Low Cost Carriers and Airports Performance: Empirical Evidence from a panel of UK Airports.

Transcription

1 Low Cost Carriers and Airports Performance: Empirical Evidence from a panel of UK Airports. March 2011 Abstract During the last decade, the proliferation of Low Cost Carriers and the related huge increase in tra c has been the most visible e ect of the deregulation of the airline market in Europe. Little attention has been paid to how airports were a ected by the changes in the new institutional environment. In this study we model the total factor productivity (TFP) for a panel of the UK largest airports over the period and investigate whether the presence of LCCs has some impact on airports TFP. Empirical results are consistent with the hypothesis that conspicuous entry of LCCs on European markets has impacted positively on the vertical chain by facilitating airports productivity improvements. This result is robust to reverse causality issues associated with the possibility that most e cient airports are those that are more likely to attract LCCs. Di erent possible arguments may explain our results: tra c increases brought about by LCCs for a given installed capacity might have generated higher TFP; more e cient organizational models might have been adopted to meet LCCs operational requirements (short turnaround times); cost reductions might have been realized in order to lower charges and attract LCCs; competition from a larger number of airports induced by LCCs wider catchment areas (with respect to full service airlines) might have exerted further pressure toward TFP improvements. Keywords: Total Factor Productivity, Airports, LCCs. 1 Introduction and relevant issues In most of developed countries the airport industry has been subject to considerable changes regarding ownership and governance structure, regulatory settings and business models. The liberalization of intra-eu tra c introduced in late 1990s abolished entry barriers and induced the proliferation of low cost carriers (LCCs), thus favouring a considerable growth of air tra c 1. 1 The implementation of the European Single Market Program has also played an important role. 1

2 These changes induced several countries to rethink their airport policies in order to tackle the need for new investments in capacity and to deal with the new operating conditions brought about by the deregulated airline sector and, in particular, by the LCCs entry. The low-cost carrier business model shared by most LCCs is centred on stripping out and avoiding all the complexity costs associated with traditional full service carriers (FSCs). This business model has several notable features: (i) using a simple pricing structure with one passenger class and fares only covering basic transportation (with optional paid-for in- ight food and drink); (ii) relying on direct selling through Internet bookings with electronic tickets and no seat reservations; (iii) operating simpli ed routes to often cheaper, less congested airports (with point-to-point rather than hub-andspoke networks); (iv) employing intensive aircraft usage (typically with 25-minute turnaround times); (v) having employees working in multiple roles (e.g. ight attendants cleaning the aircrafts and acting as gate agents); and (vi) utilizing highly standardized eets (with a maximum of two di erent aircraft types). Furthermore, in contrast to the usually very complex pricing structures operated by full service airlines (with price discrimination on each ight across multiple service classes and booking classes, outclassing fences like minimum stay requirements and date/route change penalties), for the main LCCs segmentation of prices is essentially made on the basis of just two variables: rst, the date of booking, with the lowest fares generally available several weeks/months ahead of the scheduled departure date and then rising as the departure date draws closer (i.e. inter-temporal price discrimination, rewarding early reservations and with fares increasing as the plane lls up); second, the e ective demand for a speci c ight such that typically early morning, late night, and mid-week and o -season departures are lower priced with respect to peak-travel periods ights (Piga and Bachis, 2007). The impact of low cost airlines on fares 2 and passenger numbers in the deregulated European aviation market has been striking, being the major impact the willingness of passengers to reach new airports which are typically more distant from major cities with respect to traditional hubs; LCC have also brought service to second city airports and to airports with public service obligation (PSO) or with regional development goals. Several experts argue that airlines deregulation and LCCs proliferation generated a new highly competitive operating environment that signi cantly reduced airports market power. Such market power does not seem to derive from natural monopoly characteristics of technology 3, but is most likely to be associated with land scarcity, locational and competition issues. For example Oum and Fu (2008) suggest that "the actual market power of an airport depends largely on the following factors: airport capacity available in the region as compared to the rising demand for the capacity; airline market structure and competition at the airport and in the region; share of Connecting Passengers; inter-modal competition, especially between airlines and high speed rail; the extent and nature of competition with other airports, whose tra c catchments areas overlap signi cantly with the airport under consideration". LCCs countervailing power stems from the characteristics of the business which allows them to be more adaptable and to have substantial bargaining power in negotiating fees with airports. In fact, LCCs usually operate on point-to point network and do not o er connecting services, so that it is less costly for them to switch airports whenever it is the case; moreover, they tend to have wider catchment areas than Full Service Airlines (FSAs) and their choice of an airport in a region is quite 2 Fare reductions have been as high as 80% o the standard scheduled fares of the incumbent airlines. 3 Among others, Bottasso and Conti (2010) have shown that average costs curve is U shaped for the UK airport industry. 2

