Passenger Value of Time, Benefit-Cost Analysis and Airport Capital Investment Decisions, Volume 2: Final Report (2015)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

2 DEFINING THE COMPONENTS OF VALUE OF TIME Analysis of the value of time can seem very complicated. People spend their time in many ways during travel—in a vehicle, walking, waiting, and so forth—and they do so under many different situations (for example, travel to work, travel to leisure or personal business activities, or travel in the course of work). A given time saving might occur at the beginning or the end of a trip, during a short or a long trip, or during an outbound, intermediate, or return leg of a trip. There are good reasons to think that all of these factors have strong implications for how that time saving is valued by the traveler. Furthermore, it is now widely recognized that uncertainty in travel time for a given trip is a very important factor to people, and that they place value on reducing that uncertainty, which can be quite variable and sometimes very high. Similarly, while uncertainty deals with unexpected changes in travel time (and hence arrival time), another important time component to consider is known differences between the preferred and actual (scheduled) timing of a trip, known as schedule delay. Fortunately, many of these complexities can be encompassed within a few sensible, theoretical precepts regarding how people use their time and what constraints they face. While it is true that working out the implications of such theory leads to complexity, this disadvantage is compensated for by the fact that the principles are common across many different types of travel. Studies of urban travel are the source of the greatest number of results and empirical measurements, and many results from the analyses of value of time of urban travel can be adapted to air travel. This section begins with a summary of the most prominent current theories explaining how value of time and value of reliability are determined, with attention to how they vary with specific factors. It is followed by a discussion of the attributes of air travel that are specifically relevant in light of these concepts, which are systematically applied by decomposing air trips into distinct components. The Research Team uses the theories on the values of time and reliability to describe how various factors affect travelers’ valuations of both the typical time required (value of time) and the uncertainty in that time (value of reliability) of these distinct components. This section concludes with an explanation of derived implications for strategies to empirically measure the values of time and reliability of these components. 2.1 Theory of Value of Time The most prominent approach to the value of time relies on a series of models (each successively more complete) by Becker (1965), Oort (1969), and DeSerpa (1971), with further extensions by Jara-Díaz (2000, 2003), De Borger and Van Dender (2003), and De 2 Page 14

Borger and Fosgerau (2008). Convenient unified expositions are contained in Jara-Díaz (2007, chapter 2) and Small and Verhoef (2007, section 2.6). Small and Verhoef also develop a mathematical theory of how travelers would value reliability of travel, the latter based especially on Noland and Small (1995) and Bates et al. (2001). The basic theory of valuation of travel time begins with a traveler who values purchased goods, time spent working, and time spent in several other specific activities, including leisure pursuits and various phases of travel. The traveler faces a budget constraint: goods must be purchased out of unearned plus earned income. Also, the traveler, or the household to which he or she belongs, faces an overall time constraint: total time available for work, travel, and other activities is limited. Finally, each travel activity requires a minimum amount of time determined by the nature of the transportation system. The objective is to determine how much the traveler would be willing to pay for a reduction in one of the required travel times which might occur, for example, due to a terminal upgrade at an airport. A key result is that the value of time for any travel activity consists of two components, each of which has an intuitive interpretation. The first component is the monetary reward for time spent working, typically assumed equal to the after-tax wage rate. This component is based on the traveler’s potential ability to use any travel time savings to work more. In practice, of course, the flexibility to change work hours is often limited; however, in the long run it is never completely absent because one can seek to change to a different job with more or fewer hours.4 (Another interpretation of this flexibility is that if the traveler is part of a family, changes in the traveler’s overall time budget may impact other household members, for example, through household chores, and cause them to enter or exit the workforce.) Furthermore, even in the short-term, when it might be impractical to change work hours, the traveler will probably have to trade-off household chores or leisure time against changes in travel time, and these have value to the worker, which are also closely tied to the wage rate. The second component is the monetary equivalent of the traveler’s dislike of time spent traveling relative to that spent working. That is, in the conceptual trade-off of spending less time traveling and more time working, the traveler will account for the possibility that he or she enjoys one more than the other. If work is more onerous than travel (i.e. requiring more mental or physical effort), then this component of the value of travel time will be negative—meaning it lowers the value of travel time. Empirical estimates suggest that this is typically the case for most urban travel, since the measured value of travel time is typically less than the after-tax wage rate. 4 When flexibility is limited, this component of the value of time is modified according to a term that is positive if the person is constrained to work more hours than he or she would prefer given the current wage rate and travel time constraints; and negative in the opposite case (Small and Verhoef 2007, Section 2.6.3). Page 15

