadvantage while their higher non-labour input prices (fuel, capital and materials) were responsible for a further 16% unit cost disadvantage relative to AA. Major U.S. carriers, except Continental, incurred 3±8% higher cost relative to AA because of higher labour prices. Continental's lower labour price gave it a 11% cost advantage over AA. The two Canadian carriers, Air Canada and Canadian Airlines International (CAI), also had approx. 12% cost advantage over AA due to lower labour prices. The eects of other input prices are very small for North American carriers, except CAI which had a 3.7% cost disadvantage relative to AA due to a slightly higher non-labour price. All European carriers suered substantial cost disadvantage relative to AA due to signi®cantly higher nonlabour input prices, ranging from 10% for BA and Iberia to 26% for Swissair. The eects of labour prices, on the other hand, vary across airlines. For example, B A enjoyed a 13% cost advantage over AA due to lower labour price, while Swissair had a 10% cost disadvantage due to higher labour price. Column (7) lists the contribution of eciency to the unit cost dierence. The results show that higher eciency helped to reduce unit cost for most of the major U.S. carriers (except US Air) and KLM in 1993. The reduction ranges from 1% for Continental to 10% for Northwest. However, all other carriers in our sample would have had unit cost disadvantages of varying degrees relative to AA because of their lower productive eciency. For example, ineciency alone accounted for a 20% higher unit cost for Air Canada, which negated the 12% cost advantage it had over AA from lower input prices. Similarly, Thai Airways' ineciency would have led to a 43% higher unit cost than AA were it not for the favourable impact of its signi®cantly lower input prices. ANA's lower eciency level led to a further 23% cost disadvantage in addition to the 40% cost disadvantage caused by higher input prices. In summary, except JAL and ANA, the Asian carriers including Singapore, Cathay, Korean Air, Qantas, and Thai Airways enjoy substantial unit cost advantages relative to the U.S. carriers.
Cost competitiveness of airlines
415
JAL and ANA have a substantial unit cost disadvantage vis-aÁ-vis the U.S. mega carriers. Most of the unit cost advantages of Asian carriers come from their lower input prices including labour prices. KLM enjoys a unit cost advantage among the European carriers while it has a unit cost disadvantage relative to the Asian carriers (except JAL and ANA) and some U.S. carriers. This result is consistent with the ®ndings of Bureau of Industry Economics (1994) and European Commission (1994). 5. UNIT COST COMPETITIVENESS
As discussed earlier, the observed unit cost dierence does not re¯ect the true comparative cost competitiveness between airlines because these airlines have dierent network and operating characteristics. What one needs to ask is how these carriers would be able to compete in a given market such as the trans-Paci®c and trans-Atlantic markets. A low system-wide unit cost for an airline with heavy concentration on incidental services and with a long average stage length may not constitute cost competitiveness in a speci®c market. When an airline competes in a given market, particularly in an inter-continental market, what is relevant is the marginal cost of providing a given level of services in that market. What determines cost competitiveness is the input prices the airline pays and how eciently the airline produces and markets their services. Therefore, a cost competitiveness (CC) indicator is constructed by summing the eects of input prices and eciency reported in column (6) and (7) of Table 3. Since the unit cost decomposition disentangles the eects of output mix and stage length from the eects of input prices and eciency, this indicator approximates the `true' comparative cost competitiveness of the airlines. Column (8) of Table 3 presents the cost competitiveness indicator for 1993. The cost competitiveness indicator is measured in terms of percentage above(ÿ) or below (+) that of AA. A negative number indicates that the airline is cost competitive relative to AA, while a positive number indicates vis-aÁ-vis. Continental (CO) enjoyed an over 12% cost competitiveness relative to AA in 1993. Most of its cost competitiveness came from its lower labour price. Northwest enjoyed a 5% cost competitiveness over AA. Although its high eciency gave it a 10% cost advantage, its higher input prices reduced its cost competitiveness. United and Delta were similar to AA in terms of cost competitiveness (with CC at ÿ0.1 and 1.8%, respectively). Although their higher eciency gave them a 3.8 and 5.6% cost advantage, respectively, their higher input prices created a 3.7% cost disadvantage for United and 7.4% for Delta. US Air was 20% less cost competitive relative to AA. This consists of 17 and 3% due to their lower eciency and higher input prices, respectively. Air Canada (AC) was about 8.5% less cost competitive relative to AA because of its substantially lower eciency. CAI was about 5% less cost competitive relative to AA, also due to its lower eciency level. Both AC and CAI's cost advantage from lower labour prices were overpowered by the eects of their lower eciency. Among the Asian carriers, Singapore, Korean, Cathay and Thai enjoyed higher cost competitiveness relative to AA, by factors of 16, 23, 4 and 9%, respectively. Their cost competitiveness came entirely from their signi®cantly lower input prices, while their relatively lower eciency level diminished their cost advantages. Qantas was slightly less cost competitive (2.7%) than AA because its cost advantage from lower input prices was more than o set by its lower eciency level. JAL and ANA were signi®cantly less cost competitive than AA because of the combined eects of higher input prices and lower eciency. BA was about 7% less cost competitive relative to AA due to higher non-labour prices and lower eciency. KLM was 10% less cost competitive relative to AA. This was entirely due to the cost disadvantage caused by higher input prices, albeit its higher eciency helped alleviate some of this cost disadvantage. Swissair was 38% less cost competitive relative to AA. High input prices were the main cause of this cost disadvantage. Lufthansa was 21% less cost competitive relative to AA. Their cost disadvantage was mostly caused by high input prices. Air France was also about 21% less cost competitive relative to AA. This resulted from a combined eect of high non-labour input prices and low eciency. SAS and Iberia were, respectively, 42 and 32% less cost competitive relative to AA. Their cost disadvantages resulted from high input prices and low eciency level.
416
Tae Hoon Oum and Chunyan Yu
Figures 1±3 show changes over time in cost competitiveness and its components for each of the sample airlines. Note that the comparisons are made between each of the airlines and AA in the same year, therefore the changes also partly re¯ect the changes in AA's unit costs. Although there had been considerable ¯uctuations over the years in the degrees of the relative cost competitiveness (and the eects of its components) for the airlines, the relative rankings among the airlines remained fairly stable. The most noticeable changes are with BA and Qantas, both airlines had considerably enhanced their relative cost competitive positions mainly through improvement in eciency. On the other hand, Air France's cost competitive position had deteriorated somewhat since 1990 as a result of higher input price and lower eciency. It is also noted that Swissair and KLM's eciency level had improved signi®cantly during the period, but their input prices had also risen signi®cantly. The eects of the two opposite forces appear to counteract each other. As a result, Swissair and KLM's cost competitive positions did not show any signi®cant changes. Overall, European carriers are less cost competitive relative to the U.S. airlines. This is caused by both higher input prices and lower eciency than their U.S. counterparts. The Australasian carriers, with the exception of JAL and ANA, are more cost competitive than the U.S. airlines. However, their cost competitiveness is entirely the result of low input prices rather than high eciency.
Fig. 1. Cost competitiveness (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers.
Cost competitiveness of airlines
417
Fig. 2. Eects of input prices (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers. 6. THE EFFECTS OF EXCHANGE RATE ON AIRLINE COST
So far in our discussion, all costs are expressed in a common currencyÐU.S. dollar. Since the airlines compete in international markets, it is common practice to compare their operating results in terms of a common currency. However, exchange rate ¯uctuations have had non-negligible impact on inter-country cost dierentials (denominated in a common currency). In this section, the eects of exchange rate ¯uctuations on carrier cost are examined via brief case studies of JAL and Lufthansa. Changes in Japanese and German exchange rates have played an important part in explaining the ability of JAL and Lufthansa to compete in international markets. Figures 4 and 5 plot the changes in JAL's labour and non-labour prices*, respectively, in relation to Japan's exchange rate. Japanese yen appreciated about 35% against U.S.$ between 1986 and 1993. Therefore, both JAL's labour and non-labour input prices in U.S.$ terms have increased substantially relative to AA. The pattern of input price changes in Japanese yen, however, tells a *Input prices are expressed in terms of percentage of AA's input prices.
418
Tae Hoon Oum and Chunyan Yu
Fig. 3. Eects of eciency (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers.
Fig. 4. Labour price vs exchange rate: Japan Airlines (base: AA=1.00).