3 exible. Furthermore, by increasing passengers tra c, LCCs generate larger non-aeronautical revenue sources for airports, such as catering and shopping for services normally provided as in- ight services by full fare airlines, higher car renting especially for small airports and higher commercial revenues: the existence of complementarities between the demand for aviation services and the demand for concession services is another factor which might prevent possible abuse of airports market power 4. All cited factors allows LCCs to negotiate good deals with competing airports and even to obtain subsidies from airports: this has been the case for Ryanair and EasyJet which have emerged as the two dominant LCCs in Europe. 5 It is worth recalling that airlines deregulation has also caused a signi cant change and shift in risk so that airports now have incentives to negotiate long term contracts with carriers in order to share the risk of tra c losses. Most experts of the airports industry suggest that the new competitive environment faced by airports during the last decade has exerted a strong pressure on costs and charges reduction and productivity improvements. A recent study by Bel and Fageda (2010) examines factors determining airport charges. Using data for 100 large airports in Europe, they nd, among other things, that competition from other transport modes and nearby airports imposes some discipline on the pricing behavior of airports and that LCCs, and airlines with a high market share, seem to have a stronger countervailing power. Possible more general explanations are linked to the disciplinary e ect of competition, while others are more industry speci c and linked to LCCs business characteristics: an increase in tra c for a given capacity might imply a productivity improvements, if the composition of the tra c does not cause higher production costs; LCCs short turnaround times might have required more e cient organizational models; the willingness to guarantee low charges to LLCs might have required cost reductions and e ciency improvements; LCCs wider catchment areas (with respect to FSAs) increased competition from other, and possibly more distant, airports by augmenting catchments overlaps. Indeed several studies have analyzed airports e ciency with di erent approaches (Data Envelopment Analysis, Stochastic Frontier Analysis, Total Factor Productivity Analysis) and focusing on di erent performance measures 6, but none has investigated the role on LCCs entry and just a few have analyzed if competition from othe rairports is an important e ciency driver: Yuen Chi-Lock and Zhang (2008) study the e ects of competition and policy changes on Chinese airport productivity and nd that airports subject to higher competitive pressure are more e cient, where competitive pressure is measured by the distance of a particular airport with another nearest airport 7. Several factors have been usually considered as possible e ciency drivers by the empirical literature; we recall the average craft size, the percentage of international tra c, the percentage of cargo tra c, the existence of capacity constraints imposed by regulatory, environmental and investment funding concerns, the airport size, the share of non-aeronautical revenues (as a measure of diversi cation), the hub status, the stock ownership (listed or non-listed), the location (intercontinental gateway airports tend to handle larger aircraft), area population, the degree of outsourcing, airlines load factors. 4 In 2005, other operational activities generated about 70% of the UK BAA airports pro ts: in the case of Gatwick, pro ts generated by airport charges were even negative, as in the case of Manchester Airport. This however could also re ect the method used by airports to allocate costs, or it could result from cross subsidies from commercial services to aviation operation under a single till price cap system. 5 LCCs may be of interest to multi-airport companies who can use low-cost airlines to secure business for smaller airports. 6 Pestana Barros and Dieke (2008) Fung et al.(2008) provide exhaustive surveys of this empirical literature. 7 See also Zhang et al. (2001) on Chinese airports, who employ the Her ndahl index as a proxy for market competition; and Malighetti et a.l (2008) who construct indices of network connectivity and competitive pressure based on routes substitutability. 3

4 Since during the last decade there has been a trend towards the introduction of private sector participation in the airport sector; one of the issues which has been frequently analyzed is the e ect of di erent governance structures on airport e ciency. Among others we recall the study by Oum, Yan and Yu (2008) who found, for a sample of international airports, that majority private and fully publicly owned airports are the most cost-e cient, while mixed public-private owned airports, particularly with the majority public, are less cost-e cient. Moreover, given that governance changes have often been accompanied by the introduction of some forms of regulation, some authors have tested if di erent regulatory settings are associated with airports e ciency; for example Oum et al. (2004) found that the total factor productivity of a sample of international airports is greater under the dual-till price cap than under either the single-till price cap or single-till rate of return regulation. Given that changes in airport governance followed airline deregulation and LCCs entry, we believe that, when studying airports productivity, it is important to control for operating conditions in which airports perform, together with other possible variables which may a ect airports productivity; in particular, we believe that the presence of LCCs at airports may a ect airport opearating conditions in light of the peculiarities of LLCs business. In this study we analyze the productivity of a sample of the 24 largest UK airports observed over the period In particular, we adopt a two step estimation strategy where in the rst step we conduct robust estimates of Total Factor Productivity and in the second one we estimate an empirical model where productivity is considered as a function of the ownership structure and of variables which proxy for airports chacracteristics; furthermore, we include in the model a proxy for the importance of LCCs in each airport as given by the share of LCCs passengers on total airports passengers tra c; this variable should capture the strength of the pressure toward cost reductions and better organization faced by each airport and associated to the presence of LCCs, together with possible e ects of an increase in competition among airports generated by LCCs larger catchment areas. Given that the presence of LCC on UK markets has signi cantly increased over the last decade since intra-eu tra c was liberalized in the late 1990s 8, we argue their presence in an airport might be associated with low costs and higher productivity. Moreover, in our estimation strategy we also deal with the possibility that more productive airports are also those who mostly attract LCCs, thus generating possible bias in the estimate of the e ect of the LCCs presence on productivity associated to reverse causality issues. Other airports speci c characteristic, like the composition of the tra c and the weight of passenger tra c, are included in the model; moreover, we also control for possible e ects of localization by including multiplicative e ects of time trends with dummy variables related to six di erent geographic market as de ned by the O ce of Fair Trade and the Competition Commission and by our de nitions (see data section). We also note that these market trends variables might also proxy for the (time varying) levels of competition faced by airports in their local geographic market and help to better identify the link between the presence of LCCs and airports TFP. This kind of analysis has never been conducted before. Some empirical evidence on performance determinants for the UK airport industry does exist. In particular, BAA Airports performance has been the focus of one of the rst works on the UK airport industry conducted by Parker (1999). The author compares BAA airports e ciency scores, obtained with data envelopment techniques, before and after the privatization and does not nd any signi cant 8 The rst scheduled low-cost airline in Europe, Ryanair, commenced on the Dublin London route in In August 1987, the rst full year of deregulation on the Dublin London route, passenger numbers were 92% greater than in August 1985, the last full year of pre-deregulation policies. By the late 1990s there were almost 4.5 million passengers on the route compared to under l million in a market that was static for seven years before deregulation. The route became the second busiest in the world after Tokyo Taipei.(Barrett 2004). 4