There are several crucial implications of this theory: • The value of time can be expected, on average, to rise with the wage rates of the traveler or of other members of the traveler’s household. If, as is common, the available measure is household income rather than the wage rate, the value of time is likely to rise with income. However, there is no presumption that this relationship should be proportional. • The value of time in any particular travel activity depends on factors that determine how much that time is enjoyed. Travelers will value time savings more in onerous travel situations than in situations where they are relaxed and happy, because they want to reduce the amount of time they spend unpleasantly. This implies the following corollary: o The value of time is strongly influenced by environmental conditions and by stress-inducing aspects of travel. Examples of environmental conditions are noise, discomfort, and physical exertion. Examples of stress-inducing aspects are lack of safety (perceived or actual), worry about missing connections, unfamiliarity leading to fear of making wrong choices, and high levels of crowding, to name but a few. 2.2 Theory of Value of Reliability Reliability refers to how certain the traveler can be of the time required to complete a trip. Thus, reliability is inversely related to “travel time uncertainty,” and the two terms are often used synonymously when defining valuation. Just as value of time is defined as the amount a traveler would pay to reduce travel time, the value of reliability (or the value of travel time uncertainty) can be defined as the willingness to pay to reduce that uncertainty. Measuring Reliability Unlike time, whose units are explicit, here it is important to explicitly define uncertainty. A common definition used by researchers is “the standard deviation of the realized travel times, taken across measurements of otherwise identical trips over different days”.5 Another common set of definitions focuses on the likelihood of being unexpectedly late: for 5 The standard deviation of a set of numbers (such as travel times encountered over many different days for the same trip) is a measure of how dispersed they are. Specifically, if {Ti, i=1,…,n} denotes a set of N such numbers, its standard deviation σ is defined as the square root of the average of the squared deviations from the mean T : ∑ = −= n i i TTn 1 2)(1σ . If one considers there to be an infinite number of possible realizations Ti of a random process (i.e.,. if Ti is generated by a probability distribution), the same formula applies but with each term multiplied the probability of realizing that value of Ti, and with n tending to infinity (provided this infinite summation exists for the particular probability distribution function). Page 16

example, the frequency of being 30 minutes late, or the difference between the median and the 80th percentile of the distribution of travel times across days.6 Some studies present the percentage of trips falling into different arrival bands (e.g., 20 percent of trips are up to 10 minutes early, 50 percent of trips are on-time, and 30 percent of trips are up to 20 minutes late)7, while others present a probability of delay along with the mean delay across delayed trips8. Other studies present a mean expected travel time alongside a positive or negative incremental value indicating a shift in travel time that travelers can expect to experience across a set of journeys9. Finally, in the most complex presentations, travelers are shown a number of possible outcomes (typically ten) in terms of actual travel times for a given journey.10 Reliability Based on the Value of Time One way to understand why people value reliability is based on the theory of value of time just described. If someone is nervous about missing a connection or an important meeting, that time will be more onerous and so the person will pay to reduce it. But there is a more precise way to address reliability, which seems to do quite well in explaining responses to survey questions about more and less reliable options. This is based on scheduling preferences, in particular on the inability of travelers to achieve certain desired schedules in the face of travel time uncertainty (Noland et al., 1998; Bates et al., 2001). The theory of value of reliability is best developed for the situation where there is a particular desired arrival time at the destination (for example, the time of a meeting, a doctor’s appointment, or the start of official work hours). Travelers are assumed to suffer a small penalty for each minute they arrive early (because they would have preferred spending that time at home or in some leisure activity), and a larger penalty for each minute they arrive late. They might also suffer a discrete penalty for being late at all.11 These penalties are known as “schedule delay costs”, although they would better be termed “schedule mismatch costs” because they involve the disadvantages of arriving early as well as late. 6 If {Ti, i=1,…,n} denotes a set of possible travel times, the median T m is that value for which half the values Ti are above and half below Tm; the 80th percentile T80 is that value for which 20 percent of the values Ti are above and 80 percent are below T80. 7See Hensher & Li (2012) for an example of this approach. 8 See Hess et al. (2012). 9 See Layton & Hensher (2010) for an example of this approach. 10 See Bates et al. (2001). 11 The parallels to air travel here are obvious, though not perfect. Many air trips are for the purpose of meeting someone, and some will involve catching a train or bus that departs at a specific time. Page 17

When travel time is uncertain, the traveler cannot precisely choose any arrival time, but instead must choose a travel strategy based on the distribution of uncertain travel times that are faced.12 For example, a traveler with a strong penalty for arriving late at the destination will often plan on a “buffer” by departing earlier than would otherwise be necessary. The result is that most often the traveler suffers some (relatively small) penalty for arriving early, and occasionally suffers the larger penalty for arriving late (unless the buffer is chosen to be extremely large). The trade-off then becomes how large a buffer to choose. This kind of theory leads to a mathematical characterization of the traveler’s decision. For example, under certain assumptions, the traveler chooses a buffer that makes the chance of being late equal to β/(β+γ), where β and γ are the per-minute penalties for being early and late, respectively (Small and Verhoef 2007, p. 51). This result says that a more generous buffer will be chosen (one with a lower chance of being late) when the penalty for arriving late is large relative to that for arriving early. Other more complicated results can be derived from other assumptions. In order to go further and predict the value to the traveler of reducing travel time uncertainty, it is necessary to know the distribution of uncertain travel times—that is, the range the travel times take as they vary from day to day, and the frequency with which they take various values within that range. One approach is to assume a general class of such functions, for example the exponential distribution. The problem of measuring the value of reliability as the change in expected scheduling costs has been solved analytically for several such mathematical distributions: results are in Noland and Small (1995) and Koster, Verhoef, and Kroes (2009). Values of Reliability Drawn From Observed Distributions Another possible approach is to collect enough data so that the distribution itself can be measured empirically using non-parametric techniques (i.e., techniques that do not assume any particular functional form). The expected scheduling cost can then be found analytically from properties of the estimated distribution, under certain mild assumptions.13 This approach is in its infancy, with the Research Team aware of only two such empirical measurements: Fosgerau and Fukuda (2010) and Koster, Droes, and Verhoef (2011). Both of these empirical applications are for urban road travel, but the second one concerns the 12 Even if travel time were certain, the traveler might not choose to arrive at the most preferred arrival time due to travel congestion or a mismatch with the schedules of transportation services. Indeed, this type of scheduling theory was first used to examine how travelers adjust their schedules in response to sharp rises and falls in highway congestion over the course of an urban rush hour (Small 1982). 13 These assumptions basically guarantee that the general shape of the distribution will not vary over different times of day, although its mean and inter-quartile range (or other measure of dispersion) can vary. See Fosgerau and Fukuda (2011). Page 18

