Struggling with Time: Investigating Coupling Constraints

Abstract

The time–geographic concept of coupling constraints defining when and for how long persons have to be corporeally present at a given physical location can help transportation researchers to understand better how people combining employment and domestic responsibilities coordinate and negotiate everyday trips and activities. While the usefulness of the concept has long been recognized, operationalizing the ‘when’ dimension of coupling constraints in empirical research remains difficult. The paper reviews previous measurement approaches and introduces a complementary perspective that evolves around the time–space of arrival—a time span appropriate for arrival at a given destination—and that draws on insights from human geography, sociology, and psychology. A central element of the proposed perspective is the attention to different types of time. It is shown that conceiving of clock‐time as time per se may fail to account for the multitude of temporal factors that matter to parents’, and especially mothers’, coping with coupling constraints. Another important facet of the perspective concerns the complexity of boundaries between acceptable or appropriate and unacceptable or inappropriate arrival times. These and other characteristics of time–spaces of arrival are illustrated using in‐depth interviews and a stated adaptation exercise informed by cumulative prospect theory centring on employed parents’ trips to collect their children from childcare services and elementary schools in the Utrecht region of the Netherlands.

Introduction

Both popular and scientific writings associate contemporary anxieties about the increasing pace of life, ‘time squeeze’, and ‘time famine’ with the growing numbers of people who ‘juggle’ paid employment and domestic responsibilities within a 24‐hour time budget (Hochschild, 1997; Jarvis, 2005). Indeed, adults in dual‐earner family households with young children in particular have to ‘hurry’ to complete activities within limited time frames and often feel harassed by their everyday obligations. This time squeeze has direct ramifications for the transport sector. Not only can dual‐earner family households in particular benefit from fast and reliable transportation, but these households also have much to gain from living in mixed‐use, high‐density developments providing good access to relevant locations. Both would, at least in theory, enable them to connect home, workplaces, childcare and other important activity places seamlessly.

The time–geographical concept of the coupling constraint can help one to understand better how dual‐earner family households and others combine employment and domestic responsibilities. Coupling constraints define when and for how long persons have to be corporeally present at a given physical location (Hägerstrand, 1970) and thus structure their everyday activity/travel patterns in many ways (Cullen and Godson, 1975; Kwan, 2000). Though introduced a long time ago, the concept of coupling constraints has attracted increased attention in the geographical and transportation literature during the past decade or so. There is, for instance, a renewed interest in space–time accessibility modelling, which explicitly accounts for the impact of time constraints on the set of activity locations people can access (Kim and Kwan, 2003; Miller, 2005).

Simple as the concept might seem, the operationalization of coupling constraints and particularly their ‘when’ dimension in empirical research is a deceivingly difficult task. A number of approaches have been put forward since the early 1970s, but they suffer from various limitations despite the useful insights they have generated. The present author’s recent research has sought to complement existing approaches with a perspective that draws on notions from various disciplines and research fields, including human geography, social studies of time, feminist theory, science, technology and society (STS) studies, and psychology. This has resulted in alternative conceptualizations of temporality, or the nature of time, and the differences between acceptable and unacceptable arrival times at a specific destination.

The current paper has two objectives. It seeks to describe this framework for the ‘when’ dimension of coupling constraints. In addition, it provides some empirical support for specific aspects of the approach outlined by drawing on qualitative and quantitative studies of dual‐earner family households with young children in the Netherlands. These studies concentrated on the coupling constraints associated with collecting children from childcare or primary school on days that both parents engage in paid labour. This is not only because chauffeuring trips have been studied far less frequently than commutes, but also because such trips structure activity‐scheduling processes in important ways for many parents, in particular for mothers (Tivers, 1985; Hanson and Pratt, 1995; Kwan, 2000).

