Comparing short- and long-term values of travel time savings derived from a joint modelling framework

1 The value of travel time is an important element of cost-benefit analysis for appraisal of trans2 portation project, by encapsulating the willingness to pay of the population for improvements 3 in the transport system. Those values are typically obtained from mobility choice data, in the 4 form of revealed or stated preference surveys. Although short term decisions, such as route and 5 mode choice, are typically used for this purpose, a growing number of authors is arguing that 6 long term decisions might provide more meaningful values for the evaluation of transportation 7 projects. 8 This paper uses the German Value of Time Study, that contains both short and long term 9 choice experiments, to investigate the impact of different time horizons on the valuation of time. 10 In particular, the availability in the dataset of two different long term experiments (residential 11 and workplace choice) allow to evaluate not only the impact of the time horizon, but of the type 12 of long term decision. 13 Using a joint model including all relevant choice situations, this paper investigates the 14 difference in the valuation of time coming from different kind of choice experiments. 15 The results show that the chosen time horizon does have a significant effect on the valuation 16 of travel time and cost. Another finding is that the type of long term decision and the structure 17 of the choice experiment itself also influence the valuation. 18 Dubernet, I., Dubernet, T. and Axhausen, K.W. 2 INTRODUCTION 1 Microeconomic models of time allocation have been used to derive the valuations of technologi2 cally constrained time use since the work of Becker (1), Beesley (2) and DeSerpa (3). As a result 3 the value of time has been a subject of analysis for the past five decades. The current state of 4 practice draws largely upon past British, Dutch and Scandinavian studies (4, 5) which over time 5 moved from revealed preference (RP) data, where estimates are derived from the actual choices 6 made by travellers, to a growing reliance on personalized stated preference (SP) experiments, 7 where travellers are typically required to make choices between hypothetical situations. The 8 values of time are estimated using suitably formulated discrete choice models of travel behaviour, 9 especially of route and mode choices. 10 Most value of time studies consider short term decisions framing experiments around a 11 situation where respondents are presented with variations to travel time and cost of different 12 modes or routes. The questions arises if the focus on short term decisions is the most appropriate? 13 Can for example a commuter vary much of his daily commute in the short run or is it perhaps 14 more reasonable that changes in commutes occur because of longer term decisions that people 15 make such as where to work or where to live (6)? 16 Workplace and residential location influence many other behavioural choices of travellers 17 as they define the marginal cost of further travel and the distances involved. Therefore the 18 focus of several more recent empirical studies shifted to understand and explain everyday travel 19 behaviour as a routine activity changing due to key events such as residential relocation or 20 workplace decisions. An article by Müggenburg et al. (7) reviews the theoretical framework 21 and the most important studies investigating mobility behaviour in a long-term choice context. 22 Schirmer et al. (8) give a comprehensive overview of residential location choice literature and 23 show that travel time, commuting and employment changes are significant determinants of 24 choices. 25 Trading workplace or residential location, however, represents a long term choice; it is a 26 decision that is not made easily and cannot be changed quickly. In a recently published paper 27 by Beck et al. (9) the authors compare long and short travel time valuations. They make use of 28 Swedish stated preference data where the respondents first faced a set of choices where they had 29 to make cost and travel time trade-offs for their commute with public transport or car, before 30 facing an additional set of choices where they considered increases in travel time in return for a 31 higher salary (10). They found no differences in scale between the short-term and long-term 32 trade-off scenarios. However, they discovered a significant higher travel time valuation in the 33 long run. They conclude that the time horizon over which the choice experiment is being framed 34 results in significantly different values of time. 35 This paper examines value of time measures in this specific choice context with different time 36 horizons. The paper makes use of a combined revealed and stated stated preference experiment 37 conducted in Germany in 2012 (11). The respondents were presented a series of choice situations 38 including short term decisions such as route and mode choice, as