scheduling costs of airport access via road travel. Meanwhile, an approximation to this approach would be to assume a rather general parametric distribution, such as a gamma distribution, whose solution has been found analytically by Koster, Verhoef, and Kroes (2009). Regardless of the particular distribution function assumed, reliability can be summarized using just one measure of dispersion if the shape of the distribution does not change as that measure of dispersion changes (Fosgerau and Karlström 2010). Unfortunately, it is not usually known whether the shape of the distribution will remain unchanged in the face of some investment that improves reliability, so this theoretical finding is of limited use in practice. Instead, it is more appropriate to choose a measure of dispersion that focuses on the most costly part of the travel-time distribution. One recommendation for such a measure, made by Small, Winston, and Yan (2005) and Brownstone and Small (2005), is the difference between the 80th percentile and the median, since this emphasizes those parts of the travel-time distribution most likely to result in costly missed connections or late arrivals. Such a measure can be calculated from an empirically estimated non-parametric probability distribution, as was done by Small, Winston, and Yan (2005). Other Factors Affecting Measurements of Reliability Other factors may affect the value of reliability, as well. Experiments in psychology have shown that people are relatively more comfortable in situations where they feel they have control. Thus, the value of reliability is likely to be smaller when the person undergoing a particular unanticipated delay can react to it constructively, even if only to learn information about it. For example, travel time uncertainty enroute to or from the airport in a private car may have a lower value than uncertainty about delays while on an aircraft awaiting a mechanic—especially, if the cause of the aircraft’s problem is unknown. Similarly, uncertainty in the time waiting for an aircraft to arrive at a gate from a previous flight may be valued more highly than uncertainty in the walking distance to a gate, due to the passenger’s ability to react to the latter by walking faster. Thus, theory tells us several things about how the value of reliability depends on travel and traveler characteristics: • The reliability of travel is best measured as a property indicating dispersion of the distribution of uncertain travel times the traveler may encounter on a given trip. This distribution represents the possibilities that might occur for the given trip; empirically it might be measured by historical data for travel time on different days for the same route, season, day of week, and time of day. Dispersion can be measured by a standard deviation, by the probability of exceeding some threshold, or by other properties of the distribution that reflect the relatively infrequent but disrupting occasions when travel is much longer than anticipated. Such measures of dispersion do not require assuming a particular form for the probability distribution, and, therefore, can make maximum use of the historical data without prior assumptions. Page 19

• The value of reliability is likely to depend strongly on scheduling flexibility. When people don’t care much about the time they depart or arrive, travel time uncertainty is not very important to them. Conversely, when arriving late at some event (or departing early from an event in order to allow a buffer time) is very costly, they will place a high value on being able to accurately predict their travel time. • The value of reliability is likely to depend on how much information and control the passenger has over the situation where uncertain travel times are faced. While, as discussed above, there is a growing body of evidence on the measures of travel time variability that is most conceptually accurate, those measures are not necessarily appropriate for presentation to respondents in a survey. For example, the concepts of a standard deviation or the difference between the 80th percentile and the median will not be understood by most respondents. Using them in a survey would, therefore, have detrimental impacts on response quality. Even relatively simple concepts, such as the extent of variability presented as a range around the expected value, have been observed to lead to problems in identifying a meaningful estimate of the valuation of variability (see Hess et al., 2007). Furthermore, even where complex presentations of variability may work for studies that focus mainly on such variability (e.g., as in Bates et al., 2001), such presentations may not work in surveys with a larger number of attributes or a different topical focus. This is because any attributes that are difficult to understand are likely to be ignored or misunderstood by some respondents in favor of other attributes. From this perspective, a more simple presentation of variability may be appropriate in the present context. For example, in the case of flight delays, this could use the measure calculated and widely reported by the U.S. Department of Transportation consisting of the percentage of flights that arrive at their destinations more than 15 minutes later than scheduled. Distinguishing Between Reliability and Specific Time Components Because travel time reliability is defined in terms of variations in the duration of travel time components, it is sometimes not obvious whether a particular kind of delay should be considered a separate time component or an aspect of reliability. In particular, how should the time spent experiencing a delay that may or may not cause a problem in achieving a planned objective at the destination be approached? As an example, consider three kinds of time spent in the aircraft. First, there is time spent in normal operations that occur in the absence of delays: boarding and alighting, taxiing to and from the runway, takeoff and landing, and time actually in the air. Second, some time spent on aircraft is caused by congestion or other scheduling delays, but it is covered by the buffer times that airlines add to published schedules and so are not commonly perceived as a reliability problem. Airline schedules build in extra time for boarding passengers with late connecting flights, waiting for other aircraft to move out of the way, queueing on the taxiway, and delays encountered in the air enroute to the destination. (Formerly, there was Page 20