Review of Operationalizations of Coupling Constraints

Over the years various approaches have been proposed to measure the ‘when’ dimension of coupling constraints. The most straightforward and pragmatic approach goes back to the time–geography of Hägerstrand and colleagues at Lund university in the 1960s and 1970s (Hägerstrand, 1970; Lenntorp, 1976): activities are dichotomized as fixed or flexible and the ‘when’ dimension is assumed to materialize in the clock‐times at which fixed activities start. Specific activity types are usually considered fixed—in particular paid labour, activities at home and work related, education and chauffeuring activities (Jones et al., 1983; Kitamura, 1983; Schwanen and Dijst, 2003). Drawing on the work of Cullen and Godson (1975), Kwan (2000) dichotomized activities as fixed or flexible on the basis of respondents’ ratings of the extent to which activities could have been conducted at another time or another location and how easy it would be to change the time or location of the activity. Kwan has also used the starting times of fixed activities as manifestations of coupling constraints (e.g. Kim and Kwan, 2003). While useful, this approach makes strong assumptions about punctuality. As no account is taken of arrivals earlier or later than some preferred, intended, or prearranged clock‐time, it is implicitly assumed that persons arrive exactly on time.

Transport researchers have long since relaxed the punctuality assumption characterizing classic and many recent time–geographical studies. Almost 40 years ago it was recognized that travellers coping with uncertainty about the performance of transport systems incorporate a ‘slack time’ (Gaver, 1968) or ‘safety margin’ (Knight, 1974)—a period of reserve time immediately before the required presence at a given spatial location—in their activity schedules to avoid late arrivals at a given destination. Subsequent work has elaborated travellers’ decisions regarding safety margins. Small (1982) specified a utility maximization model of scheduling choices that considers the departure time for home‐to‐work trips as a trade‐off between travel time, the extent of early, and late arrivals (both in minutes), and a fixed lateness penalty. He found that the disutility of one extra minute of late arrival exceeds the disutility associated with one extra minute of early arrival and of travel time. Noland and Small (1995) extended this model to account for travel time variability. Their work was subsequently extended to account for the inability to plan one’s activities exactly resulting from travel time uncertainty (Noland et al., 1998), and for public transport trips (Bates et al., 2001). Bates (2007) recently expanded Small’s model to incorporate heterogeneity in attitudes towards risks more comprehensively. More extensive reviews of these and related studies are available in Noland and Polak (2002) and Senbil and Kitamura (2004). This line of enquiry has recently been integrated with Hägerstrand’s time–geography (e.g. Ashiru et al., 2004; Ettema and Timmermans, 2007).

Kitamura and colleagues have sought to make the ‘when’ dimension of coupling constraints measurable by attempting to determine the earliest possible departure time and latest possible arrival time at a given destination through stochastic frontier modelling (Kitamura et al., 2000). Stochastic frontiers are unobserved boundaries, which can be estimated using econometric techniques. In the current context this means that the unobservable clock‐time representing the ‘when’ dimension of a coupling constraint is estimated as a function of the actual arrival time, which is itself a function of variables measuring travellers’ socio‐demographic background, residential location, and commute length, and a non‐negative disturbance term. These authors believe that the estimated coupling constraints “represent the earliest possible departure time or latest possible arrival time as perceived by the individual” (Yamamoto et al., 2004, p. 239). However, one might object that information on individual perceptions is only inferred from revealed behaviour rather than incorporated in a more explicit form in the model.

The approaches discussed so far assume that coupling constraints are associated with a single clock‐time. This assumption can be criticized by referring to transport studies about coping with uncertainty about the performance of transport systems, in which it is assumed that travellers hold beliefs about ranges of acceptable arrival times at a destination or indifference bands (Hall, 1983; Mahmassani, 1990; Senbil and Kitamura, 2004). This approach is rooted in Simon’s (1955) notion of bounded rationality and might, therefore, be psychologically more realistic than studies concentrating on the identification of a single preferred, prearranged, or possible clock‐time. Nonetheless, it is (implicitly) assumed that these indifference bands are delimited by identifiable clock‐times that individuals should—in principle at least—be capable of articulating discursively. It is also not clear how boundaries on indifference bands have come into existence.

The approaches discussed so far have in common that they conceptualize time as a singular and linear dimension. Time is equated to clock‐time, which means that the relevance of types of time that are related but not identical to clock‐time may not be captured. Additionally, time and space are separated from one another, and the former favoured over the latter—a practice common in both philosophy and the social sciences (Latour, 1997; May and Thrift, 2001; Crang, 2005). As will be argued below, this conceptualization of time might limit one’s understanding of the ‘when’ dimension of coupling constraints.