Cost competitiveness of airlines
419
Fig. 5. Non-labour price vs exchange rate: Japan Airlines (base: AA=1.00).
dierent story (as shown by the input prices converted using constant 1986 exchange rate). In fact, between 1986 and 1993 JAL's non-labour input prices decreased by about 24% in terms of Japanese yen, but increased by 20% in terms of U.S.$. Similar observations can be made for Lufthansa. Figures 6 and 7 plot the changes in Lufthansa's input prices in relation to the Deutsche mark. The Deutsche mark appreciated by 24% between 1986 and 1993. In Deutsche mark, Lufthansa's input prices have declined since 1990 relative to AA (as shown by the input prices measured in constant 1986 exchange rates). In U.S.$, however, there is no obvious downward trend. These dierencesÐbetween calculation in U.S.$ vs each country's own currencyÐare due to changes in exchange rates. Figures 8 and 9 plot changes in JAL's and Lufthansa's unit costs and exchange rates. In U.S.$, JAL's unit cost increased by 39% between 1986 and 1993, while its unit cost in Japanese yen decreased by 12% (at constant 1986 exchange rate). Similarly, Lufthansa's unit cost increased by 16% in 1993 from 1986's level in U.S.$ terms, but decreased by about 10% in terms of its home currency (at constant 1986 exchange rate). Exchange rate ¯uctuation is entirely beyond a carrier's control, but it has had decisive eects on JAL and, to a lesser degree, on Lufthansa's cost competitiveness. Figures 8 and 9 show that in 1993 JAL and Lufthansa's unit costs would have been 4 and 2% lower than that of AA, respectively, had there been no change in exchange rate between 1986 and 1993. Since our sample airlines are all international carriers, their cost competitiveness are all aected by exchange rate ¯uctuation, albeit to dierent degrees. The eects could be both positive and negative. Home currency appreciation may lower cost competitiveness of its carriers in the short run, but might force the carriers to improve eciency. On the other hand, home currency depreciation generally improves cost competitiveness of its carriers at least in the short run, but may lead to decreased productivity of the home carriers caused by reduced competitive pressure.
Fig. 6. Labour price vs exchange rate: Lufthansa (base: AA=1.00).
420
Tae Hoon Oum and Chunyan Yu
Fig. 7. Non-labour price vs exchange rate: Lufthansa (base: AA=1.00).
Fig. 8. Unit cost vs exchange rate: Japan Airlines (base: AA=1.00).
Fig. 9. Unit cost vs exchange rate: Lufthansa (base: AA=1.00).
7. SUMMARY AND CONCLUDING REMARKS
This study measures and compares cost competitiveness of 22 major international airlines. Cost competitiveness depends on input factor prices and eciency. Observed unit cost dierentials between airlines, however, are also in¯uenced by the eects of network and output attributes
Cost competitiveness of airlines
421
which must be removed in order to make meaningful comparisons across ®rms and/or over time within a ®rm. Therefore, a translog variable cost function is estimated and used to decompose the total unit cost dierentials into potential sources: stage length, output mix, input prices, and eciency. The cost decomposition results are then used to develop a cost competitiveness (CC) indicator after removing the eects of stage length and output mix. The study also examines the eects of exchange rate ¯uctuation on input prices and unit cost via brief case studies of JAL and Lufthansa. The main results are summarized as follows: (a) The carriers in Australasia (except JAL and ANA) are generally more cost competitive than the major U.S. carriers (Northwest and Continental are as cost competitive as some of the Australasian carriers). Lower input prices are the dominating reason for their cost competitiveness, and their cost competitiveness relative to AA has decreased over time (except Qantas and Korean Air). (b) JAL and ANA are signi®cantly less cost competitive than AA mainly because of their higher input prices. (c) Major European carriers were 7 (BA)±42% (SAS) less cost competitive than AA in 1993 because of the combined eects of higher input prices and lower eciency level. However, European carriers' eciency performance improved noticeably relative to AA during the 1986±93 period; (d) In 1993, AA, United and Delta are similar in cost competitiveness, while Northwest and Continental enjoyed, respectively, 5 and 12% cost competitiveness over AA. US Air was least cost competitive among the North American carriers, due to its lower eciency and slightly higher labour price. (e) Air Canada was about 8.5% less cost competitive relative to AA due to its lower eciency level. CAI was about 5% less cost competitive relative to AA also due to its lower eciency level. Their cost advantage from lower labour price is cancelled out by the eects of lower eciency. (f) Exchange rate ¯uctuation has had signi®cant impact on some carriers' costs. For example, JAL, and, to a lesser degree, Lufthansa's less cost competitive position is caused to a large extent by their home currency appreciation. In 1993, JAL's and Lufthansa's unit costs would have been 4 and 2%, respectively, lower than AA's had there been no change in their respective exchange rates between 1986 and 1993. Overall, input prices together (including exchange rates) have been a more important factor than eciency for determining a carrier's cost competitive position in the past. However, the importance of input prices is likely to diminish over time as airlines increase their global sourcing of labour, materials, services and other inputs, and as input prices in the developing and the Newly Industrialized Countries (NIC) continue to rise faster than in the developed countries. As liberalization of the airline industry continues, eciency will become progressively more important in determining cost competitiveness of an airline. This paper reports on the ®rst phase of our investigation on airline cost competitiveness. There are two points that need to be addressed in our future research. First, this study uses the traditional econometric method in estimating the variable cost function. Although an eciency variable is incorporated in the estimation, the resulting cost function is still an `average cost' function which may not truly re¯ect the cost frontier. In our future research, therefore, the recently developed stochastic frontier method will be applied to re®ne our empirical results on the cost function estimation, and to further investigate the causal relationship between competitiveness and eciency. Secondly, this paper illustrates the importance of the impact of exchange rates on airline cost via brief case studies on JAL and Lufthansa. Future research will incorporate a systematic investigation of the eects of exchange rates on international air transport competition. One possible way to accomplish this is to adjust the decomposition formula based on the concept of a fundamental equilibrium exchange rate which was developed by Williamson (1985), and applied by Fuss and Waverman (1992) in a cost and productivity study of automobile production. The adjusted decomposition will be able to isolate changes in cost dierences due to exchange rate ¯uctuations from those due to relative movements in factor prices and eciency within each airline.
422
Tae Hoon Oum and Chunyan Yu
AcknowledgementsÐThe authors would like to thank the seminar participants at University of British Columbia, Osaka University, Nanzan University, Canadian Transportation Research Forum Conference, Japan Airlines Headquarters, Korean Air Headquarters and Korea Transport Institute for helpful comments and suggestions. The research grant supports from the Social Science and Humanities Research Council of Canada and The Japan Ministry of Education via Osaka University are gratefully acknowledged.
