Proximity-based accessibility measures, which are also known as infrastructure-based or connectivity measures, define accessibility as the measure of physical proximity between locations. Indicators related to transport systems, such as travel distance, travel time, travel cost, travel speed, density of networks, congestion level, and location distributions, are often used (see, e.g. Geertman and Van Eck, 1995; Talen and Anselin, 1998; Wang et al., 2009). Some of these measures are only related to infrastructure supply, while others also include demand factors (e.g., travel time and chance of congestion depend on both supply and demand) (Van Wee et al., 2001). Proximity-based accessibility measures have advantages in operationalization and communication because it is relatively easy to collect necessary data
15 Different ways of terminology of these accessibility measures than the presented ones can be found
in the literature. Examples of these alternative terms will be introduced when these measures are introduced in the section 2.3.2.
16
Note that some papers (e.g. Geurs and van Wee, 2004) also categorize these hybrid models as utility-based models. Given their hybrid features, this overview does not categorize them as utility- based measures, nor as other types of measures.
for these measures and also relatively easy to explain these measures. However, they are often only suitable for appraisal of the direct effect of traffic and transport policy measures related to infrastructure development, traffic management, and pricing mechanisms, since these measures do not consider, for example, people’s preferences and behavior and factors related to activities/opportunities at locations. Take an example that when the accessibility to job opportunities for the citizens at a certain region needs to be evaluated, using proximity-based measures would be not suitable. The narrow scope of these measures also results in that they may be not sensitive enough for evaluation of spatial planning policies or ICT-related transport policies, which requires consideration of people’s preferences and behavior.
Space/time-based measures
Different from the accessibility measures defined from location perspective, space/time-based accessibility measures (also often named as people-based measures) analyze accessibility from the perspective of individuals or households, and describe to what extent people can reach locations or engage in activities (see, e.g. Dijst, 2004; Forer and Huisman, 2000; Kwan, 1998; Lenntorp, 1977; Miller, 1999; Neutens et al., 2012a). Space/time-based measures are developed based on the space-time prism (STP) theory of Hägerstrand (1970) – individual/household’s engagement of activities is subject to the spatial and temporal constraints caused by a set of chronologically ordered successive activities the time and space of which are difficult to alter. The space-time prism, which describes the potential reachable areas of activities at given pre-defined spatial and temporal constraints, is regarded as a measure for the accessibility for individual or household. Space/time-based accessibility measures, due to the recognition of space-time dimensions in individual/household’s activity- travel patterns, allow for more sensitive assessment of individual variations in accessibility. Compared to proximity-based accessibility measures, space/time-based measures prove to be effective in reflecting people’s activity-travel behavior into accessibility. One example of effects that can be captured via space/time-based measures but not via proximity-based measures is the constraints of people’s available time budget for conducting activities on people’s accessibility – a person may only have 30 minutes to do grocery shopping because of his or her work, and the effects of the available time to the person for shopping on his or her accessibility to the supermarkets in his or her region cannot be captured via proximity-based measures. However, the disadvantages of space/time-based measures lie in the difficulty to be operationalized and communicated. Using space/time-based measures often requires large amount of data at individual levels, which is difficult to collect, despite the development in GIS and spatial modeling (Kwan, 1998). It is therefore difficult to apply space/time-based measures to population groups or to a large geographical scale (Geurs and van Wee, 2004). In addition, due to their focus on short-term behavior, using space/time-based measures to evaluate transport and spatial planning policy measures is often restricted particularly when medium or long-term effects are concerned (Waddell, 2011).
Activity-based measures
Different from the space/time-based accessibility measures focusing on an individual level, some studies define accessibility as the reflection of the set of activities or opportunities available at particular locations at a more aggregate level. Such an activity-based definition is widely used for the appraisal of accessibility to different activities or opportunities such as jobs, retails, education, health services, and recreational facilities (see, e.g. Andersson and Klaesson, 2006; Kotavaara et al., 2011; Osland, 2010). A range of activity-based accessibility measures (sometimes also named as location-based measures) have been developed, for example contour accessibility measures (Breheny, 1978), potential accessibility measures
(Handy, 1994; Joseph and Bantock, 1982) and balancing-factors measures (Wilson, 1971). These measures are developed based on the so-called attraction theory that considers both the attraction of activities at locations (e.g., the number of the activities at locations) and the interactions between people and activities (e.g., the travel time needed for people to reach the activities) (Hansen, 1959; Weibull, 1976). Although the different measures developed based on attraction theory have their own advantages and disadvantages, in general, activity-based measures are considered both theoretically and practically superior to infrastructure-based measures in particular in terms of usability for evaluation of policy measures. Due to the consideration of attraction of activities at locations and the interaction between people and activities, activity-based measures are more policy-sensitive than infrastructure-based measures for evaluation of accessibility, for example, of the shops at a location. However, compared to space/time-based measures, activity-based measures do not consider spatial- temporal constraints people may face and hence may over-estimate the accessibility of the potential activities or opportunities reachable to people. These shortcomings limit the usability of activity-based measures into certain contexts, in particular, when individual characteristics or population variations are considered important. For example, activity-based measures cannot capture the effects of people’s time budget on their accessibility.