Complementary Perspective

Following the indifference‐band approach described above, the author believes that the ‘when’ dimension of coupling constraints materializes in a time span appropriate for arrival at a given physical location, or a time–space of arrival (Schwanen, 2006). However, the perspective in the current paper differs from the indifference‐band approach in terms of the conceptualizations of time(space) and of the difference between what is acceptable and what not. These differences are the consequence of another, underlying difference. More emphasis is placed on the situated character of appropriate arrival times. With this the author means that the appropriateness of arrival times is not predetermined by cognitive intentions that are part of the inner world of, and belong exclusively to, human individuals, but is instead worked out in concrete situations or dynamic, ever‐changing constellations of interacting people, artefacts, and other material items. Thus, that appropriateness is relational, emerging out of interactions between persons and their socio‐physical contexts (Thrift, 1996; Bondi, 2005). Adopting this situational and relational perspective enables the moral and ideological dimensions of temporality to be emphasized, which have so far been underexposed in transport studies. As explained below, those dimensions refer to what is socially appropriate to do at specific times and to the disciplinary and power‐related aspects inherent to (clock)time, respectively.

The remainder of this section describes the conceptualizations of time(space) and the boundaries between acceptable and unacceptable arrival times central to this paper. These ideas have been developed in tandem with, and as part of, qualitative research about how parents in dual‐earner family households with young children juggle employment and care‐giving; where appropriate text fragments from that study will be used. Details of the qualitative research are available in Schwanen (2006, 2007). Suffice it to say that 40 employed mothers and fathers in the city of Utrecht and the surrounding suburbs administered a time–space diary for a single day on which they and their partners engaged in paid labour. On the subsequent day the author interviewed them about the diary and the juggling of employment and domestic responsibilities in general. Both fathers and mothers with different types of occupation, various levels of autonomy over their employment times, and residing in neighbourhoods that differ in terms of distance to the Utrecht central business district (CBD), local accessibility to facilities, and services and auto‐orientation participated; however, higher‐educated parents and those with tenured employment were over‐represented.

Time

As in other social sciences (Adam, 1990, 2004), temporality—the way(s) that time is—tends to be taken for granted in travel demand analysis and transportation research. Following the conventions of Western thought, time—and space—are typically seen as containers, as neutral and external backdrops to events and behaviours. What is more, time is usually assumed to be linear, singular, homogeneous, and quantifiable—a phenomenon that can be represented adequately and objectively through the familiar clock and calendar. Travel behaviour analysts often equate clock‐time to time per se, which, according to many social theorists and philosophers, means that they “confuse time and the measurement of time, which is a metrical reading on a straight line” (Serres and Latour, 1995, pp. 61–62).

Philosophers, sociologists, historians, and geographers have questioned these conventional notions of time and provided a wealth of more advanced conceptualizations. Reviewing these in detail is beyond the scope of the current paper (but see Adam, 1990, 2004; Urry, 2000; May and Thrift, 2001; Crang, 2005). Instead the present paper will draw on the literature about (space)time and temporality to explore and illustrate the following, interrelated points: that the dominance of clock‐time is historically contingent; that clock‐time is but one of the times that matter to people’s travel and activity choices; that only considering clock‐time might introduce gender bias into analysis results; and that time and space should be seen as linked rather than independent.

Clock‐time is sometimes thought to reflect the ‘real’ time of nature and the universe, which would make it more objective (and scientific) than lived or experienced time. It is, however, a classic example of a socio‐technological construction (Adam, 1990, 2004; Glennie and Thrift, 1996, 2005). Standard accounts of the history of clock‐time have shown how clock‐time was developed in the Christian monasteries in medieval Europe and slowly spread from there—at first to the business and administrative elites in the later Middle Ages and Renaissance, and with the urbanization and industrialization processes and associated growth in corporeal travel in the 18th and 19th centuries to the public at large (Landes, 1983; Adam, 2004; for a somewhat different historical account, see Glennie and Thrift, 2005). Clock‐time spread rapidly during the industrial era due to its machine‐like properties of standardization, homogeneity and context‐independence, which made it a perfect instrument for the temporal coordination of places. Studies of the history of clock‐time also emphasize its importance as a medium enabling the disciplining of and exercising power over fellow humans and have them act in other people’s interests (Foucault, 1977; Thrift, 1996; see also below). Thus, clock‐time is not the natural and neutral system of reference it is commonly taken to be; its dominance is the result of the particularity of the course of Euro‐American history.