REFERENCES Baltagi, B. H., Grin, J. M. and Rich, D. P. (1995) Airline deregulation: the cost pieces of the puzzle. International Economic Review 36(1), 245±258. Bauer, P. W. (1990) Decomposing TFP growth in the presence of cost ineciency, nonconstant returns to scale, and technological progress. The Journal of Productivity Analysis 1, 287±299. Bureau of Industry Economics (BIE) (1994) International Performance Indicators: Aviation, Vol. 59. Canberra, Australia. Caves, D. W. and Christensen, L. R. (1988) The importance of economies of scale, capacity utilization, and density in explaining interindustry dierences in productivity growth. The Logistics and Transportation Review 24(1), 3±32. Caves, D. W., Christensen, L. R. and Diewert, W. E. (1982) Multilateral comparisons of output, input, and productivity using superlative index numbers. Economic Journal 92, 73±86. Caves, D. W., Christensen, L. R. and Tretheway, M. W. (1984) Economies of density versus economies of scale: why trunk and local service airline costs dier. Rand Journal of Economics 15, 471±489. Caves, D. W., Christensen, L.R., Tretheway, M. W. and Windle, R. J.(1987) An assessment of the eciency eects of U.S. airline deregulation via an international comparison. In Public Regulation: New Perspectives on Institutions and Policies, ed. E. E. Bailey, pp. 285±320. MIT Press, Cambridge, MA. Christensen, L. R. and Jorgenson, D. W. (1969) The measurement of U.S. real capital input, 1929±1967. The Review of Income and Wealth 15(1), 293±320. Distexhe, V. and Perelman, S. (1993) Technical eciency and productivity growth in an era of deregulation: the case of airlines. Paper presented at the Third European Workshop on Eciency and Productivity Measurement, CORE, Belgium, October. European Commission (1994) Expanding Horizons: Civil Aviation in Europe, an Action Programme for the Future. Report by the Comite des Sages for Air Transport, Brussels, Belgium. Encaoua, D. (1991) Liberalizing European airlines: cost and factor productivity evidence. International Journal of Industrial Organization 9, 109±124. ESG Aviation Services, The Airline Monitor, various issues. Ponte Vedra Beach, Florida, U.S.A. Fuss, M. A. and Waverman, L. (1992) Cost and Productivity in Automobile Production: the Challenge of Japanese Eciency. Cambridge University Press, New York. Gillen, D. W., Oum, T. H. and Tretheway, M. W. (1985) Airline Cost and Performance: Implications for Public and Industry Policies. Centre for Transportation Studies, University of British Columbia, Vancouver, B.C., Canada. Gillen, D.W., Oum, T.H. and Tretheway, M.W. (1990) Airline cost structure and policy implications. Journal of Transport Economics and Policy 24(2), 9±34. Good, D. H. and Rhodes, E. L. (1991) Productive eciency, technological change and the competitiveness of U.S. airlines in the Paci®c Rim. Journal of the Transportation Research Forum 31(2), 347±358. Good, D. H., Nadiri, M.I., RoÈller, L. H. and Sickles, R. C. (1993) Eciency and productivity growth comparisons of European and U.S. air carriers: a ®rst look at the data. The Journal of Productivity Analysis 4, 115±125. Good, D. H., RoÈller, L.H. and Sickles, R. C. (1995) Airline eciency dierences between Europe and the US: implications for the pace of EC integration and domestic regulation. European Journal of Operational Research 80(1), 508±518. International Civil Aviation Organization (ICAO) (1986±1993) Digest of Statistics, various annual series. Montreal, Canada. Oum, T. H. and Yu, C. (1995) A productivity comparison of the world's major airlines. Journal of Air Transport Management 2(3/4), 181±195. Oum, T. H. and Zhang, Y. (1991) Utilization of quasi-®xed inputs and estimation of cost functions. Journal of Transport Economics and Policy 25(2), 121±134. Oum, T. H. and Zhang, Y. (1995) Competition and allocative eciency: the case of competition in the U.S. telephone industry. Review of Economics and Statistics 77(1), 82±96. Ray, S. and Hu, X. (1993) A nonparametric decomposition of the malmquist productivity index: a study of airlines' data. Paper presented at the Third European Workshop on Productivity, CORE, Belgium, October. Summers, R. and Heston, A. (1991) The Penn World Table (Mark 5): an expanded set of international comparisons, 1950± 1988. Quarterly Journal of Economics 106(2), 327±368. Tofallis, C. (1996) Input eciency pro®ling: an application to airlines. Working paper UHBS 1996:21, The Business School, University of Hertfordshire, U.K. Williamson, J. (1985) The Exchange Rate System. Institute for International Economics, Washington, DC. MIT Press. Cambridge, MA. Windle, R. J. (1991) The world's airlines: a cost and productivity comparison. Journal of Transport Economics and Policy 25(1), 31±49. Xu, K., Windle, R., Grimm, C. and Corsi, T. (1994) Re-evaluating returns to scale in transport. Journal of Transport Economics and Policy 28(3), 275±286.
Pergamon PII: S0965-8564(98)00007-X
COST COMPETITIVENESS OF MAJOR AIRLINES: AN INTERNATIONAL COMPARISON TAE HOON OUM* and CHUNYAN YU
Faculty of Commerce and Business Administration, The University of British Columbia, Vancouver, B.C., V6T 1Z2 Canada (Received 15 January 1997; in revised form 2 January 1998) AbstractÐThis paper compares unit cost competitiveness of the world's 22 major airlines over the 1986±93 period. First, a unit cost index for aggregate output is computed via a multilateral index procedure. A translog variable cost function is estimated and used to decompose the unit cost dierentials into potential sources: input prices, network and output attributes, and eciency. The results of the unit cost decomposition are used to construct a cost competitiveness indicator after removing the eects of network and output attributes. Our results for 1993 are: (a) Asian carriers (except Japan Airlines and All Nippon Airways) were generally more cost competitive than the major U.S. carriers, mostly due to their substantially lower input prices; (b) Japan Airlines and All Nippon Airways were over 50% less cost competitive than American Airlines mainly because of their high input prices; (c) major European carriers were 7% (British Airways)±42% (Scandinavian Airlines Systems) less cost competitive than American Airlines, because of higher input prices and lower eciency; (d) among the U.S. carriers, American Airlines, United Airlines and Delta were similar in cost competitiveness, while Northwest and Continental enjoyed, respectively, 5 and 12% cost competitiveness over American Airlines; (e) exchange rate ¯uctuation has had considerable eects on the cost competitive position of Japan Airlines and Lufthansa. # 1998 Elsevier Science Ltd. All rights reserved Keywords: airlines, cost competitiveness, productive eciency, decomposition 1. INTRODUCTION
Since the mid-1980s, the world airline industry has experienced a period of signi®cant structural, institutional and regulatory changes. Many countries have deregulated their domestic airline industries. A series of very liberalized bilateral agreements have been signed between the U.S. and countries such as the Netherlands, Belgium, Austria, Switzerland, Finland, and Denmark. Open skies continental blocs are being formed both in Europe and in North America. A movement to create a liberalized continental bloc is under way in Asia Paci®c. As a result, more airlines are being exposed to the pressures of the market-place as the widely-based deregulation and liberalization process advances. The increased competition and recent recession have led to severe and widespread losses in the international airline industry, and forced carriers to undertake major restructuring in order to improve eciency and reduce costs. As the competition intensi®es, the ultimate ability of a carrier to survive and prosper in the globalizing airline industry depends greatly on its cost competitiveness. What constitutes cost competitiveness of an airline? In the simplest terms, an airline is cost competitive if its unit cost (average cost) is lower than that of its competitors on a sustainable basis. An airline may have a lower unit cost than its competitors because it is more ecient, pays less for its inputs, or both. That is, airlines' cost dierentials are determined by dierences in factor prices (including exchange rates) and productive eciency. Knowledge about existing levels and sources of cost dierentials are essential for analyzing public policies and carrier strategies designed to enhance airlines' competitive position. It is important to measure these dierentials and to ascertain their sources. The policies/strategies one would pursue depend on the relative importance of sources of the cost dierentials. A number of studies have focused on airline productivity and eciency. These include Caves et al. (1987), Gillen et al. (1985, 1990), Bauer (1990), Encaoua (1991), Good et al. (1993, 1995), *Author for correspondence. Fax: 001 604 822 8320; e-mail:[email protected]
407
408
Tae Hoon Oum and Chunyan Yu
Ray and Hu (1993), Distexhe and Perelman (1993) and Tofallis (1996). Few studies, however, have examined the issue of airline cost competitiveness. Windle (1991) attempted to attribute the differences in unit costs between carriers to various sources, namely output scale, input prices, operating characteristics, non-optimal capital investment, technical eciency, government ownership and ®rm-speci®c eects. His study was mainly based on the 1983 annual data for 14 U.S. and 27 non-U.S. airlines. Good and Rhodes (1991) examined three aspects of competitiveness and profitability of airlines in the Paci®c region: input prices, output prices and productive eciency, using a panel of 37 airlines over the 1976±86 period. A recent study by Baltagi et al. (1995) applied a translog variable cost function to a panel data of 24 U.S. airlines over the period of 1971±86 to analyze cost changes in the U.S. airline industry in the pre- and post-deregulation era and to identify the cost changes due to technical change, economies of scale and density, and input prices. To our knowledge, no previous study has focused on measuring and comparing cost competitiveness of the world's major airlines. This paper investigates issues related to cost competitiveness of airlines by analyzing the performance of 22 major international airlines during the 1986±93 period. Our analysis focuses on identifying the sources of airlines' cost competitiveness. Speci®cally, a translog variable cost function is estimated, and its results are used to decompose the observed unit cost dierentials among airlines into potential sources: attributes of the ®rm (network characteristics and output mix), input prices, and eciency. The results of the unit cost decomposition are used to assess and compare the cost competitiveness across the sample airlines. In addition, it is believed that exchange rate ¯uctuation has had non-negligible impact on some carriers' cost, therefore, the eect of exchange rate ¯uctuation on factor prices, consequently on unit cost, is also examined via two brief case studies. The following section describes the methodology used for cost function estimation and decomposition. Section 3 then provides a description of the data base. The results of cost function estimation and unit cost decomposition are presented in Section 4, while Section 5 examines airline cost competitiveness. The eect of exchange rates on cost competitiveness is examined in Section 6. The summary and concluding remarks are given in the ®nal section. 2. METHODOLOGIES
This section describes the methodologies for estimating a neoclassical cost function and for decomposing the unit cost dierentials between airlines into potential sources. Since capital input is not always in equilibrium in the airline industry, we follow Caves et al. (1984) and Gillen et al. (1990), and employ the following translog variable cost function to re¯ect the short run cost minimization process (with the usual restrictions on symmetry and linear homogeneity of cost in input prices):* ln VC a0 t at Tt by ln Y i i ln Ri i bi ln Wi bk ln uK 1 1 1 be ln E c ln Z dyy ln y2 i j dij ln Wi ln Wj dkk ln uK2 2 2 2 1 1 2 2 dee ln E dzz ln Z i eyi ln Y ln Wi eyk ln Y ln uK eye ln Y ln E 2 2 eyz ln Y ln Z i fki ln uK ln Wi fke ln uK ln Z i gei ln E ln Wi i gzi ln Z ln Wi
1
where VC is the cost of variable inputs; Y is the aggregate output index; W is a vector of input prices which consists of labour, fuel, and purchased services and materials inputs; K is capital stock which is treated as ®xed in the short run; u is utilization of capital stock (in this case, weight load factor); Ri 's are the revenue shares of freight and mail, non-scheduled services, and incidental servicesy; Z is average stage length; and E is an eciency index computed in Oum and Yu (1995). *Readers who are not familiar with the neoclassical translog cost functions and applications to transportation industries, please refer to, for example, Caves et al. (1984) or Gillen et al. (1990). y The revenue share variables (re¯ecting output mix), Ri 's, are incorporated only in the ®rst-order terms in order to keep the cost function simple.