5 di erence; moreover his analysis reveals the presence of economies of scale in airport operations but without specifying the relevant output range. Only recently there has been a renewed interest towards the economic analysis of the UK airport sector. Barros (2008) analyses the technical e ciency of a sample of 27 UK airports observed over the period by estimating a (translog) stochastic cost frontier which includes a trend variable. E ciency rankings suggest that the top positions are achieved by small airports, while the most important airports (Gatwick, Heathrow and Manchester) get the weakest positions; moreover costs are found to be increasing over the sample period at a decreasing rate. On the same sample Barros and Managi (2008) compute di erent Total Factor Productivity (TFP) indices and nd that the majority of UK airports did not improve e ciency over the sample period. Moreover, when e ciency scores are sorted by ownership it seems that privately owned airports perform better than publicly owned ones. Similar results are found in another work by Barros and Weber (2009) who estimate Malmquist indices of TFP and show that over the sample period UK airport became less e cient and experienced technological regress; however a clear relationship is not observed between ownership and productivity nor between regulation and productivity. A Data Envelopment Analysis conducted by Assaf (2009) on a sample of 27 airports observed in 2007 suggests that small airports are operating in a increasing returns to scale region, while large ones seem to have reached full capacity and some have exceeded their optimal size. Finally, Bottasso and Conti (2011) provide a detailed analysis of the cost structure of the UK airport industry observed over the period and suggest the existence of scale economies for tra c levels ranging between ve and fourteen million passengers ; moreover, observed cost di erentials in favour of private airports are found to disappear at the end of the sample period. 2 Data and Institutional setting The present structure of the UK airport industry has been de ned by the 1986 Airports Act which realized the privatization process announced in the 1985 White Paper on airports policy. At that time nearly all airports were owned by the public sector, either by the British Airport Authority (BAA) or by local governments 9. In 1987 all the share capital of BAA were transferred to the private sector by otation and BAA became BAA plc. Between 1993 and 1999 many (but not all) local government owned airport assets were also sold; moreover, ownership by unquoted private companies (BAA plc was the exception) has not prevented several of the airports changing hands since they were rst transferred to the private sector, some several times (see table 2). The UK airport industry is thus a mixed private-public sector industry but dominated by the private ownership of assets with a strong presence of foreign investors. 10 At the time of privatization of BAA in 1987 it was assumed that major airports had signi cant market power linked to their natural monopoly characteristics which called for the need of economic regulation. Airport designated to be regulated after the 1986 Airports Act were Manchester, Heathrow, Gatwick and Stansted. The sector speci c regulator is the Civil Aviation Authority (CAA) and pricecap regulation was applied to airports charges with regulatory reviews scheduled every ve years and 9 BAA had the largest share of the market as it owned seven airports, four in Scotland and three in the London region (Heathrow, Gatwick and Stansted), while the other largest local authority airport was Manchester. 10 When BAA was privatized the government capped the amount of shares that any one shareholder could hold to 15 per cent but, following a ruling by the European Court of Justice that this restriction impeded the free movement of capital in the European Union, it was removed in Soon afterwards the ECJ s ruling BAA group was taken over by Ferrovial. 5