The above leaves intact that clock‐time has a real and significant impact on people’s lived temporal experience. Nonetheless, clock‐time has not replaced or fully captured the times and rhythms of the cosmos, nature and our own bodies, which also bear directly on people’s everyday experience of time (Thrift, 1996; Urry, 2000; Adam, 2004). While often thought to be completely antagonistic to one another, the relations between clock time and the rhythms of the cosmos, nature, and the body are more complex than that: the latter are not entirely separate from clock‐time, but neither can they be reduced to the former—they are instead partially connected (Strathern, 1991; Urry, 2006). This implies that travel behaviour researchers should pay more attention to those kinds of time in studies of coping with coupling constraints and so avoid the fact that those times are expelled into invisibility (Law, 2004), even if they can be expressed only to a certain degree through clock‐time.

The importance of, and interconnections between, different times and rhythms was also evident in the narratives of the working parents interviewed about collecting the children from the nursery after paid work. One of them was Kate ¹ whose two children are looked after by a nursery close to home in a newly built neighbourhood on the outskirts of the city of Utrecht (Schwanen, 2006). Like many others in the Netherlands, this childcare centre has a fairly strict clock‐based time regime: it closes at 18.00 hours, but all children should have been collected by 17.45 hours. Because she has to commute home via a heavily congested road, Kate incorporates a safety margin of 25 minutes (exceeding her regular commute time of 20 minutes) into her schedule to avoid arrivals after 17.45 hours. However, even an arrival before this clock‐time can be very awkward (also Schwanen, 2006):

Somebody always has to be the last one to get picked up, sure I know that, but then if it’s yours who is then on the floor all by himself putting a puzzle together […], that is not nice. I prefer to pick him up at the same time along with lots of other parents and um, of course I know how those children react, um, I only have to hear just once that [youngest child] too had ever said, where is my mum now, while the other children were being picked up then, um, of course that cuts right to your soul, you might say, oh well. What also plays a role is that I also feel that they don’t have to be the last to be picked up.

Note that Kate’s coping with, and experience of, the coupling constraints associated with the nursery’s closing time is not at odds with Small’s (1982) model of scheduling choices discussed above. In this particular instance too, late arrival weighs (much) more heavily than arriving early, which is associated with little or no disutility at all (many interviewees reported to have a strong preference for arriving as early as possible when collecting their children from childcare). However, the example of Kate goes beyond utility‐based scheduling models, for it also indicates two more conceptual issues. ² First, it shows that time and space are difficult to separate from one another: one of the times they play a role in Kate’s and other interviewees’ coping with coupling constraints lies in the rhythms—dynamics in the presence/absence—of individuals and artefacts within the bounded space of the nursery building. As increasingly more of the children in Kate’s son’s class have been collected and the child is increasingly left behind, Kate gets a relational sense of arriving (too) late that is based in part on social comparison. This sense is caused by arrhythmia (Lefebvre, 2004): a discordance between Kate’s own rhythm and that of the people and material objects that constitute her child’s physical environment. In contrast, arriving around the same time as the other parents at the nursery constitutes a state of eurhythmia, or ‘metastable equilibrium’ (Lefebvre, 2004, p. 20). Interestingly, Kate and other interviewees were unable to articulate the event‐based (space)time associated with the dynamics of individuals and objects very well in terms of clock‐time; they could only give the range 17.30–17.45 hours as an approximation. This, then, suggests that this event‐based (space)time and clock‐time are only partially connected; one cannot be expressed fully through the other.

Second, it is no coincidence that a mother articulates the relevance of the event‐based (space)time embedded in the presence/absence of people and objects so clearly. Although not all mothers attached equal weight to this time, the participating fathers tended not to talk about this type of time at all or considered it of little relevance when specifically asked about it. The interviews suggested that fathers as a group tended to concentrate more exclusively on the clock‐time(s) set by the nursery staff. For them a late arrival usually meant an arrival after the final clock‐time stipulated by the nursery staff. These results concur with the writings of feminist social scientists about time and gendered moral rationalities. Feminist social scientists have shown that women’s experience of time in particular need not align with the linearity of clock‐time, which was developed to regulate paid labour and public life (see above)—historically activities that were mostly conducted by, and sometimes restricted to, men. Linear, future‐oriented and instrumental clock‐time is, therefore, a masculine time (Odih, 1999; Davies, 2001). Women’s lives, however, tend to have a different temporal structure; their temporal experience is often intimately associated with caring for others (Davies, 2001; Adam, 2004). Feminine time is, therefore, relational, it is “shared rather than personal and sensitive to the contextuality and particularity of interpersonal relations” (Odih, 1999, pp. 21–22) and more focused on the practicalities of the here and now, including the immediate surroundings of their children as in Kate’s case.