Cost competitiveness of airlines
409
at 's are coecients associated with year dummy variables (Tt 's) in order to capture the eects of shifts in technical eciency over time. a0 , b's, 's, c, d's, e's, f's, g's are coecients of the translog variable cost function to be estimated.* Note that the variable cost function (1) includes a variable E for eciency. E is the index of residual total factor productivity estimated by Oum and Yu (1995). It re¯ects the overall eciency level of the airlines after removing the eects of factors beyond managerial control, including average stage length and output mix. By including E in the cost function estimation, we recognize the fact that some airlines fail to be on the production frontier, that is, some ®rms are more ecient than others. Once we recognize this, failure to include an eciency indicator may lead to misspeci®cation of the model, and thus bias parameter estimates of the cost function. Therefore, we essentially use a two-step procedure to estimate the cost function. In the ®rst stage, an eciency index is estimated, and in the second stage, the estimated eciency index is used as an explanatory variable in the cost function estimation. In this way, we can explicitly examine eciency eects on airline cost. The following cost minimizing variable input cost share equations can be derived by applying Shephard's lemma to the variable cost function (1): Si
@ ln VC bi j dij ln Wj eyi ln Y fki ln uK gei ln E gzi ln Z @ ln Wi
2
To improve the eciency of estimation, it is customary to estimate the translog variable cost function (1) jointly with the variable input cost share eqn (2). In order to improve the eciency of estimation further, Oum and Zhang (1991, 1995) proposed to add the following expression for the shadow value of capital stock: ÿ Ck @ ln VC ÿ ÿ bk dkk ln uK eyk ln Y i fki ln Wi fke ln E fkz ln Z VC @ ln uK
3
where Ck is the depreciated capital cost which is approximated by the total capital cost multiplied by utilization rate. Equation (3) is basically the ®rst order condition for short-run total cost minimization which endogenizes the capacity utilization. Following Oum and Zhang (1991), we estimate the translog variable cost function (1), the cost share eqnsy (2), and the shadow price of capital input eqn (3) jointly as a system of multivariate equations using a Maximum Likelihood method. Drawing on the properties of a translog variable cost function, Caves and Christensen (1988) and Fuss and Waverman (1992) showed that the (total) unit cost dierential (including capital costs) between any two observations, 1 and 0, can be decomposed into various sources using the following formula: h i c1 ÿ c0 S 1=2 d 1y Cv d 0y Cy ÿ 1 Y 1 ÿ Y 0 + size S 1=2 d 1k Cv d 0k Cv K 1 ÿ K 0 1 1 ÿ S K ÿ K 0 ÿ Y 1 ÿ Y 0 S 1=2 d 1r Cv d 0r Cv R1 ÿ R 0 output mix + 4 S 1=2 d 1w Cv d 0w Cv W 1 ÿ W 0 input prices 1 ÿ S W1k W0k S 1=2 d 1z Cv d 0z Cv Z 1 ÿ Z 0 operating characteristics S 1=2 d 1t Cv d 0t Cv t 1 ÿ t 0 time effects S 1=2 d 1e Cv d 0e Cv E 1 ÿ E 0 efficiency *Please note that subscript is given to each regression coecient for ease of relating the coecient to the variable. y To avoid singularity of variance±covariance matrix, the materials cost share equation was dropped from the estimation. It is well known that the maximum likelihood estimates are invariant to the choice of the share equation dropped.
410
Tae Hoon Oum and Chunyan Yu
where S denotes the average share of variable cost in the total cost for observations 1 and 0, and dxiCv denotes the partial derivative of the variable cost for observation i with respect to variable x. For ease of presentation, this paper uses American Airlines (AA) as the benchmark ®rm against which to compare other airlines. 3. THE DATA
Our sample data consists of annual observations on 22 major international airlines over the 1986±93 period. The primary data are compiled mainly from the Digest of Statistics published by the International Civil Aviation Organization (ICAO), in particular, the annual series of Trac, Fleet & Personnel and Financial Data. Additional data were obtained directly from airline companies, The Airline Monitor, IATA publications, Statistics Canada publications, and a number of other sources. Airline annual reports were used to supplement, cross-check, and correct errors in the ICAO data. The airline companies were contacted for clari®cation when the various sources of data could not be reconciled. Estimation of the variable cost function (1) and decomposition of the unit cost into sources using eqn (4) require detailed data on outputs, network and output attributes, amount and utilization of capital stock, and input prices. 3.1. Outputs Five categories of airline outputs are considered: scheduled passenger service (measured in revenue-tonne-kilometres or RTK), scheduled freight service (measured in RTK), mail service (measured in RTK), non-scheduled services (measured in RTK), and incidental services (non-airline businesses). Incidental services include a wide variety of non-airline businesses such as catering services, ground handling, aircraft maintenance and reservation services for other airlines, sales of technology, consulting services, and hotel business. These are non-core activities of an airline, but use up part of the inputs reported in the data sources which are used by most researchers. Many previous researchers have ignored the presence of non-airline businesses, by failing to include the incidental outputs. Our data show that revenues from non-airline businesses account for up to 30% of total operating revenues for some airlines, with an average of 8% for the airlines included in this study. Therefore, omission of the incidental output, without excluding the inputs used to produce them, would bias performance measures in favour of the airlines who do little non-core businesses. A quantity index is constructed for the incidental output in order to include the incidental services in our analysis. The index is computed by de¯ating incidental revenues by a general price index constructed using the Purchasing Power Parity (PPP) index for GDP obtained from the Penn World Table (Summers and Heston, 1991) and the U.S. GDP de¯ator. The PPP index adjusts for changes in market exchange rates and changes in real price levels of various countries relative to the U.S., and the US GDP de¯ator ensures that the quantity index is comparable over time. A multilateral output index is formed by aggregating the ®ve categories of outputs using the following multilateral index procedure proposed by Caves et al. (1982).
