Measuring total factor productivity of airports— an index number approach

Measuring total factor productivity of airports— an index number approach

Transpn Res.-E, Vol. 33, No. 4, pp. 249±259, 1997 # 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain 1366-5545/97 $17.00+0.00 ...

154KB Sizes 2 Downloads 120 Views

Transpn Res.-E, Vol. 33, No. 4, pp. 249±259, 1997 # 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain 1366-5545/97 $17.00+0.00

Pergamon PII: S1366-5545(97)00033-1

MEASURING TOTAL FACTOR PRODUCTIVITY OF AIRPORTSÐ AN INDEX NUMBER APPROACH P. G. HOOPER, and D. A. HENSHER

Institute of Transport Studies, University of Sydney, Australia (Received 1 May 1997; in revised form 30 November 1997) AbstractÐThere is an increasing trend towards corporatisation and privatisation of airports in an e€ort to improve performance. However, the normal ®nancial reporting requirements associated with these forms of organisation are not sucient indicators of the performance of airports since pro®tability can be more a function of the exercise of market power than a sign of productive eciency. Also, there are concerns that e€orts to regulate the prices charged by airports can result in under-investment and declining service standards. This makes it important to monitor the cost-eciency, cost-e€ectiveness and service-e€ectiveness of airports. There is a growing literature on these topics, but so far there has been little attempt to apply the concepts of total factor productivity to the airport sector. We use a non-parametric index number method to illustrate how such a global measure can be used to investigate the role of disaggregated performance measures that often are very useful to managers and to those monitoring airport operations. # 1997 Elsevier Science Ltd. All rights reserved Keywords: airport, performance measures, total factor productivity. 1. INTRODUCTION

During the 1990's, there has been a trend to corporatise and privatise airports (Ashford, 1994; Haririan and Vasigh, 1994). Despite concerns that airports enjoy a signi®cant degree of monopoly power (Forsyth, 1984; Hazell, 1995), governments have expressed faith that their initiatives will result in increased commercial pressure on airport operators (Prins and Lombard, 1995; Feldman, 1996). One key reason for this has been the liberalisation of competition among airlines and the emphasis this has placed on network strategies with hierarchies of airports (Dennis, 1994). As a result, some hub airports have been brought into direct rivalry with each other for connecting trac. Often, there is scope to establish competition within the airport precincts for at least some of the services airports provide to airlines and passengers. In other cases, there are local businesses that are able to meet the needs of airport users. For some activities, such as aircraft maintenance, the airlines are free to choose among several candidate airports. Furthermore, every category of expenditure is important to the airlines in competitive markets and they have a strong incentive to use their countervailing power as major customers to negotiate lower charges (Gollin, 1996). Concerns about monopoly power have been dealt with in the United Kingdom by placing the British Airports Authority Plc (BAA) under price capping regulation using a CPI-X approach. This requires that some airport charges be increased at not more than the general rate of in¯ation less some pre-determined `X-factor'. This mechanism establishes an incentive for managers to pursue productivity improvements over time. New Zealand and Australia have adopted similar approaches to price-capping in what is described as a `light-handed' approach to regulation (Prices Surveillance Authority, 1995). The success of price-capping depends upon the initial capitalisation of the airport (Mills, 1994), but the method has the shortcoming that it favours a gradual process of productivity improvement. An even greater problem could be that the incentive to invest is weakened and the result could be declining service standards in terms of congested air-side facilities and crowded passenger terminals. Under-investment in capacity can occur regardless of *Author for correspondence. Fax: 0061 2 9351 0088; email: [email protected]

249

250

P. G. Hooper and D. A. Hensher

whether the price cap is set too high or too low in relation to long-run marginal cost (Forsyth, 1993). BAA, for example, has been criticised for its failure to make appropriate investments at Heathrow Airport (Haririan and Vasigh, 1994). In theory and in practice it can be shown that a monopolist subject to a price cap has an incentive to reduce quality (Rovizzi and Thompson, 1992). Although governments regard corporatisation and privatisation as ways to improve performance of the airport sector, these initiatives often are accompanied by a commitment to regulation. Whether this is achieved under the `light-handed'-approach of price-capping or under some other arrangement there is a continuing need to monitor quality of service and productivity. In the United Kingdom, for example, there has been extensive monitoring of BAA's service quality and productivity (Monopolies and Mergers Commission, 1991). There is a requirement to measure and compare the performance of airports both from the point of view of using resources eciently and in terms of providing services e€ectively. The extant literature on performance measurement for airports restricts its analysis to the use of partial measures that yield an incomplete representation of the important relationships. The lack of published research on overall measures of performance places a limit on our understanding of productive processes in the airport sector. The following section explains the concepts involved in performance measurement and examines the growing literature on the topic. In contrast to other areas of the transport sector, however, there has been little exploration of the use of comprehensive measures of performance in relation to airports (Hooper, 1987; Oum et al., 1992). In this issue, Gillen and Lall apply data envelope analysis (DEA) as such a measure of productivity while we illustrate the potential to apply the total factor productivity (TFP) concept in an empirical study of Australian airports. An important feature of the TFP approach is that it allows us to distinguish productivity di€erences among airports that arise from scale e€ects as opposed to those di€erences resulting from managerial performance. We conclude it is possible and desirable to measure the TFP of airports. The results have important policy implications in the changing regulatory and managerial environment and we indicate fruitful ways of extending our exploratory work. 2. MEASURING THE PERFORMANCE OF AIRPORTS

