ARTICLE IN PRESS Research in Transportation Economics xxx (2012) 1—7 ELSEVIER Contents lists available at SciVerse ScienceDirect Research in Transportation Economics journal homepage: www.elsevier.com/locate/retrec Evaluating the long term impacts of transport policy: The case of passenger rail privatisation John Preston3*, Dawn Robinsb 'Transportation Research Group, Civil, Maritime and Environmental Engineering and Science, Faculty of Engineering and the Environment, University of Southampton, Highfield, Southampton S017 IB}, UK h Geography, Faculty of Social and Human Sciences, University of Southampton, UK ARTICLE INFO ABSTRACT Article history: Available online xxx Keywords: Passenger rail Privatisation Welfare analysis Britain's national rail system was 'privatised' as a result of the 1993 Railways Act, with most of the organisational and ownership changes implemented by 1997. This paper examines the long term impacts of these changes. A key issue when examining long term changes is that of the counterfactual — what would have happened if the changes had not occurred? A simple econometric model of the demand for passenger rail services was developed and used in conjunction with extrapolative methods for key variables such as fares and train km to determine demand-side counterfactuals. Extrapolative methods were also used to determine counterfactual infrastructure and train operation costs. Although our results are sensitive to the assumptions we have made concerning the counterfactual they suggest a number of impacts. Since privatisation, rail demand has grown strongly but our analysis indicates that transitional disruptions suppressed demand by around 9% over a prolonged period (1992/3 to 2005/6), whilst the Hatfield accident reduced demand by about 5%, albeit over a shorter period (2000/1 to 2006/7). A welfare analysis suggests that although consumers seem to have gained as a result of privatisation, for most years this has been offset by increases in costs. An exception is provided by the two years immediately before the Hatfield accident. Overall the loss in welfare since the reforms were introduced far exceeds the net receipts from the sale of rail businesses. Thus although the reforms have had advantages in terms of lower fares and better service levels than otherwise would have been the case, this appears to have been offset by increased infrastructure and train operations costs. The source of these high costs remains an area of speculation but appear to be related to aspects of both market and regulatory failure. © 2012 Elsevier Ltd. All rights reserved. 1. Introduction Transport, as in many other sectors, exhibits a relative paucity of policy evaluation and where such evaluation does occur it tends to focus on short run effects. Rail privatisation in Great Britain is no exception. There was a slew of studies of the early effects (e.g. Harris & Godward, 1997; Pollitt & Smith, 2002; White, 1998) but there have been no studies in recent years. There are good reasons for this — 'evaluation research is tortured by time constraints' (Pawson, 2002). The effects of a policy change are distorted by exogenous variables such as changes in population and income and are overtaken by other policy initiatives. Undeterred, this paper draws on the recent work of Robins (2012) and attempts to evaluate the long terms impacts of the privatisation of passenger rail services in Great Britain. This paper consists of the following * Corresponding author. Tel.: +44 0 23 8059 4660. E-mail address: jpreston@soton.ac.uk (J. Preston). 0739-8859/$ - see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.Org/10.1016/j.retrec.2012.05.019 sections. In the next section, a brief history of rail privatisation is outlined. A key issue is the date at which the policy intervention is deemed to have started. In section 3, trends in key variables are outlined, with a particular focus on passenger kilometres, fares, train kilometres and costs. In section 4, a methodology is outlined for determining the counterfactual — what would have happened if privatisation had not occurred? This consists of the development of a simple econometric model of rail demand and extrapolative models of key variables such as fares, train kilometres and costs. In section 5, this methodology is applied to undertake a cost—benefit analysis of rail privatisation for the period 1995/6 to 2008/9. In section 6, the implications of these findings are discussed and, in section 7, some conclusions are drawn. 