ln Yk
X Rik R i i
2
ln
Yik Y~ i
5
where Yk is the aggregate output index for observation k; Yik is the output i for observation k; R ik is the revenue share of output i for observation k; R i is the arithmetic mean of the revenue share of output i over all observations in the sample; and Y~ i is the geometric mean of output i over all observations. 3.2. Inputs Five categories of inputs are considered: labour, fuel, ¯ight equipment, ground property and equipment (GPE), and materials. The price of labour input is measured by the average
Cost competitiveness of airlines
411
compensation per employee (including bene®ts). Fuel price is obtained by dividing total fuel cost by gallons of fuel consumed.* For ¯ight equipment, a ¯eet quantity index is constructed by aggregating dierent types of aircraft using the translog multilateral index procedure proposed by Caves et al. (1982). The leasing price seriesy for these aircraft types are used as the weights for aggregation. The annual cost for each aircraft type is estimated by the product of the lease price and the number of airplanes. Total annualized aircraft cost is then computed as the sum across all categories of aircraft. The real stock of ground properties and equipment (GPE) is estimated using the perpetual inventory method. The annual cost of using GPE is computed by multiplying the real GPE stock by a GPE service price. The GPE service price is constructed using the method proposed by Christensen and Jorgenson (1969) which accounts for interest, depreciation, corporate income and property taxes, and capital gains or losses. Since the GPE costs are small relative to the costs of ¯ight equipment, these two categories of capital inputs are further aggregated into a single capital stock series using the translog multilateral index procedure. The price of capital input is then computed by dividing total capital cost by the aggregate capital input quantity index. The last category of inputs is materials. The materials input consists of all other inputs not included in any of the input categories discussed above. As such, materials cost is the catch-all cost category, and thus includes numerous items such as airport fees, sales commissions, passenger meals, employee travel, consultants, non-labour repair and maintenance expenses, stationery, and other purchased goods and services. The materials cost is computed by subtracting labour, fuel and capital input costs from the total operating cost reported in ICAO's Financial Data. Similar to the incidental output, it is necessary to construct a materials price index in order to include the materials input in our analysis. Since the materials cost also includes uncountable items and activities, the general price index used in constructing incidental output quantity index is used as a proxy for the materials input price. It is worthwhile to note that many studies on airline productivity have excluded the materials input, presumably because the materials input is dicult to measure. Since the materials cost accounts for 35±50% of the total cost depending on airlines, the exclusion of such an important input will obviously bias the empirical ®ndings on the relative eciency of airlines. Such a bias would be further aggravated as many airlines now outsource a variety of services, which should be accounted for in the materials cost. 3.3. Characteristics of airlines An airline's unit cost is in¯uenced by variations in the network, the operating and market conditions, and the regulatory and institutional environment, over which airlines have limited control. Therefore, it is necessary to examine the eects of these variables on the observed unit cost in order to make proper inferences about cost competitiveness. For this purpose the following variables are compiled: average stage length, average weight load factor, and revenue shares of freight and mail, non-scheduled services, and incidental services.{ In addition, the residual TFP index computed in Oum and Yu (1995) is used as an indicator of overall eciency, and is considered as an important determinant of carrier cost. The 1993 key statistics for the 22 sample airlines are reported in Table 1. The list is organized by continent and by revenue size. It shows, among other things, that Qantas, Singapore Airlines, Cathay Paci®c, Japan Airlines (JAL) and KLM Royal Dutch Airlines (KLM) are long-haul carriers, and Korean Air has heavy emphasis on freight services and non-airline businesses. Overall, many Asian carriers and some European carriers generate a relatively high proportion of their revenues from freight services as compared to the U.S. carriers. European carriers (except British Airways) have heavy emphasis on non-airline businesses as well. *Although the ICAO Financial Data reports fuel expense data, it does not report fuel price or quantity. Many airlines have provided the data on quantity series of fuel consumption upon our request. Fuel consumption for some U.S. carriers are also collected from The Airline Monitor. The fuel quantity data for Canadian carriers are collected from Statistics Canada 51-002 and 51-206. As was done in Windle (1991) a fuel quantity regression model was used to estimate fuel consumption for those airlines whose fuel consumption data are not available to us. The regression included available tonne-kilometres, aircraft kilometres, load factor, aircraft hours, aircraft departure, aircraft type and year eects. y The aircraft leasing price data were kindly supplied to us by Avmark Inc. { Although a number of points served is an important characteristic of an airline network, it is not included here because we are unable to obtain a consistent time series data especially for the non-U.S. carriers. It is noted, however, that some of the previous studies involving non-U.S. carriers, such as Good and Rhodes (1991), Good et al. (1993), and Distexhe and Perelman (1993), have not included this variable either.
412
Tae Hoon Oum and Chunyan Yu Table 1. Descriptive statistics of the sample airlines, 1993 Revenue No. of Average wage Stage length Load % Pax % Fre % Inc Rev/expb (mill U.S.$) employee ('000 U.S.$) (km) factor (%) reva(%) reva(%) reva(%)
North America American United Delta Northwest US Air Continental Air Canada Canadian
14,737 14,354 12,375 8448 6624 5086 2099 1947
91,773 78,105 69,537 42,439 45,986 36,191 18,184 15,208
56 62 70 63 61 39 36 36
1566 1631 1206 1395 866 1360 1459 1630
60 67 62 67 59 63 64 68
88 88 92 87 92 90 85 85
3 5 4 7 1 3 9 8
8 6 2 4 5 6 3 3
1.04 1.02 0.98 1.04 0.98 0.99 0.97 0.94
Australasia Japan Airlines All Nippon Singapore Airlines Korean Air Cathay Qantas Thai
8591 7226 3482 3342 2957 2688 2311
22,008 13,870 14,664 15,398 13,483 15,475 19,247
107 94 25 31 52 39 19
2396 1034 4214 1617 2990 4257 1524
65 62 71 65 70 70 67
79 89 77 58 75 77 73
13 5 19 20 17 10 12
5 5 3 18 5 6 14
0.97 1.01 1.09 1.09 1.10 1.06 1.09
Europe Lufthansa British Air Air France KLM SAS Swissair Iberia
9532 8694 6826 4321 3623 3540 2745
46,338 47,705 43,258 26,859 19,439 19,788 25,676
57 35 47 52 63 67 48
1071 1712 1637 1856 702 1243 1196
66 71 68 67 63 60 67
65 89 65 64 77 66 85
13 7 14 18 5 10 7
18 0.2 20 17 17 23 7
0.999 1.09 0.92 1.06 0.99 1.01 0.96
a b
Revenue shares for passenger, freight and incidental. Ratio of operating revenue over operating expenses based on ICAO data. 4. VARIABLE COST FUNCTION AND DECOMPOSITION OF THE UNIT COSTS
In this section, we present the estimated variable cost function, and use it to decompose the unit cost dierentials into sources. The results are used to identify the eects on unit cost of input prices, output mix, average stage length, and eciency, and to evaluate and compare cost competitiveness in the following section. 4.1. Cost function estimation The maximum likelihood parameter estimates, t-statistics and other summary statistics on the variable cost function estimation [eqn (1)] are reported in Table 2. The ®rst-order coecients for input prices indicate that at the mean data, labour and fuel inputs account for 32 and 15%, respectively, of the total variable cost. This leaves the materials input to account for 53% of the total variable cost. The ®rst-order coecient for the capital input variable is negative, implying a positive shadow value of capital input.* The stage length has a statistically signi®cant negative coecient, implying that variable cost decreases with stage length. The coecient for the eciency variable is negative and statistically signi®cant, indicating that ecient ®rms are likely to have considerably lower costs. The coecient for the time shift dummy for year 1993 indicates that the eciency of using variable input has improved by 3.3% between 1986 and 1993 due to industrywide technical progress. The negative coecients for %Incidental and %Non-Scheduled indicate that, other things being equal, carriers with a higher concentration on non-airline (incidental) businesses and non-scheduled services are expected to have lower variable costs. On the other hand, carriers with higher concentration on freight services may have a cost disadvantage.y *Many empirical studies on airline variable cost function report a positive ®rst-coecient for the capital stock variable, which implies a negative value of the shadow price of capital. See, for example, Gillen et al. (1985, 1990), and Caves et al. (1987) which was used extensively by Windle (1991). y At ®rst glance this appears to be contrary to our intuition that freight services require less input than passenger services. However, this result is plausible for the following reasons. Since cargo yields per RTK are far lower than passenger yield per RTK (average yield for our sample is U.S.$1.03 per RTK for passenger vs U.S$0.33 per RTK for freight), cargo output receives very low weight in aggregating outputs. Therefore, the amount of increase in output index caused by cargo output is relatively small as compared to the amount of increase in input cost caused by cargo output. If this is the case, % freight variable would have a positive coecient as in our case.
Cost competitiveness of airlines
413
Table 2. Variable cost function estimates Variable Constant (a0 ) Output (Y) Labour (W1 ) Fuel (Wf ) Capitala (uK) Stage length (Z) % Freight.(Rf ) % Non-sch.(Rn ) % Incidentl.(Ri ) Eciencyb (E) E.output (E Y) E.labour (E Wl ) E.fuel (E Wf ) E.capital (E uK) Eciency (E E) Labouroutput (W1 Y) Fueloutput(Wf Y) Capitaloutput (uK Y) Stageoutput (Z Y) Number of observations Log-likelihood function
Coefficient 8.109 1.123 0.318 0.153 ÿ0.106 ÿ0.305 0.055 ÿ0.007 ÿ0.040 ÿ1.138 0.050 0.090 0.100 ÿ0.063 0.723 ÿ0.054 0.059 0.017 0.028
T-value 1316.2 235.96 70.08 89.38 67.48 37.61 11.80 6.95 25.83 64.63 1.02 2.47 5.92 3.72 3.50 4.59 9.93 1.54 2.30
Variable Outputoutput (Y Y) Labourfuel (W Wf ) Labourcapital (W1 uK) Labourstage (W1 Z) Labourlabour (W1 W1 ) Fuelcapital (Wf uK) Fuelstage (Wf Z) Fuelfuel (Wf Wf ) Capitalstage (uK Z) Capitalcapital (uK uK) Stagestage (Z Z) 1987 (T87 ) 1988 (T88 ) 1989 (T89 ) 1990 (T90 ) 1991(T91 ) 1992 (T92 ) 1993 (T93 )
Coefficient
T-value
ÿ0.013 ÿ0.038 0.036 ÿ0.088 0.188 ÿ0.059 0.008 0.088 ÿ0.028 ÿ0.012 0.036 ÿ0.003 ÿ0.014 ÿ0.009 ÿ0.010 ÿ0.021 ÿ0.029 ÿ0.033
0.70 5.14 4.85 8.07 7.65 12.12 2.00 15.19 7.34 1.21 2.09 0.42 2.16 1.42 1.34 3.25 4.34 4.87
178 1572.28
All variables except time dummies are in natural log and normalized at mean; in addition to the cost shares equations, the equation for the ratio of depreciated capital cost to variable cost, which is equal to negative partial derivative dk Cv , is included in the regression. a Capital is capital stock multiplied by weight load factor. b Residual TFP index from Oum and Yu (1995) is used as Eciency.