often holding in the air for landing clearance, but this has become less common as air traffic control instead has attempted to shift those delays to the ground at the departure airport.) Third, there are delays beyond those built into schedules, caused by such factors as unusual runway congestion, bad weather, faulty aircraft, and late crew arrival. There is value in reducing all three kinds of in-aircraft time, but the Research Team finds it most useful to treat the first two specifically as travel time components and the third as an aspect of reliability. The first component (routine operations) might be affected by certain investments (for example, changes in air traffic control that permit shorter air distances). The second (anticipated delays) is the result of expectations by airlines and passengers, which are affected by airport congestion but also the airlines’ choices when trading-off extra built-in scheduled time (a negative for marketing) against better on-time performance (a positive for marketing).14 The third (unanticipated delays) can be viewed as a particular realization of the uncertain travel times whose distribution forms the basis for the value of reliability, as just discussed. Thus, the first two kinds of in-aircraft time are treated as time components with value to be measured empirically, and the distribution of the third kind as a measure of unreliability, whose value is also to be measured empirically.15 Even this does not fully account for the variations in values of time that travelers may display. It is plausible that travelers place a higher value on saving those parts of in-aircraft time when they are not free to use electronics or move about the cabin. Some of this higher-valued time occurs during delays, but some occurs during routine operations, such as takeoff and landing. Ideally, one would like to measure their values separately, but doing so in a survey would be costly because one cannot ask too many stated preference questions or ask survey respondents to account for too many characteristics in such questions without creating respondent fatigue. 14 Such supply-side responses to airport reliability and delays explicitly (e.g.., how much to buffer schedules to improve on-time performance, and how much connection time to allow in itineraries) are very important to airlines as they have implications for both their passenger market shares and for operating costs (e.g., high costs for missing scheduled connections vs. additional costs of increasing scheduled times). 15 There is a further problem that the impacts of these different kinds of delays might not be additive. For example, a traveler who is especially sensitive to the third kind of delay (unanticipated delay) due to a tight scheduling constraint might find the other kinds of delay more onerous due to worry about meeting the schedule. The Research Team is not aware of any practical method to handle such spillovers from one type of travel time to another, and so the usual practice is adopted of assuming the spillovers are additive. The practice of distinguishing several types of delays makes this assumption less likely to pose a problem than in the case where they are aggregated into a single measure, because at least some parts of delay are able to be valued differently from others. Page 21

Travel Time Reliability in Air Travel In considering how to apply the lessons of past research on travel time reliability discussed in the previous section, it is worth discussing in more detail how travel time uncertainty affects air travel in particular. The airline traveler can suffer two different consequences of unexpected delays. One is the possibility of missing one’s initial flight due to delays in ground access or in the time it takes to check-in for a flight and clear security. The other is the possibility of missing a connecting flight and/or some important event or connection at the final destination (for example, a meeting or the last train of the day home), as a result of a delay in the arrival of the flight on the previous leg of the trip. These two consequences differ in terms of the costs suffered, but also in terms of precautionary travel decisions made in response to the uncertainty. Both of these differences affect the value of reliability. In the case of delays that might cause a traveler to miss an initial flight segment, the traveler can make allowances by leaving his or her ground origin earlier than otherwise would be necessary (that is, by allowing a buffer) or by taking a more expensive but more reliable travel mode (e.g., a taxi instead of public transit). This typically involves trading-off some additional inconvenience or expense against the perceived cost if the flight is missed. In the second case, that of missing a connecting flight or a planned event at the destination, the traveler can book an itinerary with more connection time or one that arrives at the destination earlier, or choose an itinerary with better on-time performance. Furthermore, the cost of missing a connection, as well as the ability to avoid it by choosing a less tightly scheduled itinerary, depends on the frequency of connecting flights, their load factors, and the policies of the airline (and the specific fare class) regarding shifting a traveler who has missed a connection to another carrier where that would reduce the delay involved. The load factor on potential alternate flights affects travelers with a connecting itinerary in two ways. First, it may affect the fare that is available if a traveler would like to book an alternative itinerary that allows more time for the connection. Second, if a traveler misses a connection, the ability to take a later flight may be constrained if those flights are full (although airlines may be able to make a seat available on a full flight by offering incentives to other passengers on that flight to volunteer to take a later flight). Unfortunately, public data on aircraft passenger load factors by flight segment do not allow any meaningful analysis of either of these effects. Finally, further delays can be caused on the egress journey, where the traveler has the option of using a more reliable egress mode from the destination airport. In addition, the perceived costs of unexpected delays may differ for both objective and subjective reasons. Objectively, for a business traveler attending a meeting or a leisure traveler attending an event at a specific time, such as a wedding, the entire trip might be wasted if delay at the destination is too much. Subjectively, travelers may be more anxious when they cannot estimate the extent or the consequences of a delay. For example, a traveler encountering delays on the ground access trip may have no way to estimate its extent (although this is rapidly changing due to new technologies); whereas the length of Page 22