In a survey of the use of performance measures in Europe, Doganis and Graham (1987) found that most airports relied upon purely ®nancial measures of performance. However, statistics such as rate of return on assets can be misleading indicators of economic and social performance. Monopolistic enterprises can make substantial surpluses when they are inecient, whereas competition can reduce pro®ts even if the individual ®rms are highly productive. Pro®tability is the result of the relationship between productivity, market power, regulatory controls and the choice of markets to serve. This is an important point because it reminds us that getting costs down does not guarantee long run pro®ts (and/or minimum subsidy). There is a need to develop demand or market indicators (often called measures of service e€ectiveness) that identify how e€ective the produced services (e.g. gate hours of service) are in servicing market demand (e.g. passenger throughput or aircraft landings). Performance has many dimensions and it is useful to distinguish between eciency and e€ectiveness. The (cost-) eciency of an enterprise represents the manner in which the physical inputs of labour, energy, maintenance materials, capital and overheads are used to produce the physical (intermediate) services such as gate hours of operation. Cost eciency is concerned with the supply-side relationships. E€ectiveness has two essential components: (i) cost-e€ectivenessÐthe relationship between inputs and consumed services (e.g. passenger throughput or aircraft landings), and (ii) service-e€ectivenessÐthe relationship between produced services (e.g. gate hours of operation) and consumed services (e.g. tonnes landing, passenger throughput or aircraft landings). Cost e€ectiveness is concerned with the demand-side relationships. The cost eciency measure is of particular interest to the operator because it relates to service levels to a large extent under their control, given passenger levels and/or landings. Government regulators also are interested in how cost e€ective each operator is in serving passengers and aircraft, the latter representing the prime purpose for being in business. The growing literature on measuring the performance of airports is addressing the limitations of ®nancial measures in capturing all of these important dimensions. Many of the studies use partial productivity measures by looking at the use of labour and capital inputs and their relationships to

Measuring total factor productivity of airportsÐan index number approach

251

output. A di€erent approach is to focus on the performance of individual processes, especially passenger terminals, baggage and cargo handling and servicing of aircraft. In recognition of the importance of customer service levels, an increasing amount of attention is being focused on passengers, cargo and airlines as major stakeholders. The particular performance measures often are expressed as ratios or as shares, but they can be measured in absolute standards such as `distance to departure gate' as an indicator of passenger convenience. Table 1 presents a summary of commonly used measures to exemplify the diversity of measures in current use. Based upon their analysis of 24 European airports, Doganis and Graham (1987) recommended a multi-faceted approach to measuring airport performance in which a battery of indicators is used. For example, information about cost performance can be provided using cost per workload unit (excluding depreciation and interest), capital cost per workload unit, and labour cost per workload unit. Relativities between di€erent airports and the same airport over time can be re¯ected in measures such as the share of labour costs in total costs. Speci®c measures of labour productivity can be measured in terms of workload units per employee or value added per employee. Capital productivity is re¯ected in similar indicators by substituting units of capital costs for employees. Airlines tend to measure the performance of airports in terms of landing charges and the share of aeronautical revenue in total earnings of the airport. Having measured airport performance in these multiple dimensions, it is possible to carry out systematic analyses of the factors contributing to di€erences among airports (Graham and Dennis, 1993). A somewhat di€erent approach to measuring airport performance has been pursued within the World Bank (Gannon and Shalazi, 1995) with distinctions made between service delivery, service quality, safety, accessibility, ®nancial sustainability and environmental quality. For example, Table 1. Airport performance measuresÐtheir scope and variety Scope of measure

Category

Examples of performance measures

Global performance of airport

Pro®tability

Ðrate of return on capital Ðrevenue to expenditure ratio Ðpro®t per workload unit (WLU) Ðcost per WLU (excl depreciation and interest) Ðoperating cost per WLU Ðcapital cost per WLU Ðlabour cost per WLU Ðaeronautical costs per WLU Ðtotal revenue per WLU Ðaeronautical revenue as a share of total Ðaeronautical revenue per WLU Ðnon-aeronautical revenue per WLU Ðvalue added per unit of capital costs ÐWLU per unit of net asset value Ðtotal revenue per unit of net asset value ÐWLU per employee Ðrevenue per employee Ðvalue added per employee Ðaircraft movements per metre of runway Ðaircraft movements per hourly capacity Ðpassengers per aircraft movement Ðservice times for check-in Ðtime to reclaim baggage Ðgate utilisation rates Ðbaggage handled per unit of time Ðbaggage service reliability over time Ðdistances to reach departure gates Ðcrowding (passenger density) Ðvariability in service times Ðpassenger service ratings Ðaverage time required to deliver freight at cargo terminal prior to aircraft departure Ðtheft and breakage rates Ðindex of aeronautical charges Ðindex of non-aeronautical charges Ðaircraft turn-round times