6 caps set in order to allow for a rate of return consistent with the cost of capital 11. This approach is supposed to provide the regulated rm with incentives for e ciency because possible savings in expenditure exceeding those anticipated at the price revision may enhance company pro tability; however, it has been noted (Littlechild 2003) that pressures on regulators have led, in practice, to a convergence of the RPI X approach with rate-of-return regulation, the latter generally regarded as encouraging overcapitalization and excess capacity. Turning to the data description, the dataset used in this study consists of a balanced panel of the major 24 UK airports observed over the period The main source of data is the statistical series "The UK Airport Industry" published each year by the Centre for the Study of Regulated Industries at the University of Bath. Additional information was obtained from airports accounts, from UK National Statistics and from the UK Civil Aviation Authority (CAA). The summary statistics reported in Table 1 suggest that in the UK airport industry there is high variability in terms of size. In particular, the size distribution of airports is highly skewed to the right: for example, if we consider the airport total number of passengers (pax), about 78% of the observations have a lower value than the mean, and similarly in the case of the work load units (wlu), de ned as one passenger or 100 kg of cargo and mail. The skewness of the size distribution is driven by the presence of two very large airports, notably Gatwick and, in particular, Heathrow. Similar conclusions can be drawn by observing data on de ated total revenues (tot rev), which are computed as the sum of charges and commercial revenues and on our output measure (y), computed as the weighted average of working load units (wlu) and other operating revenues, each weighted by its share in total revenues. The revenues shares deriving from commercial activity (share other rev) and from charges (share charge rev) are approximately equal (around 50%) and stable over time. Standard deviation suggest a certain degree of variability across airports, with Blackpool airport displaying the largest share of commercial revenues (75% on average). Concerning this issue, many studies ignore the non-aeronautical service outputs (commercial services including concessions,rentals,car parking, etc.) airports produce. Since airports inputs are not usually separable between the aeronautical and the non-aeronautical service activities,any productivity measure which excludes the non-aeronautical services outputs would bias empirical results very seriously against the airports generating a larger portion of their revenues from commercial services. Input variables used in order to estimate airports TFP are labor (l), materials (m) and capital (k). Labor is represented by the number of employees; materials are proxied by the cost of non labor inputs (operative costs other than labor costs) de ated with a speci c price index which has been computed as the weighted average of i) the Construction Output Price Index -COPI- (as a proxy for the price of materials); ii) a price index of water, gas and electricity services and iii) the RPI (as a proxy for the price of other services paid by airports). 12 The perpetual inventory method was used to reconstruct a series of the capital stock. In particular, for many airports we have exploited the capital stock revaluations that have occurred over our sample period and we have considered the revaluation 11 As far as the airport industry is concerned, there is a debate on the choice of revenues to include in the determination of the cap: when both aeronautical and non aeronautical revenues are taken into account (single-till), capacity constrained airports tend to lower aeronautical charges in order to remain under the cap, since more revenue is made on the non aeronautical side. In this situation charges are lowered when e ciency implies they should be raised and airlines will capture location and scarcity rents that could be earned by the airport 12 The weights have been proxied by the cost shares of materials, utility bills and residual services in non-wage costs as reported in BAA airports statutory accounts, respectively. 6

7 year as the baseline. 13. Overall time pattern of main interest variables con rms the positive trend of aviation markets which took place over the last decade; for example total revenues increased by approximately 15% between 2002 and 2005 and the number of passengers increased by about 20%. Other variables used in the empirical analysis are the percentage of international passengers (pip) and the percentage of passenger tra c (ppt), obtained as the number of passengers divided by wlu. These variables re ect the composition of airports tra c and depend largely on the geographic location of the airport. Both variables may in uence airports performance since they are associated with di erent amounts of revenues and costs. As far as the ownership de nition is concerned, we have de ned an airport as "public" (pub) whenever the majority ownership belongs directly or indirectly to one or more municipalities; in turn, an airport has been de ned as "mixed" (mix) whenever private investors retain the majority control but local councils hold a substantial minority stake (i.e. bigger than 20-30%) or when the airport is under public ownership but the management is fully delegated to private investors (e.g. Luton after 1998); nally, an airport that does not belong to either groups is de ned as "private" (priv). As of 2002, 9 airports in our sample were publicly owned, against 12 under private ownership and three which are de ned as mix; in turn, in 2005, 13 airports were privately owned, against 6 and 5 under public and mixed ownership, respectively. Table 2 reports, for each airport, its ownership status and its changes over the sample period whenever they occurred. Turning to variables which re ect the presence of LCC on analyzed airports, the main data source is the Civil Aviation Authority (CAA) which provided secondary data on the tra c for all the routes and all the airlines ying from all the main UK airports in the sample to the following European countries: Austria, Belgium, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal and Spain, Czech Republic, Norway, Sweden, Switzerland as well as the UK, whose domestic routes were also considered. The dataset includes monthly ights information from June 2002 up to, and including, June 2005, for a total of 37 months. For each combination of company, route and departure period (i.e., month/year), the CAA provided the number of monthly seats, the number of monthly passengers and the monthly load factors. These were broken down at the ight identi cation code level, that is, for each ight operated by all the airlines in a given month and route, so that the total number of monthly ights could also be obtained. Then, for each airline, tra c statistics were aggregated at both the route and at the airport level. This allowed the possibility to create proxies for the presence of each airline at the airport level and the market level, where a market can contain a wider set of airports located in a given geographical area. In particular, we were interested in measuring the presence of the three largest Low-Cost Carriers (hence, LCCs) operating in British airports: Ryanair, EasyJet and FlyBe. Therefore, we calculated the sum of the total number of ights and passengers of these three carriers, and then obtained their share on the total number of passengers or ights for each airport or market. To this purpose we divided the airports in our sample into six di erent geographic markets, partly on the basis of considerations developed by both the O ce of Fair Trade and the Competition Commission in the recent BAA case, and partly on our assessment based on distances between airports (two hours driving time). The markets identi ed by the Competition Commission and the OFT studies were the London area and Scotland; moreover we have identi ed four additional markets so that we have six markets in total: Mkt 1: Heathrow, Gatwick, Stansted, London City, Luton, and Southampton, 13 We have assumed a constant depreciation rate of 4.5%. The resulting capital stock series was de ated with the COPI index. 7