2. A brief history of rail privatisation There are a number of detailed histories of the events surrounding the privatisation of the railways in Britain (Freeman & Shaw, 2000; Gourvish, 2002; Harris and Godward, op cit.; Shaw, Please cite this article in press as: Preston, J., & Robins, D., Evaluating the long term impacts of transport policy: The case of passenger rail privatisation, Research in Transportation Economics (2012), http://dx.doi.Org/10.1016/j.retrec.2012.05.019 ARTICLE IN PRESS J. Preston, D. Robins / Research in Transportation Economics xxx (2012) 1—7 2000; Wolmar, 2001, 2005). Privatisation began in the 1980s with the disposal of ancillary businesses such as hotels (1981—1984), ferries (1984) and rolling stock manufacturing (1987-1989). Attention then switched to privatisation of the core business (Redwood, 1988), with the Centre for Policy Studies promoting a solution based on horizontal separation (Gritten, 1988) and the Adam Smith Institute favouring vertical separation (Irvine, 1987). The Government outlined its intentions with a White Paper in 1992 and the Railways Act in 1993. 'Privatisation' was a policy package, which had a number of inter-related elements. The first was that passenger operations were horizontally separated into 25 Train Operating Companies and a model of off-the-track competition was introduced to be administered by the Office of Passenger Rail Franchising (OPRAF). These services were franchised to the private sector between February 1996 and March 1997. The second was that the rail business was vertically segregated, with operations split from infrastructure. A new track authority (Railtrack) was established in April 1994 and privatised in May 1996 through a flotation on the stock market. Three rolling stock leasing companies (ROSCOs) were created and sold privately, whilst something over 60 other businesses were created and sold to the private sector, including infrastructure maintenance and renewal and rolling stock maintenance. The third was that open access competition would be permitted. However, for passenger services this on-the-track competition would be moderated so as to encourage competition for the franchises. One of the architects of the reform process believed that over time the majority of passenger services would be supplied by open access operations, with franchising providing a back stop for socially necessary services (Foster, 1994). Fourth, a new regulatory body, the Office of the Rail Regulator (ORR) was established to regulate Railtrack (and hence ensure fair track access charges) and to regulate competition between operators. Open access competition has remained heavily regulated, albeit with some loosening over time (see Preston, 2009). It should be borne in mind, when we refer to privatisation we refer to the broad policy package initiated by the 1993 Railways Act. Moreover, this policy package is not static. In particular, the election of a Labour administration in May 1997 resulted in a series of policy shifts (see Preston, 2008a). In opposition, Labour opposed privatisation and called for a 'publicly owned, publicly accountable' railway, although at least one observer believes that Labour was in fact complicit in the privatisation process (Engle, 2011). However, their 1997 manifesto stated 'Our task will be to improve the situation as we find it, not as we would wish it to be'. What this amounted to was the abolition of OPRAF and the creation of the Strategic Rail Authority (SRA), created by the 2000 Transport Act, but in shadow running mode before that (Gourvish, 2008), and the launch of the Ten Year Plan for Transport in July 2000, with ambitious plans for increasing rail investment. However, these plans were largely overtaken by the Hatfield accident of 17 October 2000, which revealed serious shortcomings in track maintenance procedures that resulted in a series of speed restrictions across the network and hastily arranged remedial engineering works. As detailed by Preston (2002), this led to a dramatic deterioration of Railtrack's finances, so that it was placed into receivership in October 2001 and replaced by Network Rail, a company limited by guarantee, in October 2002. The Hatfield accidents also exposed weaknesses in the franchised TOCs, with around a half failing, and having over time to be re-let (Preston, 2008b). As a result of these events, a White Paper (The Future of Rail) was published in 2004 and a Railways Act passed in 2005. This resulted in the abolition of SRA, with rail operations being brought under the direct control of the Department