These feminine and masculine times intersect with gendered moral rationalities, which are culturally defined notions of what is appropriate to do (morality) for parents as mothers and fathers (gender) informing everyday decisions and behaviours (rationality) (Duncan et al., 2003). Some feminist theorists have argued that women tend to approach moral dilemmas more contextually, empathetically and with more attention for the interconnections with others than do men, although others have criticized such thinking for its essentialist tendencies (Tronto, 1993). ³ Still, there is ample evidence that mothers’ choices with respect to the juggling of employment and care‐giving are guided first and foremost by what they consider best for their children (Duncan et al., 2003; Vincent et al., 2004). Though difficult to generalize, this tends to be different for many fathers for whom it is more ‘natural’ to let their employment role of primary income provider prevail over that of a direct care‐giver for children and other dependants (Aitken, 2000; Vincent et al., 2004). Such gender differences are to a substantial degree the product of socialization and acculturation to dominant cultural scripts about what is appropriate to do as mothers or fathers. The author’s interviews suggested that, notwithstanding differences between mothers and between fathers, these dominant cultural scripts about parenting reverberate in most mothers’ and fathers’ moral notions about time‐keeping when collecting children from the nursery. ⁴

In summary, an exclusive concern with clock‐time might result in a failure to grasp how women in particular cope with coupling constraints in everyday life. Given that clock‐time and other sorts of time are partially connected, travel behaviour researchers could at least try to express the event‐based time embedded in the presence/absence of individuals and artefacts in terms of clock‐time. While this might not result in a fully accurate representation of that time, this procedure would still be preferable to not considering this contextual time at all. The discussion has also indicated the importance of addressing the moral and ideological dimensions of time and time‐keeping—a theme that will be explored in greater detail in the following subsection.

Boundaries

This subsection centres on the difference between arrival times that are on time or acceptable and those that are too late or unacceptable using notions from STS studies and in particular from actor‐network theory (ANT) (Law, 2004; Latour, 2005). ANT originated in the sociology of science around 1980 when Bruno Latour and others started to ask what made scientists so powerful in contemporary societies. Their answer, in brief, was that scientists continually build and extend networks of heterogeneous agents, including not just human beings but also instruments, devices and other material artefacts through which all kinds of phenomena circulate (Latour, 1987). Agency—the capacity to act—is thus not located in the human mind but the effect of a continuous but not necessarily conscious process of aligning and ordering of humans and material items, or a characteristic of an actor‐network. Over time ANT has evolved beyond science studies; it now studies the processes of the formation, functioning, change and decline of networks in many different social fields. The approach has been criticized heavily for questioning, and sometimes discarding, all kinds of (in Western thought) taken‐for‐granted philosophical categories and distinctions, such as human versus non‐human, nature versus culture or technology versus society. However, its radical nature has also led to new perspectives on some of the questions of interest to sociologists, geographers, anthropologists and other social scientists.

In recent years researchers associated with ANT have shown much interest in boundaries and in boundary‐making (Mol and Law, 2005). Two insights from this field are particularly relevant for the present paper. First, it has been shown that boundaries between categories are more complex and less self‐evident than common sense may suggest. Such boundaries should not be taken for granted because they are constituted in actual practices and (hence) often unstable, multiple and porous. As shown below, the interviews suggest that these notions are also applicable to time‐keeping and time–spaces of arrival. Second, classic ANT work can be used to explain how such boundaries have come into existence through networks of texts, instruments/devices and the disciplining or drilling of people (Law, 1986; Thrift, 1996).