delay can be somewhat estimated while waiting in a check-in or security screening line, reducing anxiety while allowing the traveler to more accurately judge the danger of missing a flight—factoring in the possibility that staff will invite people with close departures to move ahead in the queue. A number of other strategies are also relevant, although difficult to analyze rigorously. Experienced travelers may make their own judgments as to likely check-in and security delays, while other travelers (or those returning home from an unfamiliar airport) may instead simply follow the standard advice offered by airlines about airport arrival time. For an early morning departure, uncertainty in ground access time (and the cost of allowing a large buffer) can be reduced by staying at a hotel near the airport the night before. (Some airport hotels encourage this and reduce its cost by offering free parking for the entire trip duration for those who stay at the hotel for one night.) Trips arriving at the end of the day encounter a different set of issues. While travelers are less likely to have an event scheduled at a specific time, a late arrival at the destination may take time out from leisure, other planned activities, or sleep. If the destination is away from home, it could mean having difficulty getting a meal or having to pay for a taxi to reach their final destination if less expensive transportation options have stopped running. Travelers missing a connecting flight may have to stay overnight in a hotel at their connecting point, which could be extremely disruptive to their plans for the following day. In all cases, late arrival may require more expensive or inconvenient ground transportation, and planning for this contingency may affect the chosen egress mode at the destination airport—for example, booking a car rental ahead of time or, if it is the home airport, leaving a car parked at the airport instead of planning to use public transit or to be picked up by a family member. Finally, one empirical result from a study of reliability in urban travel by Noland et al. (1998) found that no discernible additional costs could be attributed to planning efforts (such as additional mental effort required to organize one’s schedule in light of travel time uncertainty) once scheduling costs were accounted for in the manner described in the previous section. However, based on the discussion above, one can postulate that, for some types of air travel, the additional effort required to account for contingencies of uncertain travel times might be substantial. An example would be trips requiring multi- modal connections on an unfamiliar local transit system with infrequent service. Thus two additional conclusions are reached: • Once scheduling costs are accounted for, other planning costs have not been verified to be important in urban travel as a distinct source of value of reliability. But they might be important for air travel for complex trips with tight schedules. • The detailed trade-offs involved in the choice of flights deserve further empirical research in order to fully understand how air travelers value reliability. One important research need is to learn how travelers obtain and use information about the variability in travel times and flight schedule reliability. For example, travelers Page 23

can now obtain airline on-time statistics fairly easily, but how many travelers actually use this information and how do they extrapolate estimates of the schedule delay costs they may incur from the data? 2.3 Valuation of Individual Components of Time in Air Travel A good way to advance understanding of the value of time and value of reliability in air travel is to note that such travel involves many distinct travel segments, each involving time that might be made shorter or longer by actions taken by air carriers or terminal operators. Given the theory developed above, some likely differences across these components in the values of time and reliability can be identified and—equally important—in the factors that affect those values. This initially led to the Research Team to divide the time spent in making an air trip into groups of components, which are listed below. The boundaries between the trip components were later modified as the stated preference survey for this project was finalized: 1. Ground-side access time 2. Time spent in flight check-in, security, and walking to gate 3. Time spent in gate area before boarding 4. In-aircraft time 5. Transfer connection time 6. Baggage pickup and terminal egress time 7. Ground-side egress time, including deplaning and walking to baggage claim or terminal exit (but not picking up baggage) Component 1: Ground-side access time This part of the trip is similar to an urban trip: indeed in many cases it is an urban trip. Thus, the extensive theoretical and empirical literature on the values of urban travel time and its reliability is relevant. This is especially true for the literature on journey to work and other work trips because they, like travel to an airport, potentially involve a target (preferred) arrival time and substantial perceived costs of missing this target. One difference, however, is that the penalty for lateness is likely to be even higher for air travel (due to a missed a plane) than for journey to work or other work trips (being late for work, a meeting or appointment). Unreliability, as perceived by the traveler, may be especially high on a return air trip from an airport in an unfamiliar city. Lack of knowledge of the public transit system or of the typical patterns of highway congestion may add substantially to the perceived dispersion of Page 24

ground-side access time. Survey evidence is needed to measure this perceived dispersion, which may differ from objectively measured dispersion. Component 2: Time spent in flight check-in, security, and walking to gate This portion of the trip potentially involves high stress due to uncertainty and unfamiliarity with the physical layout and with required procedures. For example, travelers cannot know ahead of time the length of various queues for check-in (particularly if they need to bypass ticket kiosks and check bags or otherwise interact with an airline representative) or to pass security screening. Furthermore, there is only limited opportunity to use this time productively, and the environment is often noisy and crowded. Thus, this component of time is likely to have a high value (i.e. travelers would pay a lot to reduce it); and this value depends strongly on a traveler’s situation, experience with air travel, and reaction to stress. Furthermore, this part of the trip can contribute a lot to overall travel time uncertainty, making it a natural target for policies or investments designed to reduce uncertainty. Component 3: Gate area time Time spent at the gate, including patronizing nearby services, is probably the main use of any buffer time that was allowed and not lost through unexpected delays in the previous components of the trip. If this experience is relatively pleasant, it will reduce the overall cost of unreliability of the entire trip, by permitting a larger buffer. For example, people may reduce the chance of missing their flight by planning to arrive in time for a leisurely meal, then adjusting their plans to a more hurried or carry-on meal if they are delayed during ground-side access or flight check-in and security. Hence, investments in improving the environment for this part of the trip, or in making it more productive through mobile communications technology, may have an especially high payoff. However, it is worth remembering, from our theoretical review, that the valuation of any component of time depends not just on an absolute measure of how pleasant it is, but on how pleasant it is relative to whatever is given up when this component of time is increased. For buffer time at an airport, the relevant comparison is most likely with time spent at home (for outbound trips) or with time on the purposes of the trip (for return trips). The enjoyment of these alternative time uses and, therefore, the valuation of time at the gate, may be quite different for single-day trips than multi-day trips. Component 4: In-aircraft time Time spent in the aircraft will be valued based on factors similar to those for urban trips— including dependence on trip purpose and income. The value will be less when conditions are more pleasant; those conditions depend on load factors, seat spacing, temperature, noise level, quality of staff service, in-flight entertainment, and the like. Many of these factors will be functions of the specific airline and type of aircraft used for the flight. For purposes of how travelers perceive disutility , the most important parts of in-aircraft time are probably dwell time at the gate (after boarding) and delays on the taxiway system Page 25