4.2. Unit cost decomposition The 1993 unit cost (cost per unit of aggregate output) dierentials between each airline and American Airlines (AA) are decomposed into dierent sources using eqn (4), and the results for 1993 are summarized in Table 3. Column (1) lists the observed unit cost dierence expressed in percentage dierence of each airline's unit cost relative to American Airlines (AA). For example, in 1993, Delta's unit cost was 13.5% higher than AA's while Singapore Airline's unit cost was 30.6% lower than AA's. Asian carriers, except Japan Airlines (JAL) and All Nippon Airways (ANA), have substantially lower unit costs than the major U.S. carriers while most of the U.S. carriers, in turn, have substantially lower unit costs than the European carriers. KLM is an exception in that its unit cost is about 10% lower than Delta's. Columns (2)±(7) in Table 3 report on the decomposition of the unit cost dierences, that is, the contribution of each source to the observed unit cost dierence. Each of the entries listed under `Sources of dierence' is measured as the percentage dierence in total unit cost that would result if the only dierence between that particular airline and AA were that particular source. For example, for United, in column (4) under `labour price', 3.3 indicates that if the only dierence between United and AA in 1993 were the price of labour, then United's unit cost would have been 3.3% higher than AA's. Columns (2) and (3) show the eects of stage length and output mix (scheduled passenger, freight, non-scheduled, and incidental businesses). Variations in stage length alone account for a substantial portion of the observed unit cost dierences, especially for the carriers at extreme ends of the scale. For example, if other things are equal, Singapore Airlines' and Qantas' unit cost would be 23 and 25% lower than AA's, respectively, because of their long stage length. SAS and US Air, on the other hand, would have 25 and 18% higher unit cost than AA, respectively, again if other things were equal. Output mix, in general, has only limited eects on the observed unit cost. The only noticeable exception is British Airways (BA). BA has very little incidental and nonscheduled services, thus its unit cost would be 11% higher than AA's if other things were equal. Columns (4)±(6) of Table 3 are the percentage dierences in unit costs between each airline and AA attributable to, respectively, the dierences in labour price, other input prices, and all input prices together. It is noted that Asian carriers, with the exception of JAL and ANA, enjoyed signi®cant cost advantage relative to AA due to their lower labour and non-labour input prices. For
414
Tae Hoon Oum and Chunyan Yu Table 3. Unit cost decomposition and cost competitiveness, 1993 (% above and below AA's unit cost) Observed unit cost difference (1)
Sources of difference Firm characteristics
Input prices
Efficiency
Stage Output mix Labour Other inputs All inputs (2) (3) (4) (5) (6)
(7)
Cost competitiveness (8)=(6)+(7)
North America America United Delta Northwest Continental US Air Air Canada Canadian
0.0 ÿ1.7 13.5 ÿ3.7 ÿ10.7 40.9 12.7 3.6
0.0 ÿ1.2 7.7 3.3 3.8 17.6 1.9 ÿ1.1
0.0 0.2 3.1 2.1 ÿ2.6 ÿ4.2 3.1 1.5
0.0 3.3 7.7 3.9 ÿ10.7 3.0 ÿ12.4 ÿ12.3
0.0 0.4 ÿ0.3 0.7 ÿ0.7 ÿ0.4 0.9 3.7
0.0 3.7 7.4 4.6 ÿ11.4 2.6 ÿ11.5 ÿ8.6
0.0 ÿ3.8 ÿ5.6 ÿ9.9 ÿ0.7 17.4 19.9 13.7
0.0 ÿ0.1 1.8 ÿ5.3 ÿ12.1 19.9 8.5 5.0
Australasia Japan Airlines All Nippon Singapore Airlines Korean Air Cathay Qantas Thai
50.1 114.8 ÿ30.6 ÿ25.2 ÿ18.3 ÿ24.6 ÿ20.8
ÿ13.0 13.0 ÿ23.0 ÿ0.8 ÿ17.4 ÿ24.9 0.6
3.0 3.9 6.8 0.1 4.4 0.5 ÿ0.1
22.6 18.8 ÿ16.9 ÿ14.7 ÿ2.1 ÿ8.9 ÿ22.4
15.9 21.5 ÿ3.3 ÿ9.0 ÿ4.2 ÿ0.1 ÿ29.8
38.4 40.4 ÿ20.3 ÿ23.8 ÿ6.4 ÿ8.9 ÿ52.1
14.3 23.1 3.9 0.8 2.6 11.6 42.9
52.7 63.5 ÿ16.3 ÿ22.9 ÿ3.8 2.7 ÿ9.3
19.3 29.2 21.9 3.3 81.5 46.4 36.9
ÿ1.2 11.2 ÿ2.4 ÿ6.2 25.0 5.4 7.6
ÿ0.8 ÿ2.0 11.0 1.7 ÿ4.7 ÿ2.2 2.8
ÿ5.2 0.6 ÿ12.8 0.6 4.4 9.4 ÿ4.9
14.0 16.3 9.9 15.5 21.0 26.1 10.0
8.8 16.8 ÿ2.9 16.0 25.4 35.5 5.1
12.4 3.8 10.2 ÿ5.3 17.0 2.8 16.4
21.2 20.6 7.3 10.7 42.4 38.3 21.5
Europe Air France Lufthansa British Airways KLMa SAS Swissaira Iberia a
1992.
example, over 90% of Korean Air's unit cost advantage over AA came from its lower input prices. On the other hand, JAL's high labour price caused a 23% unit cost disadvantage while their higher non-labour input prices (fuel, capital and materials) were responsible for a further 16% unit cost disadvantage relative to AA. Major U.S. carriers, except Continental, incurred 3±8% higher cost relative to AA because of higher labour prices. Continental's lower labour price gave it a 11% cost advantage over AA. The two Canadian carriers, Air Canada and Canadian Airlines International (CAI), also had approx. 12% cost advantage over AA due to lower labour prices. The eects of other input prices are very small for North American carriers, except CAI which had a 3.7% cost disadvantage relative to AA due to a slightly higher non-labour price. All European carriers suered substantial cost disadvantage relative to AA due to signi®cantly higher nonlabour input prices, ranging from 10% for BA and Iberia to 26% for Swissair. The eects of labour prices, on the other hand, vary across airlines. For example, B A enjoyed a 13% cost advantage over AA due to lower labour price, while Swissair had a 10% cost disadvantage due to higher labour price. Column (7) lists the contribution of eciency to the unit cost dierence. The results show that higher eciency helped to reduce unit cost for most of the major U.S. carriers (except US Air) and KLM in 1993. The reduction ranges from 1% for Continental to 10% for Northwest. However, all other carriers in our sample would have had unit cost disadvantages of varying degrees relative to AA because of their lower productive eciency. For example, ineciency alone accounted for a 20% higher unit cost for Air Canada, which negated the 12% cost advantage it had over AA from lower input prices. Similarly, Thai Airways' ineciency would have led to a 43% higher unit cost than AA were it not for the favourable impact of its signi®cantly lower input prices. ANA's lower eciency level led to a further 23% cost disadvantage in addition to the 40% cost disadvantage caused by higher input prices. In summary, except JAL and ANA, the Asian carriers including Singapore, Cathay, Korean Air, Qantas, and Thai Airways enjoy substantial unit cost advantages relative to the U.S. carriers.