8 Norwich; Mkt 2: Manchester, Humberside, Nottingham, Birmingham, Blackpool, Leeds and Liverpool; Mkt 3: Glasgow, Edinburgh and Aberdeen; Mkt4: Bournemouth, Cardi, Bristol and Exeter; Mkt 5: Newcastle and Durham; Mkt 6: Belfast. Table 3 indicates that the share of passengers of the 3 largest LCCs in an airport tends to be very similar in markets 1, 2 and 5, where about 40-50% of passengers on both UK domestic and inter- European routes y with one of the 3 largest LCCs. A largest presence is recorded for market 4, while in market 6 the preponderance of LCCs presence is quite striking, since the share of LCCs passengers in an airport reaches values up to 80%. However, the relevant variation across airports in a market, as revealed by the measures of standard deviation in Table 3, indicate that the presence of LCCs is quite heterogeneous across airports in a given market, and that therefore it is an important factor that needs to be controlled for in the empirical analysis. Finally LCCs presence has increased in most markets over the sample period; in particular for market 4 their passengers share has grown by 50%. 3 Empirical results In order to analyze if changes in the operating environment generated by the entry of LCCs on markets have some impact on airport productivity levels we decided to adopt a two step approach. In the rst step we obtained di erent measures of airports Total Factor Productivity by estimating a Cobb Douglas production function with di erent econometric techniques together with standard Ordinary Least Squares. Simultaneity and endogeneity issues involved in production functions estimate have been accounted for by applying both GMM techniques and the Levinshon and Petrin (2003) approach. 14. Three di erent measures of TFP obtained with the cited econometric techniques are labeled as T F P 1 (OLS), T F P 2 (LP ), T F P 3 (GMM) in Table 4 and become the dependent variable of the second step of our analysis: TFP 1 has been obtained with OLS, TFP 2 with the Levinshon and Petrin method and TFP 3 with GMM estimator. 15 In second step regressions we allow TFP to depend on di erent variables which account for ownership structure and tra c composition and on the presence of LCCs in each airport. On the basis of the de nitions of public and private airports given in the data section we included in the model two ownership dummies (dpriv and dpub), while airports with mixed ownership are considered as the reference group. The standard argument suggested by the literature on the e ects of privatization on rms performance is linked to possible heterogeneity in managers objective function associated with di erent ownership forms; moreover, di erent governance structures generate di erent agency relationships among stakeholders which may as well in uence productivity. Tra c composition is included as a possible variable a ecting productivity, since di erent kinds of tra c have di erent implications on airports costs and revenues. Usually, high percentages of international passengers (pip), and lower percentages of air cargo (high ppt), are expected to increase production costs and lower total factor productivity given that international tra c and passengers tra c require more services, space and resources, with respect to domestic tra c and cargo tra c. We have included in the empirical model the share of passengers of the 3 largest LCCs in each airport (labeled as LCCs in Table 4) as a variable proxying for changes in operating conditions faced by airports. We argue that LCCs presence in an airport might be associated with lower costs and higher productivity, since such airlines requires low charges and high e ciency in operating conditions 14 See Van Biesebroeck (2007) for details on the methodologies. 15 We have also estimated a translog pruduction function, which provides a second order approximation to an unknown functional form. Results are very similar to these presented in the paper and are available from the authors upon request. 8