before reaching the runway, which may be extra onerous due to the stress of not being sure if the plane will take off on-time. To the best of the Research Team’s knowledge, the issue of differences in the perceived value of time involved in delays while in the aircraft compared to those incurred waiting to board the aircraft or in-aircraft travel time generally has not been investigated. Component 5: Transfer connection time For connecting flights, time spent deplaning and walking to the departure gate (have similar characteristics to the same components at the departure airport (#2 above, in the sense that both activities require walking from place to place within the airport. The major difference being that component #5 only involves movement between gates and does not include processing as in checking bags or passing through security described in component #2. However, the penalty for lateness on the inbound flight may be even greater than in the case of a non-stop flight—missing a connection means not only delay at the connecting airport but also the potential of several hours additional delay or even an overnight stay in a location that was not planned as part of the intended itinerary. As was the case with a gate or taxiway delay prior to departure, any time spent waiting to deplane from the inbound aircraft has a very high value. There is also a similarity between this component and component #3, as arriving at a connecting gate early then would entail waiting time. Component #5 has been studied in detail for air travel (Theis et al., 2006). The theory in this work suggests that each traveler determines an optimal amount of connection time that considers the likelihood of missing a connection, the benefits of having adequate time to eat, use a phone, and deal with personal comfort—all being traded off against a general desire to reduce overall travel times. The empirical work finds that air passengers on average perceive additional benefits from up to 15 minutes above the minimum connection times allowed by airlines, but that these benefits gradually turn to disbenefits as connection times are lengthened beyond that amount. It is reasonable to expect that this additional time will vary as a function of overall flight time, with passengers on long-haul flights potentially having a greater desire for slightly longer connection times. As described by Theis et al., connection times are primarily an issue for airline operations planning and less so for airport design. Component 6: Baggage pickup and terminal egress time This component is similar to flight check-in—it involves waiting for an uncertain amount of time in an environment in which there are limited opportunities for other activities (except perhaps making phone calls (parts of #2). The importance of reliability is probably diminished, compared to making the original flight, but may not be entirely absent due to the desired schedule at the destination, particularly making connections to scheduled ground transportation. The commonly experienced lack of information on expected delivery times for baggage may increase the valuation for this component. Page 26

In the case of international trips, immigration and customs add another layer of unpleasantness and/or uncertainty, similar to check-in and security at the origin airport. Component 7: Ground-side egress time, including deplaning and walking to baggage claim or terminal exit (but not picking up baggage) The time spent getting from the destination airport to the actual trip destination has characteristics similar to ground access time at the origin (#1). The importance of reliability is likely to be smaller than at the origin because most of the trip has now been completed. But the actual extent of unreliability may be greater if the non-home destination city is unfamiliar: for example, the traveler may be unsure of ground transportation options, and may worry about getting lost if driving a rental car or the reliability of local transit. Each of the foregoing travel time components may be further divided into sub-components that travelers may perceive as having different disutility. For example, waiting during the ground-side access trip may be perceived as having a greater disutility per minute (higher value of time saved) than in-vehicle time, while time spent walking to the gate within the terminal building may have a higher perceived disutility per minute than time spent riding a people mover or moving walkway. However, quantifying differences in the perceived disutility for these sub-components presents empirical challenges, both because of the large number of potential sub-components and the relatively small time differences involved in most practical situations. The perceived value of travel time for each of the components and sub-components of an air trip are not of equal importance from the perspective of airport capital investment decisions, or may be relevant to some types of project but not others. Reduced air passenger in-aircraft delays forms one of the major benefits of airside capacity improvements. Similarly, a reduction in time spent moving within the airport terminal complex and, in particular, a reduction in the walking distance (and hence time spent walking), forms a significant part of the justification for construction of automated people- movers at airports. It follows from these two observations that care is required in assessing the value of different travel time components where there may be significant differences in the perceived disutility of different sub-components. Failure to properly account for these differences could lead to the use of inappropriate value of travel time savings in the context of a particular project. Adjustments for the Stated Preference Survey In the course of designing the stated preference survey, to be described below in Chapter 3, the aforementioned time components were modified to better distinguish between those that were likely to have similar perceived values of time savings and to maintain a manageable stated preference survey. The resulting components used in the stated preference experiments were as follows: Page 27