Cost-eciency

Cost-e€ectiveness (revenue earning) Partial productivity measures

Capital productivity Labour productivity

Performance of particular processes

Runways Passenger processing Baggage handling

Customer-service measures

Passengers

Cargo Airlines

Note: A `workload unit' is equal to one passenger or 100 kg of cargo. Sources: Doganis and Graham (1987); Seneviratne and Martel (1991); BIE (1994); Gannon and Shalizi (1995).

252

P. G. Hooper and D. A. Hensher

ecient service delivery can be measured in terms of charges on aircraft and passengers whereas service quality is re¯ected in the usage density of terminal space and average queuing delays to aircraft. Many of these measures focus attention on sectoral and social performance rather than the management of the airport. An illustration of this is that air trac control usually is a separate function to the management of the airport and yet it can have a large bearing on the use of airport capacity. Also, airline management practices can have an impact on the airports ability to process aircraft, passengers and cargo. The Bureau of Industry Economics, in reviewing the performance of Australian airports (BIE, 1994), preferred to focus on key processes, including aircraft handling and passenger processing. Labour productivity was measured in terms of aircraft handling and customs processing, but indicators were evaluated for speci®c functions such as air trac control and ®re ®ghting and rescue services. The BIE found the measurement of capital productivity to be more problematic. An indicator such as aircraft movements per metre of runway or of passengers handled per aircraft movement does not take account of the mix of heavy and light aircraft. Also, some airports require cross-runways because of weather conditions. The BIE adopted an approach involving a wide range of capital productivity measures. For example, the performance of airport terminals was measured in terms of passengers per check-in desk and passengers per immigration desk. These indicators re¯ect capital utilisation and separate measures were presented to deal with passenger processing, the key dimensions being timeliness, reliability and airport amenities. Baggage and cargo handling eciency can be measured in terms of delivery times, but additional indicators were used to incorporate customer satisfaction rankings. The measurement of the performance of airport terminals has received the greatest amount of attention in published research with a good deal of emphasis on the satisfaction of consumer needs (Seneviratne and Martel, 1991, 1994; Lemer, 1992). Detailed indicators of these kinds can assist managers of airports as well as providing public evidence about the performance of airports in terms in all three dimensions, cost-eciency, coste€ectiveness and service-e€ectiveness. However, selective use of partial measures can ensure that an enterprise is portrayed in the best light, when in fact the links with an overall index of performance are quite tenuous if not misleading. Some partial indicators can have a strong association with overall performance, but the challenge is to identify the relevant partial measures and to understand how these respond to the actions of managers as opposed to exogenous factors. For example, the measured performance of an airport can be a€ected by organisational factors (e.g. how much control the airport operator has over terminals) and by operating conditions (e.g. the amount of snow to be cleared). Performance also is a function of the composition of airline trac (e.g. connecting vs terminating passengers) and the mix of passengers and cargo. If these types of factors are not accounted for in the way inputs and outputs are measured and in the way exogenous factors are dealt with, performance measures can be misleading. A common way to deal with the problem of multiple products is to develop an index of overall output. In the airport case it is usual to refer to `workload units' according to which one passenger is regarded as being equivalent to 100 kilograms of cargo. The measurement of output is less of a problem when performance of individual functions within an airport are examined, but workload units can be misleading at the overall airport level since the method of aggregation is arbitrary from a productivity measurement point of view. Arbitrary methods of aggregation across di€erent classes of output can obscure key features of the production process (Jara Diaz, 1982). Turning to inputs, a choice can be made to measure labour in terms of hours worked rather than hours paid, but the former is dicult to measure (Doganis and Graham, 1987). In either case, it has to be recognised that there are substantial di€erences between classes of labour employed and it is desirable to weight each category by cost shares. Also, many airports have sub-contracted the provision of services and considerable attention must be paid to the inclusion of all inputs required to operate the airport. Capital input tends to be measured by airports using accounting information that embodies depreciation rules that can have little relation to actual use of resources. Also, capital costs as measured by airports can be dominated by interest costs and, in turn, these can be more a function of the way the airport has structured its ®nances. Di€erences in the circumstances of airports can invalidate simple comparisons of performance while changes in those circumstances over time can make it dicult to interpret trends in performance of a single airport. These issues of comparability have been investigated in terms of the