Utility-based measures17
Utility-based measures are founded on economic theory. Two streams of utility-based measures can be distinguished – generalized cost measures and random utility-based
measures (also named as LogSum-measures)18.
One stream of measures is so-called generalized cost measures, which are recently introduced as new accessibility measures for the Netherlands (Hoogendoorn-Lanser et al., 2012). Generalized cost measures estimate the generalized cost for traveler’s journey between locations. The generalized cost includes not only monetary costs such as the fares of public transport journeys or the fuel costs for car journeys, but also non-monetary costs, which are expressed in monetary terms, related to the aspects such as travel time, comfort, and quality during journeys (see, e.g. Groot et al., 2011; Hilbers et al., 2007). Generalized cost measures have advantages in operationalization and communicability. Data can be collected at reasonable costs. These measures can be expressed as comparable indices as well as absolute values, making it relatively easy to explain these measures to practitioners and policy makers. General cost measures are also considered with high policy relevance for appraisal of transport policies. For example, they are applicable to different transport modes, enabling comparison of accessibility between modes. These measures can provide insights into both the average accessibility of each traveler and the overall accessibility of the complete traffic flow (Hoogendoorn-Lanser et al., 2012). However, while it is relatively easy to convert some non-monetary factors such as value of time into monetary costs, it is a complex and not unanimous process to convert others (e.g., comfort, quality) into monetary costs
17 Note that, as introduced in section 2.3.1, there are composite accessibility measures that are
developed within utility-based paradigm but based on other types of accessibility measures.
18 Note that the perspective for utility-based measures is not consistent across studies in the literature.
For example, while some studies consider generalized cost measures also a type of utility-based measures (see, e.g. Hoogendoorn-Lanser et al., 2012), some studies use utility-based measures only to indicate random-utility measures (see, e.g. Crozet et al., 2012).
(Hoogendoorn-Lanser et al., 2012). Choosing appropriate converting coefficients for particular contexts is difficult. Using different converting coefficients in generalized cost measures to evaluate one same transport or spatial planning policy may result in different, or even contradictory, results. In addition, generalized cost measures cannot be used to effectively measure accessibility to activities/opportunities, as generalized cost measures largely focus on travels between locations only. This type of measures in turn seems not suitable for appraisal of spatial planning policy measures. Finally, it is also debatable whether generalized cost measures shall be considered as measures of accessibility or as measures of people’s valuation of accessibility.
Another stream of utility-based measures defines accessibility as the utility outcome of a set of activity-travel choices. These random utility-based measures are developed based on the utility theory from the perspective of user benefits within random utility paradigm (Ben- Akiva and Lerman, 1979, 1985). The accessibility within the random utility paradigm is formalized as the expected maximum utility that an individual can derive from a set of activity-travel alternatives (when analysts’ perspective is adopted)19 (see, e.g. Handy and Niemeier, 1997; Niemeier, 1997; Sweet, 1997). These random utility-based measures outperform the other accessibility measures in theoretical soundness and behavioral realism. These measures consider individual’s choice-making and therefore capture individual’s behavior in the measures. An additional advantage of these measures lies in their usability in economic evaluations. The random utility-based measures can be translated into LogSum- measures20 and be linked to micro-economic theory, allowing for calculations of consumer surplus for economic evaluations. Compared to other types of accessibility measures, these random utility-based measures are able to compute transport-user benefits of transport and land-use projects or policies. Random utility-based measures also show diminishing returns – the non-linear relationships between accessibility improvements and user-benefit changes. In addition, the random utility-based measures, which can be analyzed via discrete choice models, have advantages in operationalization. Because of the advantage in quantitatively modeling complex behavioral choices, discrete choice models have been used in modeling traveler behavior and decision-making to a dominant extent (McFadden, 2001). However, despite these advantages, random utility-based measures are relatively difficult to interpret and communicate (Geurs and van Wee, 2004). The random utility theory is not easy to understand and this results in difficulty to communicate these measures to, especially, practitioners and policy-makers who do not have knowledge of this theory. It is also not without question whether random utility-based measures can be essentially considered as measures of accessibility or better as measures of people’s valuation of accessibility (i.e., user benefits) (Geurs and Ritsema van Eck, 2001). In addition, while random utility-based measures are often used assuming decision-theory equivalent to decision-utility, there is a discrepancy between what these measures of accessibility aim to measure (experienced utility) and what they actually measure (decision-utility) (Chorus and de Jong, 2011).
19 Note that when an individual’s perspective is adopted (in this case, there is no unobserved utility
components by analysts), the accessibility should be defined as the maximum utility that an individual can derive from a set of activity-travel alternatives (see, Ben-Akiva and Lerman, 1985).
20
When the random utility component is i.i.d. Gumbel distributed, the accessibility can then be written
as the LogSum of the deterministic utilities of the alternatives in the choice set(see, Ben-Akiva and