for Transport (DfT). At ****** Long Distance Operators London and SE Operators 'Regional Operators ™ All Operators Fig. 1. Passenger kilometres by sector. the same time some regulatory functions of the ORR1 with respect to Network Rail's performance and capacity transferred to the DfT, although ORR retained responsibility for determining track access and charges and became responsible for safety regulation, a hot potato that had already been passed from Railtrack to the Health and Safety Executive. These arrangements were further refined by yet another White Paper in 2007, Delivering a Sustainable Railway, in which the DfT would draw up a High Level Output Specification (HLOS) for five year control periods (the current control period, CP4, runs from 2009 to 2014) and issue a Statement of Funds Available (SOFA). The ORR then determine how much of the HLOS Network Rail can deliver. This analysis suggests at least four periods to the 'privatisation' era covered by this paper. The first, 1992 to 1995, was the preparatory phase. The second from 1995 to 2000 was the initial privatisation phase, in which OPRAF and Railtrack are key players. The third, from 2000 to 2005, is the Hatfield phase in which OPRAF is replaced by SRA and Railtrack by Network Rail. The fourth phase is the post Hatfield phase, commencing in 2005, associated with greater control of the railways by DfT. A fifth phase, with an emphasis on budgetary control, may have been initiated by the Conservative Liberal Democratic Coalition Government that came to power in May 2010 and published a Rail Command Paper in March 2012 (DfT, 2012), but this is beyond the scope of this paper. 3. Key trends A large amount of useful data is published by ORR, principally in National Rail Trends and as a result some of the key trends in the privatised passenger rail market are relatively well known and have been reported at previous conferences (Nash & Smith, 2007; Preston, 1999; White, 2001). With respect to demand, measured in terms of passenger kilometres, we can see from Fig. 1 that there is a clear break in the long run series around 1995/6. Between 1979/ 80 and 1994/5 total demand was down slightly (by 7%) but between 1995/6 and 2008/9 there has been strong growth (of 69%). Between 1979/80 and 1994/5 the demand for long distance services was down around 20%, the demand of London and South East services had barely changed and the demand for Regional services had increased slightly (by 8%). Since 1995/6 all three sectors have seen strong growth, with London and the South East up 82%, Long Distance services up 62% and Regional services up 53%. An 1 The Office of the Rail Regulator became the Office of Rail Regulation in 2004, as a result of the 2003 Railways and Transport Safety Act. Please cite this article in press as: Preston, J., & Robins, D., Evaluating the long term impacts of transport policy: The case of passenger rail privatisation, Research in Transportation Economics (2012), http://dx.doi.Org/10.1016/j.retrec.2012.05.019 ARTICLE IN PRESS ]. Preston, D. Robins / Research in Transportation Economics xxx (2012) 1—7 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 lijiv/'---^- ' \ / * - * f S B Í ......Long Distance Operators RevPKm Regional Operators RevPKm •London and SE Operators RevPKm Fig. 2. Real revenue per passenger km (£ per 10 km, 2008 prices). important question, and one we will examine in the rest of the paper, is the extent to which this demand growth can be attributed to privatisation. Fig. 2 shows that prior to 1995/6, there had been a trend for real mean fares (measured by revenue per passenger km) to increase, particularly for London and the South East. However, privatisation was also accompanied by fares regulation. The price of season tickets and saver returns on long distance services (or the equivalent day return ticket) were protected. Initially, fare increases of baskets of fares were limited to RPI (1996-1998), then by RPI - 1% (1999—2003). Following the financial crisis instigated by Hatfield this was increased to RPI + 1% (2004-2011) and there was a proposal for this to be increased to RPI + 3% from 2012 (Steer Davies Gleave, 2011), although this has since been dropped. It should be noted that the revenue measure used here is really one of yield. Improved yield management techniques may be responsible for the rise in long distance fares, particularly since Hatfield. Overall, between 1979/80 and 1994/5 real fares increased by 13%. Between 1995/6 and 2008/9, they increased by only around 4%. Fig. 3 shows