Measuring total factor productivity of airportsÐan index number approach

253

impact of composition of trac and airport size on a set of partial productivity measures for 25 airports in di€erent parts of Europe (Doganis et al., 1995). In brief, performance measures were regressed against independent variables such as work load units to re¯ect scale. The average labour cost per employee was introduced to capture the e€ects of variations in input prices, while the proportion of transfer and transit passengers, among others variables, was used to re¯ect the composition of business. The relative variability of passenger trac indicated the e€ect of departures from a uniform workload. This systematic investigation of partial productivity measures provides rich insights into the airport business, including limited evidence of economies of scale below 5 million workload units per annum. Labour productivity appears to continue increasing beyond the stage where economies of scale are fully exploited, raising some interesting questions about the nature of capital investments at large airports. Nevertheless, we are left with the problem of interpreting a set of measures, a problem that has been overcome in many other transport applications through the use of a measure of total factor productivity (TFP). We will see below that TFP indices can be derived which represent cost-eciency or cost-e€ectiveness; the di€erence being in the selection of the measure of output. However, we emphasis that TFP is not a measure of service-e€ectiveness and it must be supplemented using the types of measures discussed above. Provided sucient data on inputs, outputs and factors in¯uencing performance are available, it is possible to estimate TFP using parametric methods. This approach has the advantage that we can test for economies of scale and density and investigate the impact of variations on input and output prices on performance. We are aware that attempts have been made to estimate parametric TFP models for BAA's group of airports (Tolofari et al., 1990; Monopolies and Mergers Commission, 1991). However, the lack of suitable data for airports leads us to consider non-parametric methods using index numbers or data envelope analyses (DEA). All that is required with DEA is information about the quantities of inputs and outputs for a decision-making unit and an optimisation procedure (linear programming) is used to reveal an eciency frontier. Departures from this frontier de®ne relative eciency. DEA is applied to the airports sector by Gillen and Lall in this issue. We pursue an alternate approach using index number methods. In essence, the method compares an index of outputs to an index of inputs, allowing comparisons between the same airport in di€erent periods, or between di€erent airports. The weights in constructing the indexes can be the revenue shares and cost shares as indicators of the importance of outputs and inputs in the production process, the result being an axiomatic measure of performance (Diewert, 1991). However, the equivalence between parametric estimates of TFP and index number estimates can be established by imposing prior conditions on the nature of the technology and competition. Because of this, the weights in the index number approach have an economic interpretation whereas the weights in DEA measures of TFP are an outcome of an optimisation procedure. Although DEA is appealing because of its parsimonious data requirements, it has a built-in bias to overstate performance when the combined number of outputs or inputs is large relative to the number of decision making units (Charnes et al., 1989; Diewert, 1993). A further advantage of the index number approach is that it is possible to attribute changes in TFP to inputs and outputs. 3. THE TOTAL FACTOR PRODUCTIVITY APPROACH

A simpli®ed Tornquist index (Diewert, 1989) permits analysis of the productivity of a single airport over a period of time. However, there is considerable value in comparing the performance of di€erent airports allowing for the possibility of economies of scale and scope. In the case of BAA, for example, it has been claimed that each one percent increase in passenger movements results in only a 0.4% increase in stang (Monopolies and Mergers Commission, 1991) and others have claimed there are signi®cant economies of density (Graham and Dennis, 1993). A multilateral index allows greater scope to investigate these types of e€ects and makes it possible to compare airports. Provided sucient sample points are available, an output- or scale-adjusted Tornquist index can be estimated (Hooper 1987). If it is acceptable to assume that the impact of scale does not vary over time for a single airport, a less demanding approach is to estimate a scale-adjusted TFP index from data pooled from a set of airports. The de®nition of the appropriate multilateral index (Caves et al., 1982) is:

254

P. G. Hooper and D. A. Hensher

  ÿ  1 Xÿ ÿ  TFPk 1 Xÿ Rki ‡ Ri ln Yki ÿ ln Yi ÿ Rbi ‡ Ri ln Ybi ÿ ln Yi ln ˆ TFPb 2 2 ÿ  1 Xÿ ÿ  1 Xÿ Wkn ‡ Wn ln Xkn ÿ ln Xn ‡ Wbn ‡ Wn ln Xbn ÿ ln Xn ÿ 2 2

…1†

where k b i n Ri

= each individual observation, k=1, ..., K = base observation (a particular or average observation) = outputs, i=1, ..., I = inputs, n=1, ..., N = weights for each output Ri = arithmetic mean of output weights over all airports and years Wn = arithmetic mean of input weights over all airports Wn = weights for each input and years ln Yi = unit measure of output ln Yi = geometric mean of unit measure over all airports and years ln Xi = geometric mean of unit measure over all airports and ln Xn = unit measure of input years