Cost competitiveness of airlines
415
JAL and ANA have a substantial unit cost disadvantage vis-aÁ-vis the U.S. mega carriers. Most of the unit cost advantages of Asian carriers come from their lower input prices including labour prices. KLM enjoys a unit cost advantage among the European carriers while it has a unit cost disadvantage relative to the Asian carriers (except JAL and ANA) and some U.S. carriers. This result is consistent with the ®ndings of Bureau of Industry Economics (1994) and European Commission (1994). 5. UNIT COST COMPETITIVENESS
As discussed earlier, the observed unit cost dierence does not re¯ect the true comparative cost competitiveness between airlines because these airlines have dierent network and operating characteristics. What one needs to ask is how these carriers would be able to compete in a given market such as the trans-Paci®c and trans-Atlantic markets. A low system-wide unit cost for an airline with heavy concentration on incidental services and with a long average stage length may not constitute cost competitiveness in a speci®c market. When an airline competes in a given market, particularly in an inter-continental market, what is relevant is the marginal cost of providing a given level of services in that market. What determines cost competitiveness is the input prices the airline pays and how eciently the airline produces and markets their services. Therefore, a cost competitiveness (CC) indicator is constructed by summing the eects of input prices and eciency reported in column (6) and (7) of Table 3. Since the unit cost decomposition disentangles the eects of output mix and stage length from the eects of input prices and eciency, this indicator approximates the `true' comparative cost competitiveness of the airlines. Column (8) of Table 3 presents the cost competitiveness indicator for 1993. The cost competitiveness indicator is measured in terms of percentage above(ÿ) or below (+) that of AA. A negative number indicates that the airline is cost competitive relative to AA, while a positive number indicates vis-aÁ-vis. Continental (CO) enjoyed an over 12% cost competitiveness relative to AA in 1993. Most of its cost competitiveness came from its lower labour price. Northwest enjoyed a 5% cost competitiveness over AA. Although its high eciency gave it a 10% cost advantage, its higher input prices reduced its cost competitiveness. United and Delta were similar to AA in terms of cost competitiveness (with CC at ÿ0.1 and 1.8%, respectively). Although their higher eciency gave them a 3.8 and 5.6% cost advantage, respectively, their higher input prices created a 3.7% cost disadvantage for United and 7.4% for Delta. US Air was 20% less cost competitive relative to AA. This consists of 17 and 3% due to their lower eciency and higher input prices, respectively. Air Canada (AC) was about 8.5% less cost competitive relative to AA because of its substantially lower eciency. CAI was about 5% less cost competitive relative to AA, also due to its lower eciency level. Both AC and CAI's cost advantage from lower labour prices were overpowered by the eects of their lower eciency. Among the Asian carriers, Singapore, Korean, Cathay and Thai enjoyed higher cost competitiveness relative to AA, by factors of 16, 23, 4 and 9%, respectively. Their cost competitiveness came entirely from their signi®cantly lower input prices, while their relatively lower eciency level diminished their cost advantages. Qantas was slightly less cost competitive (2.7%) than AA because its cost advantage from lower input prices was more than o set by its lower eciency level. JAL and ANA were signi®cantly less cost competitive than AA because of the combined eects of higher input prices and lower eciency. BA was about 7% less cost competitive relative to AA due to higher non-labour prices and lower eciency. KLM was 10% less cost competitive relative to AA. This was entirely due to the cost disadvantage caused by higher input prices, albeit its higher eciency helped alleviate some of this cost disadvantage. Swissair was 38% less cost competitive relative to AA. High input prices were the main cause of this cost disadvantage. Lufthansa was 21% less cost competitive relative to AA. Their cost disadvantage was mostly caused by high input prices. Air France was also about 21% less cost competitive relative to AA. This resulted from a combined eect of high non-labour input prices and low eciency. SAS and Iberia were, respectively, 42 and 32% less cost competitive relative to AA. Their cost disadvantages resulted from high input prices and low eciency level.
416
Tae Hoon Oum and Chunyan Yu
Figures 1±3 show changes over time in cost competitiveness and its components for each of the sample airlines. Note that the comparisons are made between each of the airlines and AA in the same year, therefore the changes also partly re¯ect the changes in AA's unit costs. Although there had been considerable ¯uctuations over the years in the degrees of the relative cost competitiveness (and the eects of its components) for the airlines, the relative rankings among the airlines remained fairly stable. The most noticeable changes are with BA and Qantas, both airlines had considerably enhanced their relative cost competitive positions mainly through improvement in eciency. On the other hand, Air France's cost competitive position had deteriorated somewhat since 1990 as a result of higher input price and lower eciency. It is also noted that Swissair and KLM's eciency level had improved signi®cantly during the period, but their input prices had also risen signi®cantly. The eects of the two opposite forces appear to counteract each other. As a result, Swissair and KLM's cost competitive positions did not show any signi®cant changes. Overall, European carriers are less cost competitive relative to the U.S. airlines. This is caused by both higher input prices and lower eciency than their U.S. counterparts. The Australasian carriers, with the exception of JAL and ANA, are more cost competitive than the U.S. airlines. However, their cost competitiveness is entirely the result of low input prices rather than high eciency.
Fig. 1. Cost competitiveness (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers.
Cost competitiveness of airlines
417
Fig. 2. Eects of input prices (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers. 6. THE EFFECTS OF EXCHANGE RATE ON AIRLINE COST
So far in our discussion, all costs are expressed in a common currencyÐU.S. dollar. Since the airlines compete in international markets, it is common practice to compare their operating results in terms of a common currency. However, exchange rate ¯uctuations have had non-negligible impact on inter-country cost dierentials (denominated in a common currency). In this section, the eects of exchange rate ¯uctuations on carrier cost are examined via brief case studies of JAL and Lufthansa. Changes in Japanese and German exchange rates have played an important part in explaining the ability of JAL and Lufthansa to compete in international markets. Figures 4 and 5 plot the changes in JAL's labour and non-labour prices*, respectively, in relation to Japan's exchange rate. Japanese yen appreciated about 35% against U.S.$ between 1986 and 1993. Therefore, both JAL's labour and non-labour input prices in U.S.$ terms have increased substantially relative to AA. The pattern of input price changes in Japanese yen, however, tells a *Input prices are expressed in terms of percentage of AA's input prices.
418
Tae Hoon Oum and Chunyan Yu
Fig. 3. Eects of eciency (% above or below AA's unit cost); (a) North American carriers; (b) European carriers; (c) Asian carriers.
Fig. 4. Labour price vs exchange rate: Japan Airlines (base: AA=1.00).
Cost competitiveness of airlines
419
Fig. 5. Non-labour price vs exchange rate: Japan Airlines (base: AA=1.00).
dierent story (as shown by the input prices converted using constant 1986 exchange rate). In fact, between 1986 and 1993 JAL's non-labour input prices decreased by about 24% in terms of Japanese yen, but increased by 20% in terms of U.S.$. Similar observations can be made for Lufthansa. Figures 6 and 7 plot the changes in Lufthansa's input prices in relation to the Deutsche mark. The Deutsche mark appreciated by 24% between 1986 and 1993. In Deutsche mark, Lufthansa's input prices have declined since 1990 relative to AA (as shown by the input prices measured in constant 1986 exchange rates). In U.S.$, however, there is no obvious downward trend. These dierencesÐbetween calculation in U.S.$ vs each country's own currencyÐare due to changes in exchange rates. Figures 8 and 9 plot changes in JAL's and Lufthansa's unit costs and exchange rates. In U.S.$, JAL's unit cost increased by 39% between 1986 and 1993, while its unit cost in Japanese yen decreased by 12% (at constant 1986 exchange rate). Similarly, Lufthansa's unit cost increased by 16% in 1993 from 1986's level in U.S.$ terms, but decreased by about 10% in terms of its home currency (at constant 1986 exchange rate). Exchange rate ¯uctuation is entirely beyond a carrier's control, but it has had decisive eects on JAL and, to a lesser degree, on Lufthansa's cost competitiveness. Figures 8 and 9 show that in 1993 JAL and Lufthansa's unit costs would have been 4 and 2% lower than that of AA, respectively, had there been no change in exchange rate between 1986 and 1993. Since our sample airlines are all international carriers, their cost competitiveness are all aected by exchange rate ¯uctuation, albeit to dierent degrees. The eects could be both positive and negative. Home currency appreciation may lower cost competitiveness of its carriers in the short run, but might force the carriers to improve eciency. On the other hand, home currency depreciation generally improves cost competitiveness of its carriers at least in the short run, but may lead to decreased productivity of the home carriers caused by reduced competitive pressure.
Fig. 6. Labour price vs exchange rate: Lufthansa (base: AA=1.00).
420
Tae Hoon Oum and Chunyan Yu
Fig. 7. Non-labour price vs exchange rate: Lufthansa (base: AA=1.00).
Fig. 8. Unit cost vs exchange rate: Japan Airlines (base: AA=1.00).
Fig. 9. Unit cost vs exchange rate: Lufthansa (base: AA=1.00).