that there has been long term growth in Regional and London and South East services (measured in terms of train km) but there was a slight decline in Long Distance services up to 1995/6 but there has been an increase since. Overall between 1979/ 80 and 1994/5 there was an 8% increase in train km, but since 1995/ 6 there has been a 34% increase. It should be noted that data on seat km are not available and train km is only a proxy measure of increased capacity. Fig. 4 (based on Smith, 2006) shows that in terms of total costs per train kilometre, there was a decrease between 1979/80 to 1995/ 6 in real terms of almost 20% but since the Hatfield accident there was a large increase of almost 60% up to 2003/4 but there have been 20D. J? > J" J" J* J3 <*""' <& <& (£' # ef>V <&' <£' $~ <&" ****** Long distance operators *» « Regional operators ■London and SE operators (inc peak) .n rc T Oi C" O' O'i y. 'Ji 'J' 'J'i 'J' Ci 'J. 'J' 0"i 'J: C-'. 'Ji Bi O O O O U% rjT. fTi ~ —i i-r, Ě í t í í g ^—Costs TKm/f Fig. 4. Total costs per train km (£, 2008 prices). Source: Smith (2006). some decreases since. This cost explosion has been accompanied by an increase in Government support. This fluctuated around the £2 billion mark (in 2008 prices), although there is a downward spike around 1995/6 as a result of the receipts from privatisation sales (Fig. 5). Following the Hatfield accident there was a rapid increase to £7 billion in 2006/7, although there has been some fall back since. 4. A methodology for determining the counterfactual Building on the work of others (e.g. Wardman, 2006), this methodology consists of a demand forecasting model to determine the extent to which changes can be associated with the privatisation policy package and to assess changes in consumer surplus and total revenue. Extrapolative moving average models are used to determine the counterfactual trends in fares, train kilometres, operating costs and capital costs. The demand forecasting model was based on a simple negative exponential or semi-log time series formulation as follows: Ln PKMt = a + j8RPKMt + yTKMt + <5GDPt + 0PRTV + ^HAT + pSTRIKE (1) where PKMt = Passenger Kilometres in year t, RPKMt = Real Revenue per Passenger Kilometre in year t, TKMt = Train Kilometres in year t, GDPt = Real Gross Domestic Product in year t, PRIV = Privatisation Dummy Variable (1992/3 to 2005/6), HAT = Hatfield Dummy Variable (2000/1 to 2006/7) and STRIKE = Strikes Dummy Variable (1982/3 and 1991/2). The estimated coefficients of equation (1), using data from 1979/80 to 2008/9, and some diagnostic statistics are given in Table 1. All parameters are statistically significant (at the 1% level), the model explains over 98% of variation in the data and autocorrelation does not appear to be a problem. Over 100 different model specifications were tested, in particular with respect to functional & & 4 .<& _<£■»' _<&' Jf Fig. 3. Train km (millions). -—~ Total government support excluding PTE grants Total government support including PTE grants Fig. 5. Total government support to the rail industry (£ million, 2008 prices). Please cite this article in press as: Preston, J., & Robins, D., Evaluating the long term impacts of transport policy: The case of passenger rail privatisation, Research in Transportation Economics (2012), http://dx.doi.Org/10.1016/j.retrec.2012.05.019 ARTICLE IN PRESS 4 J. Preston, D. Robins / Research n Table 1 Forecasting model parameters. Coefficient Value r-statistic )7 1 Transportation Economics xxx (2012) 1—7 Actual & Counterfactual TKm 17.106 -2.817 7.093 3.614 a 2.923 ß -5.690 y 0.0024 <5 3.68762E-I 8 -0.092 -8.575 p -0.051 -3.117 p -0.063 -3.283 Adjusted R2 0.983 Durbin-Watson 1.453 form (linear and log linear models were also tested), the specification of the independent variables (e.g. the reciprocal of TKM was tested), the use of lagged variables, alternative explanatory variables (e.g. car ownership, car journey times, petrol prices) and alternative specifications of the Dummy Variables. The model given in Table 1 was believed to be the most parsimonious. The dummy variables in the model indicate that privatisation suppressed demand between 1992/3 and 2005/6 by around 8.8% (1 - exp 6) whilst the Hatfield accident suppressed demand between 2000/1 and 2006/7 by a further 5.0% (1 - exp fi). The strikes in the years 1992/3 (ASLEF) and 1991/2 (Signalmen) were estimated to reduce demand by around 6.1% (1 - exp p). A feature of the negative exponential specification is that demand elasticities are directly proportional to the