An important feature of an index number is that it should be invariant to the selection of the base year. The TFP formula in (1) enables any pair-wise comparison of two ®rms in one year or years within a business. Also, it displays characteristicity, meaning that the weights enable symmetric treatment of all ®rms and/or time periods so that a comparison throughout a cross-section or panel of data is possible. A comparison between entities that is independent of the airport or year chosen as the base gives an index appeal in benchmarking. The appropriate input weights in the equation for the multilateral index are the contributions of each input to costs (i.e. cost shares). These are readily available, but the output weights should be the cost elasticities. Unless these elasticity weights are available from prior research they have to be estimated using a statistical cost function with the same properties as the aggregator function that was used to derive the weightings in the index number approach. In the majority of empirical studies the absence of such elasticities has led to the use of revenue shares as proxies. This is a valid assumption where a business displays constant returns to scale across all outputs (which negates the di€erential weighting) and where all outputs are priced at marginal cost (allowing output-speci®c revenue to have some correspondence with total cost). Also, this is a valid approach when an axiomatic approach is assumed, meaning that estimation of TFP only requires the assumption that a ®rm is commercial. Without a knowledge of cost elasticities it is not strictly possible to distinguish changes in TFP due to scale e€ects from other sources of productivity di€erences. Although there is some evidence that unit costs fall over some range in output, it does seem likely there are constant returns to scale for larger airports and that the key source of cost reductions arise through increases in trac density. Failure to allow for this, even ex-post, is a potentially serious concern. The revenue-weighted TFP index is more aptly regarded as a measure of gross total factor productivity (GTFP) and it has the limitation that it does not distinguish among sources of relative productivity. GTFP includes eciency and/or e€ectiveness gains that come about as a result of exploitation of scale economies and/or other in¯uences on production and costs, as well as gains due to true shifts in knowledge or our technical ability to produce things. By adopting some parametric (statistical) analysis, it is possible to analyse the sources of variation in GTFP across airports and/or over time. Ex-post adjustment can be approximated by running a simple regression model where the dependent variable is TFP and the independent variable is output, the latter required to identify the possible in¯uence of scale economies. This approach permits investigation of additional sources of variation including exogenous market characteristics (e.g. location), work practices, network design predetermined by government to satisfy community service obligations, and management practices. This decomposition of GTFP separates out variations in GTFP explainable by these variables. Caves et al. (1981) demonstrate that a Cobb-Douglas TFP regression is dual to a Cobb-Douglas neoclassical cost function. Thus

Measuring total factor productivity of airportsÐan index number approach

255

using TFP regressions one can obtain estimates of cost elasticities for a restricted cost function, thus permitting decomposition of TFP.

4. AN APPLICATION OF THE TFP APPROACH

The authors had access to information for six Australian airports for the four ®nancial years dating from 1989/91 and, for the purpose of exploratory analysis, used these data to estimate the multilateral index of TFP de®ned above. We measured output as de¯ated income, with a distinction between aeronautical and non-aeronautical revenue. Since this is relying upon demand-side information, the implication is that the resulting TFP index is best considered as a measure of cost-e€ectiveness. The resulting aggregate quantity re¯ects all sources of income received by airports from airlines, passengers, and other users of airport services. De¯ating revenues by an output index is an acceptable method of forming an output quantity index (Oum et al., 1992). The de¯ator for aeronautical output was obtained from Prices Surveillance Authority (1993), Table K.1, de®ned as the aggregate price index for aeronautical income. The non-aeronautical component of aggregate output was calculated by de¯ating the non-aeronautical income by the Australian Bureau of Statistics (ABS) implicit price de¯ator for gross national expenditure. Implicit in the de®nition in the current context however is that the airport is pricing eciently (i.e. at competitively established prices) but, since monopoly pricing is an issue of concern, it is problematic to derive an output measure from income. Indeed better measures of output quantity would have been landings for aeronautical output and passenger plus meetergreeter throughput and the volume of cargo handled for non-aeronautical output, and gate service hours provided for intermediate output. Revenue (shares) would only be used in weighting outputs (although once again it is assumed that prices re¯ect marginal costs). The perpetual inventory method used in the majority of TFP studies was adopted in the current study to obtain a measure of capital input. This involves measuring the total stock of capital, the ¯ow of services from the capital stock, and the price associated with the ¯ow of capital services. Given that airport infrastructure has a long life, the matching of the assets annualised value against the output supplied will be very sensitive to the timing of the investment over the period in which productivity is being measured. In the early years of an asset's life it is likely that excess capacity will prevail and hence show up as a contributing factor to low annual productivity. In the later years of an asset's life it might show up through an impact on high levels of congestion and hence a shortage of capacity which can reduce output and hence a€ect productivity in a di€erent way. Further allowance has to be made for additions to the capital stock in the construction phase during which there is no increment to output levels. This is one of the reasons why TFP is potentially limiting as a forecasting tool. Its greatest value is for monitoring performance over long periods of time with careful identi®cation of the sources of variation in gross TFP. Indivisible capital inputs and the timing of investments are important in¯uences of variation in TFP over time. The current data set covers a period when there were no major capital investments in runways and terminals. Also, the airports were operated by the same organisation so that there were no inconsistencies in accounting practices. Labour input was measured as expenditure on wages and salaries in constant dollar values (1988/89). The labour de¯ator is the ABS wages series (ABS Cat. No. 6302). The input de¯ator for non-capital and non-labour was an implicit de¯ator for gross national expenditure. Residual expenditures after accounting for labour and capital were expressed in constant values and introduced as `other' inputs. Gross TFP levels are presented in Table 2 and Fig. 1. Potentially the rankings can be misleading where the gross TFP indices are very close, but they provide a broad indication of the relative productivity of each airport in each year. In any comparison of the rankings of each airport over time it is important to identify the actual level of the TFP index, because some of the operators have indices with very similar values. It does not take a very large change in output or inputs to change the adjacent rankings. It is important that this point is recognised, allowing for some degree of `grouping' of operators with very similar TFP indices in any interpretation of the ®ndings. A preferred measure is the ranking for the average over the last four years, also given in Table 2.