1) Airport ground access time 2) Terminal access time from ground transportation 3) Check-in and security time 4) Time to reach the gate area from security 5) Time in the gate area until boarding the flight 6) Flight time, including making any flight connections Since the stated preference experiments only addressed the choice of flights and the trip to and through the departure airport, the time components used in the experiments did not include the time spent at the destination airport or the ground egress trip. 2.4 Productive Use of Time Spent in Travel In the previous section, it was noted that valuation of time may depend on how pleasantly the time is spent. Similarly, it can depend on how productive this time is for doing tasks such as making phone calls, reading, or working on a laptop. Productivity deserves special mention here because it particularly affects business travelers and so is likely to be highly valued and a prominent consideration by air travelers. Furthermore, it can be influenced by investments made by an airport authority or airline, for example, through the availability of power outlets or Wi-Fi service in a terminal or on an airplane. In addition, the productivity of waiting time can affect the value of reliability throughout the trip because waiting at the gate is a likely use of any extra buffer time the traveler decides to build into his or her itinerary. If time spent waiting at the airport is nearly as enjoyable as time at home or as productive as time at the office, then travelers may be inclined to include more buffer time in their plans and, thereby, reduce the cost of unreliability. However, there are some limitations to the possibilities of reducing values of time and reliability by productivity-enhancing technology. One is privacy. Travelers may avoid making phone calls on sensitive business matters where other people can overhear them, or using a laptop for company business in situations where people sitting next to them can see their computer screen. A second limitation is space and comfort. Many airline seats and even crowded departure lounges make it difficult to work on a laptop computer, while access to power in order to preserve battery charge can be a further limitation (witness travelers sitting on the floor to be within reach of a power outlet). A third consideration is the continuity of the activity in terms of location—productive work often involves some setup time which makes a block of uninterrupted time in one location more productive than several shorter blocks of time separated by movement from one place to another. Indeed, the fact that an air journey is inherently split into so many different components reduces the ability to use travel time productively and, thus, keeps values of time and reliability from being very low. This is in contrast with the conditions on high speed rail trips, where seating is generally more spacious, power outlets and Wi-Fi are commonly available, and Page 28

the trip involves longer blocks of uninterrupted time, a fact that Eurostar actively uses in its marketing on the London to Paris and Brussels routes. 2.5 Other Considerations Several additional factors are likely to affect values of time and reliability for air travel. The effect of income on value of time is likely to be strong for any of these components of air travel, in part because higher-income people often are concerned to use their time productively and, as a result, are more willing to pay for amenities, such as a quiet and uncrowded environment. But, as noted earlier, there is no reason to assume that value of time is strictly proportional to income or, for that matter, that the relationship with income is similar across different travel time components. Safety is an additional factor. Of course, safety is valued for its inherit importance, but it can also be related to the value of time. The level of perceived safety can affect the value of time spent in situations in which it produces feelings of anxiety or stress, but the extent of this is unknown. Further research could help clarify the precise role that perceived safety plays in air travel decisions. The fact that air passengers often travel in parties of more than one has implications for assessing the value of time. For example, do children have the same value of time as their parents? The presence of young children drastically changes the way the adults traveling with them use their time during trips, with implications for the perceived values of time and reliability for the entire party. Also, because the costs of some ground access and egress modes do not vary proportionally with the size of the travel party, observed differences in ground access and egress mode use by large versus small travel groups may reflect these differences in cost, and this, in turn, affects the way empirical studies can infer values of time from people’s choices. Although the primary focus of the current project is on the value of passenger travel time, benefit-cost analysis also has to consider the costs to airlines of flight delays. Apart from the direct costs of increased fuel use and additional crew time when aircraft are delayed, there are additional delays elsewhere in the network when one aircraft is delayed for whatever reason. To reduce some of these problems, airlines include some buffer time in their schedules, which, of course, is a cost to them and to travelers even if they succeed in eliminating unplanned delays. A particular difficulty for airline scheduling arises in cases where there is a large difference between the airport capacity in good weather and bad weather. Competitive considerations cause the airlines to schedule their flights to take full advantage of the good weather capacity, particularly when those conditions occur for a large part of the time. However, this can result in situations where bad weather causes a huge increase in delays to levels that greatly exceed the buffer allowances in the schedule. Rather than have large Page 29

delays ripple through the rest of their network, airlines often cancel flights in order to bring demand closer to the available capacity. In some cases, they are forced to cancel the flights anyway because the aircraft or crews scheduled to operate them are delayed somewhere else in the network. However, this imposes a different set of costs to recover from the disruption and to handle the passengers who would have flown on the cancelled flights. 2.6 Implications for Measurement Time components The discussion above suggests that some of the time components probably have similar characteristics with regard to value of time and reliability, and so can be grouped together to make analysis and discussion more tractable. A grouping into four broad components, while distinguishing three major types of trip purpose, can serve as a clear framework for measuring values of time and reliability. This leads to the matrix shown in Table 5. However, it should be noted that this breakdown does not cover all the major variables identified as being important, in particular trip length (especially single-day versus multi- day) and portion of the trip (outbound, intermediate, or return). Furthermore, certain types of situations may warrant singling out one particular component for a more intensive analysis. Table 5: Major combinations of time components for measuring values of time and reliability. Time component Component numbers Airport Access/Egress 1, 8 Security, check-in, & moving within airport 2, 5, 7 Waiting at gate 3, 6 In-aircraft time 4 Survey needs As noted, the values of time and reliability are likely to depend strongly on technology, whether for productivity or entertainment. Furthermore, the relevant technologies are changing rapidly, creating the danger that any measured values will be soon outdated. To cope with this problem, any stated preference (SP) survey of hypothetical choices in air travel situations would ideally specify the characteristics of the technology assumed to be available. By varying these characteristics across situations about which travelers are queried, one could, in principle, estimate models that enable measured values of time and reliability to be updated as new technologies become implemented. But, in practice, this could make the survey unduly complicated, and is perhaps less important than providing a more detailed categorization of travel time components. The SP survey performed in the Page 30