well as long-term decisions 39 with residential and workplace location. In particular, the long term experiments asked the 40 respondent to make trade-offs between transport measures and a set of workplace or residence 41 attributes. The contribution of this work is to determine if the different time horizons lead 42 to significantly different time valuations but no scale differences as found in (9). In addition, 43 the data used in this paper contains two types of long term experiments, thus allowing to test 44 whether the framing of the question as a workplace or residential choice problem influences the 45 valuation of time. Finally, the high number of residence and workplace attributes in the German 46 data allows to investigate the relative importance of travel related variables compared to other 47 attributes of the locations, and how this influences the valuation of time. 48 Dubernet, I., Dubernet, T. and Axhausen, K.W. 3 The remainder of this paper is structured as follows: section 3 outlines the survey and is 1 followed by the description of the method used to estimate a joint shortand long-term model 2 (section 4); section 5 outlines the results of the modelling before presenting the final discussion 3 and outlook in section 6. 4 DATA DESCRIPTION 5 The data used for this analysis is taken from the German VOT Study. The German Federal 6 Ministry of Transport and Digital Infrastructure (BMVI) has recently published the 2030 7 Federal Transport Investment Plan (Bundesverkehrswegeplan, BVWP), its mediumto long8 term investment strategy for the country’s transport infrastructure serving longer distance travel 9 (12). As part of this, it updated the overall methodology of its central evaluation tool, cost-benefit 10 analysis (CBA). Within one project the German VOT Study values of travel time savings and 11 reliability for personal and business travel were estimated and recommended for the BMVI (11). 12 The design of the German VOT study builds on the experience of time valuation studies in 13 Switzerland. Swiss studies followed a variant path, when compared to international practice by 14 employing more complex SP experiments including multiple modes and multiple elements of 15 the generalized costs of travel in a series of overlapping choice contexts (13–16). 16 Two complementary samples one for business and one for non-business trips were 17 collected. This paper focuses on the non-business sample as only those respondents received 18 long-term choice experiments. A detailed description of the survey, collected data and response 19 behaviour can be found in (17). 20 After the pre-test in May 2012 the two-step survey was carried out in six subsequent waves 21 from July to October 2012. In the first step RP data on three trips undertaken by the respondents 22 were collected in a computer assisted telephone interview (CATI). The purposes of the RP trips 23 were pre-specified: commuting to work and the trips to most important shopping and leisure 24 (< 50 km) destination. Also information on the last long-distance trip over 50 km distance was 25 collected, and, if the latter was ground-based, data on the most recent air trip was also collected. 26 The gathered trip information was complemented with the usual socio-demographic information 27 and information about mobility tools. Out of the reported trips a reference trip was chosen 28 randomly. 29 The SP experiments were constructed around the reference trip. Information about the 30 non-chosen options were added. The non-chosen alternatives and their attributes were based on 31 information from a number of sources. Door-to-door car travel times were computed based on 32 the average travel times reported by Tom-Tom Stats and a NavTeq – network for Germany using 33 the MATSim software package (18). The average car travel costs were calculated based on the 34 2012 ADAC (General German Automobile Club) price-per-kilometer estimate for an average 35 sized car in each car segment (range from mini to caravan) (19). The travel times, headways, 36 transfers and prices on public transport including air travel were obtained from the relevant 37 websites with an internet bot. 38 The respondents received the SP experiments within a maximum of two weeks of having 39 participated in the CATI. The participants could choose to respond in a paper-and-pencil form 40 or with a web-based survey. Respondents received three different SP experiments either a 41 mode choice or route choice experiment, one reliability and one long-term experiment. In total 42 they were offered 24 single choice situations. Each type of SP experiment contained 8 choice 43 situations. 44 In the mode choice experiments the respondent had to choose between three modal alter45 natives. The modes offered depending on the reported reference mode were either walking, 46 Dubernet, I., Dubernet, T. and Axhausen, K.W. 4 cycling, car, local public transport (PT) and the various long distance public transport modes: 1 train, air and the newly deregulated coach o