256

P. G. Hooper and D. A. Hensher Table 2. Gross TFP levels, 1988±89 for the six FAC airports

Year

Sydney

1988/89 1989/90 1990/91 1991/92 Average

0.81 0.88 0.81 0.90 0.85

(4) (3) (5) (2) (1)

Melbourne 0.72 0.66 0.65 0.74 0.69

(8) (11) (12) (7) (3)

Adelaide

Brisbane

0.43 0.45 0.53 0.61 0.50

0.71 0.71 0.75 1.00 0.79

(21) (18) (16) (14) (4)

(9) (10) (6) (1) (2)

Perth 0.44 0.44 0.48 0.64 0.50

(19) (19) (17) (13) (4)

Hobart 0.28 0.32 0.40 0.55 0.39

(24) (23) (22) (15) (6)

Notes: Brisbane has been assigned a score of 1.00 in 1991/92. Rankings in brackets.

Fig. 1. Gross Total Factor Productivity (capital: Perpetual Inventory Method).

Table 2 suggests that although Brisbane airport had the highest GTFP in 1991/92, its productivity in the three prior years was not as good in a relative sense as Sydney, which on average across the period has the highest GTFP. The di€erences are sucient to have con®dence in the top ranking of Sydney airport over this period of four years. Because of the ¯uctuations in adjacent years over such a short period, growth rates are not very informative for each year. Recognising that GTFP can be in¯uenced by a variety of factors, it is useful to eliminate sources of productivity di€erence that make meaningful comparisons problematic. We considered tonnes landed, total movements, total international and domestic passenger throughput, total employees, and labour costs. A partial correlation matrix showed that there was a very high correlation between all measures of output (including the overall output index), typically exceeding 0.92, and between the output measures and labour quantity and cost (typically partial correlations of 0.82 or higher). Under these conditions, we opted to run two regression models. The ®rst had the output index as an explanatory variable and the second included airport-speci®c dummy variables as the explanatory variables. The dummy variables in the second model represent the airport-speci®c mean e€ect of in¯uences that distinguish airports. We could not combine the airport dummy variables with the output index because of multicollinearity (i.e. too high a partial correlation between pairs of explanatory variables in contrast to their correlation with the dependent variable). The results in Table 3 are used to obtain an output-adjusted TFP index for each airport. The ratio of output-adjusted to gross TFP indices are plotted in Fig. 2. The most important result in Fig. 2 is the in¯uence that size of operation (as measured by the output index) has on variations in TFP. The larger airports (Sydney, Melbourne and Brisbane) have a reduced output-adjusted TFP and the smaller airports (Hobart, Adelaide and Perth) have a higher output-adjusted TFP, bringing the di€erences closer together across all the airports. This is an important ®nding (Table 4). Table 5 presents a summary of the output-adjusted TFP after normalising on the 'best' year and airport (Brisbane, 1991/92).

Measuring total factor productivity of airportsÐan index number approach

257

Table 3. Identifying di€erences in gross TFP, 24 observations (heteroskedasticity correction), dependent variable=TFP Explanatory variable Model 1 Constant Output index Adjusted R2 Model 2 Constant Sydney airport Melbourne airport Adelaide Airport Brisbane Airport Perth Airport Adjusted R2

(1,0) (1,0) (1,0) (1,0) (1,0)

Parameter-estimate

t-value

0.92 0.19 0.75

34.1 10.9

0.51 0.61 0.40 0.15 0.53 0.15 0.75

7.3 8.2 5.5 1.8 5.0 1.7

Fig. 2. Ratio of output-adjusted TFP to gross TFP. Table 4. A comparison of gross TFP and output-adjusted TFP Year

Airport

88±89 89±90 90±91 91±92 88±89 89±90 90±91 91±92 88±89 89±90 90±91 91±92 88±89 89±90 90±91 91±92 88±89 89±90 90±91 91±92 88±89 89±90 90±91 91±92