That boundaries on time–spaces of arrival are constituted in actual practices is already evident from the third section. The text fragment from the interview with Kate points toward a relational boundary between arriving on time and too late that, emerging out of the rhythms of parents collecting their children and social comparison. Other parts of the interview with Kate also suggested that this relational boundary is relatively unstable and porous—the event‐based (space)time may be relevant under normal conditions but less so in unusual circumstances causing unforeseen delays (extremely heavy traffic, for instance). Boundaries on time–spaces of arrival are also multiple; different versions (can) coexist at a single moment (Mol and Law, 2002). In addition to the event‐based (space) time, there are two clock‐based boundaries at 17.45 and 18.00 hours. The latter corresponds with the official ending time of the staff’s workday; the former with the time the children should have been collected and gives the staff the opportunity to tidy up the nursery for the next day. These clock‐times are imposed by the nursery staff who try to control parents’ rhythms of collecting their children through standardization and regularity (Glennie and Thrift, 1996): parental practices are disciplined to be or become similar to one another, as well as to become routinely repetitive over time. Of the boundaries on time–spaces of arrival discussed, the clock‐based boundaries tend to be the most stable ones, because they are produced and sustained through networks of interwoven texts, instruments and coercive disciplining practices (Law, 1986; Thrift, 1996). Nursery regulations, the contracts between the childcare centre and parents, and sometimes the centre’s website contain texts informing parents about the clock‐times at which children should have been picked up and about the repercussions of violations of the rules. These can vary from rebukes by the nursery’s head to monetary penalties to the dissolution of the contract between the nursery and family household. Parents and nurseries generally were reluctant to talk about what actually occurs in situations of repeated late arrivals; however, it became clear that increasingly more nurseries have adopted a system of yellow and red cards as in football, whereby staff members can give a yellow card as a serious warning to parents after repeated late arrivals. A second yellow card means red and thus expulsion or the end of the contract. Thus, parents are disciplined or drilled through the employment and circulation of various instruments and devices, such as paperwork (contracts, letters, booklets, etc.) and money, as well as clocks and watches (Latour, 1997). It is in its function as a seemingly neutral, common metering system that clock‐time’s properties of standardization and homogeneity are exploited to its fullest extent. The disciplining is also borne out by the fact that parents usually try to mobilize others—partner, parents (in‐law), other relatives or neighbours—when they think they will not arrive before 18.00 hours (Schwanen, 2007).

Although such systems of documents, devices and drills have resulted in the internalization of the final clock‐based boundary in parents’ minds, there is still a degree of indeterminacy in actual situations as parents can resist the nursery’s attempts to order their arrivals at the childcare centre (De Certeau, 1984). The interviews show that some parents arrive later than 18.00 hours every now and then, although a delicate balance exists between what nurseries tolerate or not: the arrival should not be too late; individual parents should not arrive late too often; and the total number of parents arriving late should be small. Thus, even the final clock‐based boundary is imbued with some porosity.

The indeterminacy and openness of clock‐based boundaries is also evident in situations where they shift on relatively short notice and lose their significance as determinants of travel behaviour altogether. Here is another example from the qualitative study, which focuses on collecting children from school at 15.00 hours rather than from the nursery in the early evening. Mila works in Utrecht and commutes by bicycle. As she finds leaving her office to collect the children when her colleagues are still at work rather difficult; she regularly leaves later than intended (thereby sacrificing her planned safety margin). On the day before the interview a meeting with a colleague made her leave later than intended:

I left at almost ten to 3 from work though that is actually too late to pick up the children at 3 o’clock. […] This does come with feelings of stress um, because you think when you leave: will I be on time? And will I not have an accident? What happened was that I didn’t cross [the street] at the traffic lights but 100 meters further and because of that I hit the curb stone really hard and I thought: oh, it wouldn’t be handy if I get a flat tire or fall unexpectedly. This just didn’t happen but I did feel upset. […] In fact I took more risk at the beginning and when this had happened I thought: no, this is not smart. […] Then I thought I’d rather arrive late, because there are other parents and they know the situation and they’ll wait, for this is not quite the first time of course.

This excerpt makes clear that boundaries on acceptable arrival times can shift swiftly as a result of (unforeseen) events encountered in the physical world. It thereby illustrates the situation‐specific, contingent and open nature of time–spaces of arrival. It is the interactions between human minds and various material items as well as the knowledge gained from experience about the practices of other parents at the school playground who will look after her children that collectively account for the dynamism in, and shifting boundaries on, Mila’s space–time of arrival.