7. SUMMARY AND CONCLUDING REMARKS
This study measures and compares cost competitiveness of 22 major international airlines. Cost competitiveness depends on input factor prices and eciency. Observed unit cost dierentials between airlines, however, are also in¯uenced by the eects of network and output attributes
Cost competitiveness of airlines
421
which must be removed in order to make meaningful comparisons across ®rms and/or over time within a ®rm. Therefore, a translog variable cost function is estimated and used to decompose the total unit cost dierentials into potential sources: stage length, output mix, input prices, and eciency. The cost decomposition results are then used to develop a cost competitiveness (CC) indicator after removing the eects of stage length and output mix. The study also examines the eects of exchange rate ¯uctuation on input prices and unit cost via brief case studies of JAL and Lufthansa. The main results are summarized as follows: (a) The carriers in Australasia (except JAL and ANA) are generally more cost competitive than the major U.S. carriers (Northwest and Continental are as cost competitive as some of the Australasian carriers). Lower input prices are the dominating reason for their cost competitiveness, and their cost competitiveness relative to AA has decreased over time (except Qantas and Korean Air). (b) JAL and ANA are signi®cantly less cost competitive than AA mainly because of their higher input prices. (c) Major European carriers were 7 (BA)±42% (SAS) less cost competitive than AA in 1993 because of the combined eects of higher input prices and lower eciency level. However, European carriers' eciency performance improved noticeably relative to AA during the 1986±93 period; (d) In 1993, AA, United and Delta are similar in cost competitiveness, while Northwest and Continental enjoyed, respectively, 5 and 12% cost competitiveness over AA. US Air was least cost competitive among the North American carriers, due to its lower eciency and slightly higher labour price. (e) Air Canada was about 8.5% less cost competitive relative to AA due to its lower eciency level. CAI was about 5% less cost competitive relative to AA also due to its lower eciency level. Their cost advantage from lower labour price is cancelled out by the eects of lower eciency. (f) Exchange rate ¯uctuation has had signi®cant impact on some carriers' costs. For example, JAL, and, to a lesser degree, Lufthansa's less cost competitive position is caused to a large extent by their home currency appreciation. In 1993, JAL's and Lufthansa's unit costs would have been 4 and 2%, respectively, lower than AA's had there been no change in their respective exchange rates between 1986 and 1993. Overall, input prices together (including exchange rates) have been a more important factor than eciency for determining a carrier's cost competitive position in the past. However, the importance of input prices is likely to diminish over time as airlines increase their global sourcing of labour, materials, services and other inputs, and as input prices in the developing and the Newly Industrialized Countries (NIC) continue to rise faster than in the developed countries. As liberalization of the airline industry continues, eciency will become progressively more important in determining cost competitiveness of an airline. This paper reports on the ®rst phase of our investigation on airline cost competitiveness. There are two points that need to be addressed in our future research. First, this study uses the traditional econometric method in estimating the variable cost function. Although an eciency variable is incorporated in the estimation, the resulting cost function is still an `average cost' function which may not truly re¯ect the cost frontier. In our future research, therefore, the recently developed stochastic frontier method will be applied to re®ne our empirical results on the cost function estimation, and to further investigate the causal relationship between competitiveness and eciency. Secondly, this paper illustrates the importance of the impact of exchange rates on airline cost via brief case studies on JAL and Lufthansa. Future research will incorporate a systematic investigation of the eects of exchange rates on international air transport competition. One possible way to accomplish this is to adjust the decomposition formula based on the concept of a fundamental equilibrium exchange rate which was developed by Williamson (1985), and applied by Fuss and Waverman (1992) in a cost and productivity study of automobile production. The adjusted decomposition will be able to isolate changes in cost dierences due to exchange rate ¯uctuations from those due to relative movements in factor prices and eciency within each airline.
422
Tae Hoon Oum and Chunyan Yu
AcknowledgementsÐThe authors would like to thank the seminar participants at University of British Columbia, Osaka University, Nanzan University, Canadian Transportation Research Forum Conference, Japan Airlines Headquarters, Korean Air Headquarters and Korea Transport Institute for helpful comments and suggestions. The research grant supports from the Social Science and Humanities Research Council of Canada and The Japan Ministry of Education via Osaka University are gratefully acknowledged.
REFERENCES Baltagi, B. H., Grin, J. M. and Rich, D. P. (1995) Airline deregulation: the cost pieces of the puzzle. International Economic Review 36(1), 245±258. Bauer, P. W. (1990) Decomposing TFP growth in the presence of cost ineciency, nonconstant returns to scale, and technological progress. The Journal of Productivity Analysis 1, 287±299. Bureau of Industry Economics (BIE) (1994) International Performance Indicators: Aviation, Vol. 59. Canberra, Australia. Caves, D. W. and Christensen, L. R. (1988) The importance of economies of scale, capacity utilization, and density in explaining interindustry dierences in productivity growth. The Logistics and Transportation Review 24(1), 3±32. Caves, D. W., Christensen, L. R. and Diewert, W. E. (1982) Multilateral comparisons of output, input, and productivity using superlative index numbers. Economic Journal 92, 73±86. Caves, D. W., Christensen, L. R. and Tretheway, M. W. (1984) Economies of density versus economies of scale: why trunk and local service airline costs dier. Rand Journal of Economics 15, 471±489. Caves, D. W., Christensen, L.R., Tretheway, M. W. and Windle, R. J.(1987) An assessment of the eciency eects of U.S. airline deregulation via an international comparison. In Public Regulation: New Perspectives on Institutions and Policies, ed. E. E. Bailey, pp. 285±320. MIT Press, Cambridge, MA. Christensen, L. R. and Jorgenson, D. W. (1969) The measurement of U.S. real capital input, 1929±1967. The Review of Income and Wealth 15(1), 293±320. Distexhe, V. and Perelman, S. (1993) Technical eciency and productivity growth in an era of deregulation: the case of airlines. Paper presented at the Third European Workshop on Eciency and Productivity Measurement, CORE, Belgium, October. European Commission (1994) Expanding Horizons: Civil Aviation in Europe, an Action Programme for the Future. Report by the Comite des Sages for Air Transport, Brussels, Belgium. Encaoua, D. (1991) Liberalizing European airlines: cost and factor productivity evidence. International Journal of Industrial Organization 9, 109±124. ESG Aviation Services, The Airline Monitor, various issues. Ponte Vedra Beach, Florida, U.S.A. Fuss, M. A. and Waverman, L. (1992) Cost and Productivity in Automobile Production: the Challenge of Japanese Eciency. Cambridge University Press, New York. Gillen, D. W., Oum, T. H. and Tretheway, M. W. (1985) Airline Cost and Performance: Implications for Public and Industry Policies. Centre for Transportation Studies, University of British Columbia, Vancouver, B.C., Canada. Gillen, D.W., Oum, T.H. and Tretheway, M.W. (1990) Airline cost structure and policy implications. Journal of Transport Economics and Policy 24(2), 9±34. Good, D. H. and Rhodes, E. L. (1991) Productive eciency, technological change and the competitiveness of U.S. airlines in the Paci®c Rim. Journal of the Transportation Research Forum 31(2), 347±358. Good, D. H., Nadiri, M.I., RoÈller, L. H. and Sickles, R. C. (1993) Eciency and productivity growth comparisons of European and U.S. air carriers: a ®rst look at the data. The Journal of Productivity Analysis 4, 115±125. Good, D. H., RoÈller, L.H. and Sickles, R. C. (1995) Airline eciency dierences between Europe and the US: implications for the pace of EC integration and domestic regulation. European Journal of Operational Research 80(1), 508±518. International Civil Aviation Organization (ICAO) (1986±1993) Digest of Statistics, various annual series. Montreal, Canada. Oum, T. H. and Yu, C. (1995) A productivity comparison of the world's major airlines. Journal of Air Transport Management 2(3/4), 181±195. Oum, T. H. and Zhang, Y. (1991) Utilization of quasi-®xed inputs and estimation of cost functions. Journal of Transport Economics and Policy 25(2), 121±134. Oum, T. H. and Zhang, Y. (1995) Competition and allocative eciency: the case of competition in the U.S. telephone industry. Review of Economics and Statistics 77(1), 82±96. Ray, S. and Hu, X. (1993) A nonparametric decomposition of the malmquist productivity index: a study of airlines' data. Paper presented at the Third European Workshop on Productivity, CORE, Belgium, October. Summers, R. and Heston, A. (1991) The Penn World Table (Mark 5): an expanded set of international comparisons, 1950± 1988. Quarterly Journal of Economics 106(2), 327±368. Tofallis, C. (1996) Input eciency pro®ling: an application to airlines. Working paper UHBS 1996:21, The Business School, University of Hertfordshire, U.K. Williamson, J. (1985) The Exchange Rate System. Institute for International Economics, Washington, DC. MIT Press. Cambridge, MA. Windle, R. J. (1991) The world's airlines: a cost and productivity comparison. Journal of Transport Economics and Policy 25(1), 31±49. Xu, K., Windle, R., Grimm, C. and Corsi, T. (1994) Re-evaluating returns to scale in transport. Journal of Transport Economics and Policy 28(3), 275±286.