current research project handled changing technology by asking the respondents about the actual technologies they used or encountered. Surveys also need to pay careful attention to travelers’ scheduling desires, constraints, and general degree of flexibility, because these strongly affect the value of reliability. This was done by asking questions about these matters with respect to a recent trip and then asking for hypothetical choices for a similar trip. Alternatively, they can be specified as characteristics of some hypothetical trip, although this runs the danger of the respondent considering the situation to be impractical or incomprehensible. The better the survey explains the scheduling constraints that are assumed to apply, the more likely the responses will provide valid information about preferences under those constraints. This kind of strategy suggests designing surveys to describe scheduling constraints in a way that corresponds to what people actually worry about during their travel. For example, most travelers know what it means to miss a flight, to be late for a meeting, to feel rushed during a connection, to miss out on a planned meal, or to be forced onto an uncomfortable egress mode. To the extent the survey can describe conditions in these terms, respondents are more likely to understand the situation they are being asked to consider in their hypothetical choices. Finally, our review suggests keeping an open mind about how income, or any other measure of financial well-being, is considered. It is clear that both overall household income and the respondent’s own wage rate are likely to affect the values of time and reliability; thus, it is useful to ascertain both of these quantities in a survey, as well as the traveler’s position within the household. Then the challenge to the analyst is to specify models that use this information in a parsimonious yet informative way so that the variation of values of time and reliability with financial well-being can be accurately measured without overly restrictive preconceptions. 2.7 Recent Empirical Studies of Air Traveler Values of Time As described in Hess et al. (2007), there are significant challenges in using data that describe air travelers’ chosen itineraries alone to estimate values of time. In particular, it is difficult to infer what fares and other itineraries are available to travelers when they make their choices; thus, determining their time/cost trade-off is correspondingly difficult. Work based on such data is traditionally marred by problems with identifying significant and meaningful parameters values, notably for the cost coefficient (e.g., Pels et al., 2001, 2003; Hess & Polak, 2005, 2006a, 2006b). In addition, it is difficult to determine the effects of travel time components other than flight time without having information about the details of those components. As a result, more recent work uses data from air passenger surveys, which include both information about past trips made by the travelers and stated preference (SP) data describing how they might make those trips under changed circumstances. Page 31

Evidence in the literature suggests that data from SP surveys can be used to produce reliable measures of monetary valuations for different service attributes (Louviere, et. al., 2000). In particular, regular air travelers are accustomed to making complex choices, reducing any detrimental impacts of the hypothetical setting of SP surveys. Further steps can be taken to mitigate presentational effects, as described in Collins et al. (2011), who used an SP survey mimicking a typical online booking system. This study is also of interest given its in-depth study of the valuations of in-flight amenities, which show very high valuations on long-haul flights for better entertainment systems and greater seat pitch, relating to earlier points in the report about productive use of travel time. The studies conducted to date distinguish values for ground access/egress times, flight times, connection times and flight reliability, but not specifically for the times to check-in and clear security, nor for the initial wait time at the origin airport. Using data from a survey designed to collect information about actual trips made by air travelers combined with stated preference data, Adler et al. (2005) find that air travelers’ values of access/egress and flight time, as well as flight reliability, exhibit considerable heterogeneity across the population of travelers. They represented flight reliability by using the U.S. Department of Transportation’s on-time performance metric (percentage of flights arriving no more than 15 minutes late), which is convenient because it is readily available to air travelers for most flights at the time of booking. In the above cited Adler 2005 study, which accounts for random differences in preferences among travelers, the value of flight time is found to be approximately $70 per hour for business travelers and about half that for non- business travelers. The value of access/egress time is only about 10-15% less than flight time, for both business and non-business travelers. The value of reliability is found to be $38 per 10 percentage point change in on-time performance for business travelers, but only $7 per 10 percentage points for non-business travelers (with a high degree of heterogeneity for the latter). These results are broadly confirmed by Hess et al. (2007), with the additional finding that for holiday travelers, the sensitivity to on-time performance increases with flight distance. A later study by Warburg et al. (2006), using the same survey data, identifies the key traveler characteristics, both demographic and trip-related, that cause systematic differences in values of time across the population. In addition to trip purpose (business vs. non-business), these factors include income, travel party composition, fare reimbursement, and duration at the destination, among others. As noted earlier, Theis et al. (2006) use a survey instrument that collects and provides more detail about connection conditions to estimated values for connection times. This study finds that the estimated values vary depending on the duration of the connection, with very short times quite onerous to travelers and very long times about equally onerous as other travel time elements. Figure 3 illustrates these effects in terms of three components of “disutility”: the general preference for shorter overall travel times, balanced by the desire to make a connecting flight and to have time at the connection to deal with personal comfort and communication. This figure is intended only to show the general shape of the Page 32

effects. The work described in Theis et al. (2006) includes statistical estimation of the numerical values of these effects. Figure 3. Disutilities Associated with Connecting Time Source: Georg Theis, et al., Risk Averseness Regarding Short Connections in Airline Itinerary Choice, April 2006. Finally, work by Hess and Adler (2011) shows how values of time compare across four surveys that were conducted over a 5-year period. This work suggests that values of air travel time have changed over this period, likely as a result of such factors as changes in the composition of air travelers and changes in airport and in-flight amenities, some of which can be forecast and some of which may be difficult to anticipate. An implication of this finding is that values of time need to be periodically updated. Hypothesis: Three components of disutility associated with scheduled elapsed time in connecting itineraries  Value of time  Transfer success rate  Rush  Total Scheduled transfer time Scheduled transfer time Scheduled transfer time Scheduled transfer time DU DU DU DU MCT window of indifferenceDU – disutility MCT – minimum connecting time as published by airport Page 33