9 at airports; moreover, their presence enlarges the potential for competing airports since they have wider catchment areas with respect to FSAs. The estimated equation can be written as: ln T F P it = + ln T F P it 1 +LCCs it +dpriv it +dpub it +pip it +ppt it + j (tmkt j )+e i +u it (1) where e i re ects a time invariant airport speci c component term capturing non observable airport characteristics which a ect TFP; u it is an IID random component with zero mean, uncorrelated with itself and with the regressors. In equation 1 we have included the rst lag of TFP in order to capture persitence in TFP levels and to include some dynamics in the model In order to take into account regional and/or market level heterogeneity we included a full set of interaction terms between market dummies and a time trend (t mkt j ) that proxy for time varying unobserved heterogeneity in the geographic market j where the airports operate. The inclusion of these market-trend interactions contorl for shocks that may hit the markets where the airports are located, as well as for di erences in the economic and competitive environment they face: as a result, they should capture both the regional speci c business cycles and the (changes in) intensity of competition in each particular market, or in other locational related variables (e.g. population density, transport infrastructure, income wealth of the local population, economic development and so forth). Indeed, a recent detailed analysis conducted by the CAA and by the UK Competition Commission (2008) reveals that there is signi cant potential for airport competition in both the London and the North West (Manchester) regions of the UK as a result of the presence of signi cant overlaps in catchment areas for passenger tra c 16 ; such characteristics may be captured by market interaction dummies. Our empirical strategy for the second step of the analysis has been to rst estimate the model by OLS on pooled data. Table 4 reports OLS results obtained by employing three di erent estimated values of TFP as dependent variable on the basis of estimation results from the rst step. OLS estimates provide a very high and signi cantly positive coe cient for lagged TFP and a positive and signi cant coe cient for LCCs passengers share of about 0.2, thus suggesting that the presence of LCCs operating in an airport my have a positive impact of TFP; however none of other explanatory variables turn out to be signi cant. Indeed OLS techniques do not account for problems of endogeneity which in our model are generated by unobserved heterogeneity, feedback e ects and measurement errors. Given that some of our explanatory variables are potentially correlated with unobservable timeinvariant, airport level component represented by e i, we remove this rst source of endogeneity by appropriately transforming the model ( rst di erence and within transformation). Another source of endogeneity is associated to the presence of the lagged dependent variable and the share of passengers of the 3 largest LCCs in each airport among explanatory variables. The presence of these variables makes both the within group and the rst di erence estimators biased and inconsistent. In this case the standard solution is to rst di erence the data and use longer lags of ln T F P t 1 (e.g. ln T F P t 2 ) as instruments (for the lagged dependent variable) in a GMM fashion, along the lines proposed by Arellano and Bond or Arellano and Bover; moreover suitable instruments for LCCs need to be employed. However, when the number of cross sections is small, as in our case where N=24, these estimators can 16 Nevertheless the CC suggested that common ownership of three London airports by BAA was likely to adversely a ect competition and in a successive report (2009) ruled BAA would have to divest itself of airports in either Gatwick or Stansted, and either Edinburgh or Glasgow. In December 2009 BAA sold Gatwick to Global Infrastructure Partnership. 9

10 be severely biased and imprecise as their consistency properties hold for large cross sectional dimension; furthermore, the panel is also short (T=5), and using longer lags as instruments reduces considerably the observations actually used in the estimation. For this reason we have followed a di erent strategy, i.e. presenting di erent estimation methodologies that separately tackle the endogeneity problem of ln T F P t 1 and of LCCs and comparing the results as a robustness check. As far as the share of LCCs passengers is concerned, if we assume the current period disturbances to be uncorrelated with past and current values of LCCs, but potentially correlated to its future values, we can apply an IV method, after transforming appropriately the model, and use LCCs it 1 as an instrument for LCCs. In fact, there might reasons to believe that, even after controlling for time invariant unobserved airport heterogeneity and market speci c time trends, a past shock to productivity might be correlated with entry of low cost airlines as of time t: for example, poor past performance of an airport might have discouraged entry or expansion of LCCs in that airport; in this case, LCCs would not be strictly exogenous but just predetermined. We follwed the above estimating approach and applied an IV method after rst di erencing the model. Table 4 show results labeled as "IV FD": the lagged TFP variable loose its signi cance in all estimates, while the impact of the presence of LCs is always positive and of higher magnitude with respect to OLS estimates, ranging between 2.5 and Interestingly, the dummy related to public ownership of airports assumes positive values, thus implying that airports whose majority ownership belongs directly or indirectly to one or more municipalities are more e cient than those where private investors retain the majority control but local councils hold a substantial minority stake. This result is consistent with Oum, Yan and Yu (2008) ndings on a sample of international airports. The assumption of predetermination on LCCs may in some cases be violated: in fact, if a current shock to TFP leads to successive entries of low cost airlines, the assumption of predetermination would fail and we would have to use longer lags as valid instruments (t-2), which is not advisable in such a short and small panel; furthermore, if low cost airlines forecast future TFP and decide to entry or not (or to grow) in a particular airport on the basis of these forecasts, then LCCs would be fully endogenous and lagged values would no longer be valid instruments (because current values of LCCs would be correlated with future shocks to TFP). An alternative approach is therefore to search for "external instruments". Estimates reported in Table 4 and labeled as "IV FE" are obtained with IV techniques applied after within transformation of the model, where the LCCs has been instrumented with value added growth in the NUTS1 region where the airport is located and with the average share of low cost carries in markets di erent from that where the airport is located. The rationale of using value added growth in the region as an instrument is that in areas with strong economic growth low cost carries might nd pro table opportunities to enter the market or to expand, while it is unlikely that regional value added growth is correlated with an airport TFP, especially after controlling for airport xed e ects and market speci c time trends. Moreover, using the share of LCC in other markets rests on the assumption that LCCs might be correlated with developments in other markets which in turn should not be correlated with the TFP of an airport located in a di erent market. Regression results broadly con rm those obtained with IV FD as far as lagged TFP is concerned; the ownership dummy for public airports is no longer signi cative and coe cients of the LCCs variable are found to be lower with respect to the IV FD case (around 0.9). Turning to the endogeneity issue, and to the associated problem of nite sample bias, stemming from the presence of the lagged depend variable in the empirical model, we tried to tackle it by applying the least square dummy variable (bias) corrected estimator (LSDVC) proposed by Kiviet (1995), which appears particularly suitable when the conditions underlying the use of the GMM-di of 10