Towards Quantification

The preceding section has foregrounded the relevance of different times and of norms and ideology in how parents deal with coupling constraints, as well as the importance of situation‐specificity, contingency and indeterminacy in such coping. Some explanation for how differences between acceptable and unacceptable arrival times have come into existence has also been provided. While foregrounding the complex interactions between human beings, material objects, norms and emotions relevant to people’s negotiating of coupling constraints is theoretically appealing, commonalities and regularities in people’s experiences and tactics should not be disregarded. These are, after all, starting points for policy‐making. Furthermore, although qualitative research yields crucial insights, many researchers and policy‐makers will be concerned about the generalizability and transferability of the outcomes. Quantitative analysis using data for larger groups of respondents might alleviate some of these concerns.

The literature review in the second section indicated that transport studies about how travellers cope with uncertainty about the performance of transport systems have made important contributions to scientific understanding of how people cope with coupling constraints. Within this field, cumulative prospect theory (CPT), proposed by Kahneman and Tversky (2000), is rapidly gaining momentum as an alternative to expected utility theory (EUT) ⁵ (Von Neumann and Morgenstern, 1944), which has also been used in research about travellers’ coping with uncertainty (e.g. Batley, 2007). Drawing on CPT enabled the author to obtain further insights into parents’ coping with coupling constraints in at least two respects. First, it allows the theoretical framework outlined in the third section to be extended through the concept of the reference point (see below) and because it can be used to address the ramifications of unreliability and variability in travel times to the nursery more directly than the sociological and geographical approaches drawn upon allow. Second, using CPT enables quantitative assessments of travellers’ coping with uncertainty in a way that is better compatible with the theoretical framework above. Most importantly, unlike EUT, CPT recognizes that choices are situation dependent through its accommodation of reference dependence and does not radically separate rationality from emotionality (Kahneman and Tversky, 2000; Kahneman, 2003). At the conceptual level CPT appreciates the relevance of feelings and bodily sensations to everyday common‐sense decisions more directly than does EUT. Nonetheless, it still presumes that meaningful behaviour is the outcome of conscious thought; such pre‐reflective forms of acting as habitual action—so important in repetitive travel behaviour (Verplanken et al., 1997; Fujii and Gärling, 2003)—are not taken into consideration directly (though they do feature prominently in some of Kahneman’s other work). In short, CPT may have its imperfections for analysing the impact of unreliable transport systems in parents’ coping with coupling constraints, but it does allow one to investigate the notions about temporality, gender, and the multiplicity of boundaries between on‐time and late arrivals for larger groups of parents than qualitative, in‐depth methods would allow.

Reference Points

CPT was developed as a descriptive‐realistic approach for understanding how people make choices whose outcomes are not known in advance, such as whether or not to take an umbrella (Kahneman and Tversky, 2000). Like other decision theories, CPT assumes that such issues can be conceptualized as gambles whereby each possible outcome is associated with a corresponding probability. CPT builds on EUT in assuming that decision‐makers will choose the option with the highest utility or value is chosen but differs in terms of how values are defined. CPT proposes decision‐making as a two‐step process. In the editing phase possible outcomes are coded as gains or losses relative to a reference point and probabilities transformed into decision weights. ⁶ The option with the largest sum of edited outcomes multiplied by decision weights is then chosen in the evaluation phase. CPT accommodates these operations through two functions whose parameters can be determined empirically: an ‘S’‐shaped value and an inversely ‘S’‐shaped weighting function (Figure 1). Note that collectively these two functions reflect a fourfold pattern of risk attitudes (Tversky and Kahneman, 1992): people are risk aversive when facing gains with medium and high probabilities and (very) unlikely losses, but risk seeking in the case of (very) unlikely gains and losses with medium and high probabilities.

Figure 1 Cumulative prospect theory’s (CPT) value and weighting functions (Tversky and Kahneman, 1992)

Figure 1 shows that the reference point is a crucial concept in CPT. At zero the value function is kinked, suggesting that losses loom larger than gains. This phenomenon is called loss aversion and reflects that the (anticipated) emotions of losing an object or phenomenon bear heavier on choices than those associated with obtaining the very same object or phenomenon. Closely associated with loss aversion is the phenomenon called preference reversal: outcomes framed as losses may result in different choices than outcomes framed as gains, even if they are objectively identical. The phenomena of loss aversion and preference reversal are not only interesting because there is robust empirical evidence to support them; they also imply that it is not the inherent qualities of outcomes that determine how they are valued but instead their relations to other outcomes or the situations of which they are part.

While this relationalism aligns CPT with the theoretical framework outlined in the third section, the core concept of the reference point has remained somewhat under‐theorized in CPT. According to Kahneman and Tversky (2000, p. 16), the:

reference point is largely determined by the objective status quo, but […] also affected by expectations and social comparisons.