Sydney Sydney Sydney Sydney Melbourne Melbourne Melbourne Melbourne Adelaide Adelaide Adelaide Adelaide Brisbane Brisbane Brisbane Brisbane Perth Perth Perth Perth Hobart Hobart Hobart Hobart

Gross TFP

Output-adjusted TFP

Ratio of output adjusted to gross TFP

1.07 1.16 1.07 1.18 0.95 0.87 0.86 0.97 0.56 0.59 0.69 0.80 0.94 0.93 0.99 1.32 0.57 0.57 0.64 0.84 0.37 0.42 0.53 0.73

0.92 1.02 0.90 0.98 0.89 0.82 0.79 0.88 0.80 0.85 0.91 0.99 0.98 0.97 0.99 1.29 0.74 0.75 0.77 0.95 0.89 0.95 1.00 1.14

0.86 0.87 0.84 0.83 0.94 0.94 0.91 0.90 1.43 1.44 1.31 1.24 1.04 1.05 1.00 0.98 1.28 1.31 1.22 1.12 2.44 2.27 1.91 1.56

258

P. G. Hooper and D. A. Hensher Table 5. Output-adjusted TFP levels, 1988±89 for the six FAC airports

Year

Sydney

1988/89 1989/90 1990/91 1991/92 Average

0.72 0.79 0.70 0.77 0.74

(8) (3) (9) (5) (3)

Melbourne 0.69 0.64 0.61 0.68 0.66

(9) (12) (12) (10) (5)

Adelaide

Brisbane

0.62 0.66 0.71 0.77 0.69

0.76 0.76 0.77 1.00 0.82

(13) (11) (8) (5) (4)

(6) (6) (5) (1) (1)

Perth 0.57 0.59 0.60 0.73 0.62

(16) (15) (14) (7) (6)

Hobart 0.70 0.74 0.78 0.88 0.77

(9) (7) (4) (2) (2)

Notes: Brisbane has been assigned a score of 1.00 in 1991/92. Rankings in brackets.

When di€erences in the level of output are taken into account there is a signi®cant change in the overall rankings as well as a shift in the relative productivity between airports. In particular, on an output-adjusted TFP index, Brisbane is now the best performing airport, followed by Hobart, and then Sydney airport. Sydney was the best on gross TFP, attributable in large measure to its size. Although size alone is not the determinant of productivity di€erences, it clearly is a most signi®cant distinguishing feature of each airport. A larger airport participates in and provides a wider range of activities. The fact that the output-adjusted and normalised TFP indices in Table 5 have a range of 0.57 to 1.00 compared to 0.28 to 1.00 on the gross index (Table 2) supports this view. The output-adjusted indices of TFP improve for the smaller airports relative to the larger airports, with Hobart's ranking improving from 6 to 2. However, Perth airport does not do well, with the adjusted output e€ect resulting in a drop in rank from fourth to sixth. Other reasons for these changes can be conjectured and investigated in the future. 5. CONCLUDING COMMENTS

Corporatisation and privatisation of airports could result in a more commercial approach to management with resulting improvements in performance, but this cannot be evaluated purely on the basis of ®nancial indicators. The need to develop appropriate service and productivity indicators has been recognised and there is a small, but growing literature on the subject. Though there have been appeals to measure `overall productivity', there is little evidence that the tools of productivity measurement that have been applied in other parts of the transport sector have had serious application in the case of airports (Tretheway, 1995). This paper has pointed out the shortcomings of partial productivity measures and it has emphasised the need to distinguish between the various sources of overall productivity di€erences among airports. We regard our work as exploratory and recommend that more analysis be undertaken that should as a minimum consider a time period of 10 years. Such work should seek to improve the de®nition of output both for produced services and consumed services and it should resolve the issue of annualised cost of capital to a single de®nition. Major advancements would be the estimation of a multilateral TFP index approach across time and airports and a decomposition of the resulting gross TFP index to identify the in¯uence of economies of scale, scope and density and other contextual in¯uences such as support facilities. The current study, though, is indicative of the potential for and desirability of more detailed investigation. REFERENCES Ashford, N. (1994) Airport management in a changing economic climate. Transportation Planning and Technology 18, 57± 63. Bureau of Industry Economics (1994) International Performance Indicators. Aviation. Research Report 59, Bureau of Industry Economics, Australian Government Publishing Service, Canberra. Caves, D. W., Christensen, L. R. and Tretheway, M. W. (1981) U.S. trunk air carriers, 1972±1977: a multilateral comparison of total factor productivity. In Productivity Measurement in Regulated Industries, eds T. G. Cowing and R. E. Stevenson, pp. 47±76. Academic Press. Caves, D. W., Christensen, L. R. and Diewert, W. E. (1982) Multilateral comparisons of output, input, and productivity using superlative index numbers. The Economic Journal 92(March), 73±86. Charnes, A., Cooper, W. W., Sun, D. B. and Huang, Z. M. (1989) Polyhedral cone-ratio DEA models with an illustration application to large commercial banks. CCS Research Report 643. College of Business Administration, The University of Texas at Austin, Austin, TX. Dennis, N. (1994) Airline hub operations in Europe. Journal of Transport Geography 2(4), 219±233. Diewert, W. E. (1989) The Measurement of Productivity. Working Paper, Department of Economics, University of British Columbia, Vancouver, BC.