11 Arellano and Bond are unlikely to be met (Judson and Owen (1999)). However this estimator requires strict exogeneity of the other regressors so that it does not take into account possible bias arising from the presence of other endogenous explicative variables. Estimates obtained by applying LSDVC estimator are shown in Table 4: coe cients of lagged TFP are positive and signi cative and range between higher values obtained with OLS and lower values from IV estimates 17. Since Kiviet estimates on lagged TFP are closer to those provided by the rst di erence IV estimator, it may be the case that in this empirical application, the bias of the latter is not too important. LSDVC estimates con rm the positive impact of LCCs on productivity with a similar magnitude to IV FE estimates; moreover the positive coe cient for the public ownership dummy is con rmed. Overall results suggest that the presence of LCCs has a positive impact on airports total factor productivity in all estimated models. Our preferred model is the one estimated with Fixed E ect IV, as it accounts for endogenity issues by relying on external instruments and provides estimates for the parameter of the lagged TFP similar to those obtained with LSDVC estimator. The average value for the coe cient of the LCCs variable obtained with FE IV is about 0.9, which implies that a standard deviation increase of the share of passegngers of LCCs in an airports would have a contemporaneous positive impact on TFP of about 27%; moreover, its long run impact, obtained by diving the contemporaneous e ect by 1 minus the average coe cient of lagged TFP, results to be slightly higher and close to 30%. Finally, there is partial evidence of higher productivity of public airports with respect to those characterized by mixed ownership. 4 Conclusions The last decade has been a period of great changes for the European airport industry. Airports had to adapt to new operating conditions brought about by the deregulation of the airline industry, which has favoured proliferation of LCCs and huge increases of tra c. The new environment is characterized by higher pressure on airports toward costs reduction and productivity improvements necessary to attract LCCs and to compete on enlarged relevant markets. Changes in ownership and governance structure, di erent organizational models and new vertical relationships between airports and airlines are some of the adjustments observed in the European airport industry and, in particular, in the UK case. In this study we analyze if Total Factor Productivity estimated for 24 of the largest UK airports is a ected, among other things, by the presence of LCCs which has been proxied by the share of LCCs passengers on total passengers tra c at each airport. The empirical analysis has been conducted with a two step approach, where TFP has been estimated with di erent econometric techniques in the rst stage. In the second step of the analysis a dynamic panel model has been estimated: rst step tted values of TFP have been regressed on a set of explanatory variables which include ownership forms, airport speci c characteristics and the share of LCCs on total passengers. Empirical results are consistent with the hypothesis that LCCs entry on European markets has stimulated airports productivity improvements together with fare reductions for passengers. This 17 In dynamic panel models it is well known that the xed e ects will downwardly bias the coe cient of the lagged dependent variable. This is called Hurwicz bias (after Leonid Hurwicz, who noted this e ect for time series half a century ago). It is sometimes called Nickell bias, after Stephen Nickell who rediscovered and applied it to panels 20 years ago. 11

12 result is robust to reverse causality issues associated with the possibility that most e cient airports are those which are more likely to attract LCCs. Di erent possible arguments may explain our results: tra c increases brought about by LCCs for a given installed capacity might have generated higher TFP; more e cient organizational models might have been adopted for meeting LCCs operative standards (short turnaround times); cost reductions might have been realized in order to lower charges and attract LCCs; competition from a larger number of airports induced by LCCs wider catchment areas (with respect to FSAs) might have exerted further pressure toward TFP improvements. Possible policy implications of our ndings are linked to regulatory issues and, in particular, add some elements to the debate on the opportunity of regulating the industry. Some authors argue that there is no need for regulation since possible abuse of airports market power is prevented by di erent factors, like the existence of complementarities between the demand for aviation services and the demand for concession services. More in general, market power might have been reduced by the increasing degree of competition between airports which took place during the last decades as a consequence of the liberalization of European air transports and of the growing importance of Low Cost Carriers. Our ndings, coupled with the evidence of diseconomies of scale for high tra c levels which undermine the natural monopoly argument for economic regulation, reinforce the view against the need for regulation of the airport industry. 12