relevant policy variables. At the mean values in the data, the elasticity of demand with respect to RPKM was computed to be -0.62, with respect to TKM it was calculated to be 0.90 and with respect to GDP it was found to be 0.39. These values are broadly consistent with some other studies (e.g. Whelan, Harvey, & Cartmel, 2010), although less consistent with others (e.g. Wardman, 2006). In particular, our fare and income elasticities are low (in absolute terms) and we are not able to model the impacts of car competition, in part due to multicollinearity between TKM, GDP and car competition variables at our aggregate level of analysis. It is possible that our approach is overestimating the impact of TKM on the growth in passenger demand and underestimating the impact of external factors such as income and car competition. It is thus possible that our approach will overestimate the benefits of the privatisation package. The counterfactual estimates for fares, train kilometres and infrastructure and train operating costs are based on trend analysis of five year moving averages (after Burton, Carrol, & Wall, 2002) and are illustrated by Figs. 6—9. It can be seen that the counterfactual assumes strong growth in both fares and train km but with the fares growth being greater than the actuality but train km growth being e "u r --^ t— , 300 \ ~ ' m ' —- 250 200 J—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—,—1—,— c"' c" 1 ij"* er, 13"' cr> c"" er» c-^ cr» c-"> ~ri cti c, c"> o o o o o o o o o i-H tH tH tH tH tH tH tH tH i-H tH tH tH tH tH tH f^j M r*\l H"^l r*\l -CfflCm--Ac TKm Fig. 7. Actual and counterfactual train kilometres (millions). less. The counterfactual also posits declining infrastructure and train operating costs per train kilometre whereas the actuality has been strong growth. Infrastructure costs consist of renewals (which exhibited the strongest growth), maintenance and other operating costs whilst train operating costs consist of operating costs and rolling stock capital costs. These cost calculations are based on the work of Smith (1996/7 to 2006/7) (Smith, 2006; Smith & Wheat, 2007), with the series extended using company accounts and the TAS Rail Industry Monitor. An important issue is the period of time over which our counterfactual trends are estimated. We use a relatively long period (five year moving averages from 1982/3 to 1994/5) which picks up the impacts of the commercialisation of British Rail. If we used a short period (for example the five years from 1988/9 to 1992/3 used by Pollitt and Smith (2002)), it is likely that the welfare impacts of privatisation would be enhanced as there is some indication that the benefits of commercialisation were diminishing over time, although the impacts of the Organising for Quality (OfQ) reforms that were implemented in the early 1990s but were largely pre-empted by the privatisation process, are unknown (Gourvish, 2002, 382—3). One of the key indices is the change in passenger kilometres. The actual change between 1994/5 and 2008/9 was 77%. Equation (1) forecasted a slightly lower growth of 70%, whilst equation (1) run with our counterfactual estimates of fares and train km and with actual GDP forecasts growth of 22%. This suggests that around 69% of the demand growth since 1994/5 can be attributed to privatisation, in part due to fares that were lower and service levels that were higher than they otherwise would have been. The rest of the growth would have happened anyway, in part due to increasing incomes. 0.13 0.125 0.12 0.115 E 011 £ 0.105 ™ 0.1 0.095 0.09 0.085 0.0S Actual & Counterfactual RevPKm 00 CO 00 CD E3 55 c3 c8 T- T> Ti IJi 30 C C C. C C> OOOOODOQOO <-« N N (S (S N N (N (N -CfRevPKm — — Ac RevPKm Fig. 6. Actual and counterfactual revenue per passenger km (£, 2008 prices). Actual & Counterfactual Total Infrastructure Operating Costs per TKm T 1 [ 1 I- \ rf5 > J" J? > o.^ > # J* J" J> J" > > — ~ Total Operating Costs per TKm (£/TKm) -CF Total Operating Costs perTKm Fig. 8. Actual and counterfactual infrastructure costs per train km (£, 2008 prices). Please cite this article in press as: Preston, J., & Robins, D., Evaluating the long term impacts of transport policy: The case of passenger rail privatisation, Research in Transportation Economics (2012), http://dx.doi.Org/10.1016/j.retrec.2012.05.019 ARTICLE IN PRESS ]. Preston, D. Robins / Research in Transportation Economics xxx (2012) 1—7 Actual & Counterfactual Train Operating Costs per TKm 14 -i- 5 4 ^-1-1-1-]-1-1-1-1-1-1-1-1-1-1-1-1-1-T-1-T-1-1-T-1-1-1-1-1-1-1 ^