Measuring total factor productivity of airportsÐan index number approach

259

Diewert, W. E. (1991) The Measurement of Productivity in Regulated Industries. Discussion Paper No. 91-20, Department of Economics, University of British Columbia, Vancouver, BC. Diewert, W. E. (1993) Data envelope analysis: a practical alternative? Paper presented to a workshop on Measuring the Economic Performance of Government Enterprises, 12 February 1993, under the sponsorship of Swan Consultants Pty. Ltd., Canberra. Doganis, R. and Graham, A. (1987) Airport Management: The Role of Performance Indicators. Transport Studies Group, Polytechnic of Central London, London. Doganis, R., Lobbenberg, A. and Graham, A. (1995) The Economic Performance of European Airports. Research Report 3, Department of Air Transport, College of Aeronautics, Cran®eld University. Feldman, D. (1996) Commercial magnetism. Airline Business 12(12), 36±39. Forsyth, P. (1993) Corporatisation, privatisation and the deregulation of Australia's airports. In Transport Policy Workshop Perspectives 1993, eds R. Daniels and M. Nyathi, pp. 185±202. Institute of Transport Studies, University of Sydney, Sydney. Forsyth, P. (1984) Airlines and airports: privatization, competition and regulation. Fiscal Studies 14(3), 61±74. Gannon, C. and Shalizi, Z. (1995) The Use of Sectoral and Project Performance Indicators in Bank-Financed Transport Operations. Discussion Paper, The World Bank, Environmental Sustainable Development, Report TWU 21, Washington, DC. Gollin, A. (1996) Airport reform: eight years down the runway. Paper presented at the 20th Australasian Transport Research Forum, 28±29 August, Auckland. Graham, A. and Dennis, N. (1993) Factors a€ecting airport performance. Paper presented at the Regional Science Association International British Section, 24th Annual Conference, September 1±3, 1993. Haririan, M. and Vasigh, B. (1994) Airport privatization: procedures and methods. Transportation Quarterly 48(4), 393± 402. Hazel, R. (1995) Airport economics. In Handbook of Airline Economics, eds D. Jenkins and C. P. Ray, pp. 113±120. Aviation Week Group, a Division of The McGraw-Hill Companies. Hooper, P. G. (1987) Productivity change in transport: a survey. Transport Reviews 7(4), 341±367. Jara Diaz, S. R. (1982) The estimation of transport cost functions: a methodological review. Transport Reviews 2, 257±278. Lemer, A. C. (1992) Measuring performance of airport passenger terminals. Transportation Research A 26(1), 37±45. Mills, G. (1995) Airports: users don't pay enough Ð and now here's privatisation. Economic Papers 14(1), 73±84. Monopolies and Mergers Commission (1991) A Report on the Economic Regulation of the South-East Airport Companies (Heathrow Airport Ltd, Gatwick Airport Ltd and Stansted Airport Ltd.). Report presented to the Civil Aviation Authority, June 1991, Civil Aviation Authority, London. Oum, T. H., Tretheway, M. W. and Waters II, W. G. (1992) Concepts, methods and purposes of productivity measurement in transportation. Transportation Research A 26(6), 493±505. Prices Surveillance Authority (1993) Inquiry Into The Aeronautical And Non-Aeronautical Charges Of The Federal Airports Corporation. Matter No. PI/92/7, Report No. 48, Prices Surveillance Authority, Melbourne. Prices Surveillance Authority (1995) Regulation of Airport PricingÐIs the New Zealand Approach Applicable to Australia. Discussion Paper, Prices Surveillance Authority, Melbourne. Prins, V. and Lombard, P. (1995) Regulation of commercialized state-owned enterprises: case study of South African airports and air trac and navigation services. Journal of Air Transport Management 2(3/4), 163±172. Rovizzi, L. and Thompson, D. (1992) The regulation of product quality in the public utilities and the citizen's charter. Fiscal Studies 13, 74±95. Seneviratne, P. N. and Martel, N. (1991) Variables in¯uencing performance of air terminal buildings. Transportation Planning and Technology 16, 3±28. Seneviratne, P. N. and Martel, N. (1994) Criteria for evaluating quality of service in air terminals. Transportation Research Record 1461, Transportation Research Board, Washington, DC, 24±30. Tolofari, S. R., Ashford, N. and Caves, R. E. (1990) The Cost of Air Services Fragmentation. Report TT9010, Department of Transport Technology, Loughborough University, Loughborough. Tretheway, M. (1995) Book review (The Economic Performance of European Airports, R. S. Doganis, A. Graham and A. Lobbenberg). Journal of Air Transport Management 2(3/4), 207.