13 References [1] Abrate G. and Erbetta F. E ciency and Patterns of Service Mix in Airport Companies: An Input Distance Function Approach. Transportation Research Part E, forthcoming. [2] Armstrong M., Cowan S. and Vickers J. Regulatory Reform: Economic Analysis and British Experience. MIT Press: Cambridge; [3] Assaf A. Bootstrapped Scale E ciency Measures of UK Airports. Journal of Air Transport Management 2009;1-3. [4] Barrett, S.D. Product di erentiation supplier rents and sustainability in the deregulated European airline market a case study. Journal of Air Transport Management 2000, 5, [5] Barrett S. How do the demands for airport services di er between full-service carriers and low-cost carriers? Journal of Air Transport Management 2004, 10, [6] Barros C. P. and Weber W.L. Productivity Growth and Biased Technical Change in UK Airports. Transportation Research Part E 2009; 45, 4, 64. [7] Barros C.P. and Manage S. Productivity Change of UK Airports: Technical University of Lisbon, Working Paper 2008, 22. [8] Barros C. P. The Technical E ciency of UK Airports. Journal of Air Transport Management 2008; 14, [9] Bel G. and Fageda X. atization, regulation and airport pricing: an empirical analysis for Europe. 2010, Journal of Regulatory Economics, 37: [10] Bernstein J.I. and Sappington D.M. Setting the X-factor in Price Cap Regulation Plans. Journal of Regulatory Economics 1999; 16, [11] Caves D., Christensen L.R. and Swanson J. Economic Performance in Regulated and Un- Regulated Environments: A Comparison of Us and Canadian Railroads. Quarterly Journal of Economics 1981; 96, [12] Competition Commission. BAA Market Investigation: Emerging Thinking. London [13] Doganis R., Graham A. and Lobbenberg A. The Economic Performance of European Airports. Department of Air Transport Research Report 1995; 3, Cran eld University. [14] Friedlaender A., Berndt E., Wang Chan J. and Vellturo C. Rail costs and capital adjustments in a quasi-regulated environment. Journal of Transport Economics and Policy 1993; 27, 2, [15] Fung M., Wan K., Hui Y. and Law J. Productivity Changes in Chinese Airports Transportation Research Part E 2008; 44, [16] Kiviet J.F. On Bias, Inconsistency, and E ciency of Various Estimators in Dynamic Panel Data Models. Journal of Econometrics 1995, 68: [17] Judson R.A. and Owen A.L. Estimating Dynamic Panel Data Models: A Guide for Macroeconomists. Economics Letters

14 [18] Littlechild S. Re ections on Incentive Regulation. Review of Network Economics 2003; 2, 4, [19] Malighetti P., Martini G., Paleari S. and Redondi R. The e ciency of European airports: Do the importance in the EU network and the intensity of competition matter? Working Paper University of Bergamo, 2008, n.4. [20] Martin J.C., Roman, C. and Volte-Dorta, R. A Stochastic Frontier Analysis to Estimate the Relative E ciency of Spanish Airports. Journal of Productivity Analysis 2009; 31, [21] Newbery D.M. atization, Restructuring and Regulation of Network Utilities. MIT Press [22] Niemeier H.M. Gateway Airport Investment and Development of Airline Services for a Global Economy. OECD Joint Transport Research Centre, forum 2009, Leipzig. [23] O ce of Fair Trade. UK airports. Report on the market Study and Proposed Decision to Make a Market Investigation Reference, London [24] Oum, T.H., Zhang, A., Zhang, Y., Alternative forms of economic regulation and their e ciency implications for airports. Journal of Transport Economics and Policy 2004, 38, [25] Oum T.H. and Fu X. Impacts of Airports on Airline Competition: Focus on Airport Performance and Airport-Airline Vertical Relations. OECD Joint Transport Research Centre, Discussion Paper 2008; 17. [26] Oum T.H., Yan J. and Yu C. Ownership Form Matter for Airport E ciency: A Stochastic Frontier Investigation of Worldwide Airports. Journal of Urban Economics 2008; 64, [27] Parker D. The Performance of BAA Before and After atization. Journal of Transport Economics and Policy 1999; 33, 3, [28] Pels E., Nijkamp P. and Rietveld P. Ine ciencies and Scale Economies of European Airport Operations. Transportation Research Part E 2003; 39, [29] Piga, Claudio A. and Enrico Bachis. (2007). Pricing Strategies by European Traditional and Low- Cost Airlines: or, When Is It The Best Time To Book On Line?, in Darin Lee (ed.), Advances in Airline Economics, Volume 2: The Economics of Airline Institutions, Operations and Marketing, Elsevier: Amsterdam, ch. 10, [30] Salazar de la Cruz F. A DEA approach to the airport production function. International Journal of Transport Economics 1999; XXVI, 2, [31] Starkie D. and Thompson D. atizing London s Airports, Institute for Fiscal Studies, London [32] Starkie D, A New Deal for Airports?, in Robinson C. (ed.), Regulating Utilities: New Issues New Solutions, 2001 IEA, London. [33] Starkie D. The Airport Industry in a Competitive Environment: a United Kingdom Perspective. OECD Joint Transport Research Centre 2008; Discussion Paper, 15. [34] Yuen Chi-Lock and Zhang A. E ects of competition and policy changes on Chines airport productivity: An empirical investigation. Journal of Air Transport Management 2008,