The existence of multiple rather than a single reference point in real‐life situations has also been recognized (Kahneman, 1992). However, the existence of reference points is primarily explained in terms of limits of the capacity of the human mind to process information (Kahneman, 2003); there is little appreciation for their socially constructed nature.

The understanding of time–spaces of arrival—and PT’s conceptualization of reference points—can be advanced if the socially constructed, temporal boundaries in the third section are identified as potential reference points against which arrival times are evaluated. This makes it possible to test empirically to what extent the different times discussed there function as reference points. Specifically, the researcher can first estimate the coefficients characterizing CPT’s value and weighting functions using specifically estimated stated adaptation data, then manually manipulate the reference point(s) as presented to the respondents, and finally re‐estimate the coefficients on the manipulated data to compare the goodness‐of‐fit (GOF) of the re‐estimated set of coefficients against the original coefficients. ⁷ An improvement of the model’s GOF after re‐estimation suggests that the new (set of) reference point(s) corresponds more closely with parents’ perceptions and presumably also to their real‐life decision‐making processes. If no improvement occurs, this is taken to indicate that the original reference point(s) resonate better with participants’ real‐life experiences.

Stated Adaptation Exercise

The multiplicity of reference points has been tested using specifically collected stated adaptation data. The endeavour is described only briefly here for sake of brevity; Schwanen and Ettema (2007) provide a detailed motivation of the format and procedures followed. Participants were asked to indicate how they would cope with an unforeseen delay as a consequence of transport failure (e.g. due to congestion, train delay or vehicle breakdown) when travelling from the workplace to the nursery to collect the children. Since the qualitative phase had indicated that one’s partner is the most likely person to be mobilized in this specific situation (Schwanen, 2007), participants could choose between travelling to the nursery themselves (coded as zero in the analysis) and letting their partner collect the children in their place (coded as 1). Choices were supposed to be based on comparisons of their own possible arrival times and associated probabilities and those of their partner. On the basis of the qualitative phase, the choice to mobilize one’s partner was interpreted to indicate that participants considered their own possible arrival times inappropriate or unacceptable.

The stated adaptation exercise consisted of 11 questions divided into three blocks that all introduced a specific reference point. The first block comprised questions in which arrival times were framed relative to the clock‐based time that the children should have been collected (set at 17.45 hours). The second consisted of questions with arrival times relative to the clock‐based time at which the nursery formally closes (set at 18.00 hours). Finally, the third block contained questions with arrival times relative to the event‐based time that almost all children of the child’s group have been collected and s/he is left behind more or less alone (set at 17.35 hours after discussions with test‐phase participants). Figure 2 shows a sample question from with arrival times framed relative to the time that the nursery formally closes.

Figure 2 Sample question from the experiment.

The stated adaptation exercise was incorporated in a mail‐out/mail‐back survey among dual‐earner family households with young children about how they organize and conduct the chauffeuring of children to/from childcare providers and primary school. Respondents were recruited via 19 nurseries and five elementary schools in the Utrecht region in the Netherlands. A total of 455 parents from dual‐earner family households whose youngest child was zero to 8 years old and attended daycare or after‐school care for at least one day per week were selected for the analysis in this paper. There was an equal split of mothers and fathers, though white, middle‐class, higher‐educated parents with children aged zero to 3 years were over‐represented (Table 1). These statistics are largely consistent with national‐level statistics about the use of nurseries, which are most widespread among highly educated mothers (Portegijs et al., 2006). Empirical tests have shown that the over‐representation of higher‐educated parents has had no impact on the study outcomes.

Table 1. Characteristics of participants in the stated adaptation exercise

		%
Educational attainment	low/medium	19.8
	high	80.2
Country of origin	Netherlands	94.9
	other	5.1
Gender	mother	49.7
	father	50.3
Number of children	one child	48.8
	two children	45.3
	three children	5.9
Age of child(ren)	at least one child aged 0–1 years	63.3
	at least one child aged 2–3 years	50.8
	at least one child aged 4–5 years	14.3
	at least one child aged 6–8 years	7.0
Childcare use	daycare only	84.5
	day and after‐school care	12.2
	after‐school care only	3.3
Residential location	city of Utrecht	37.6
	surrounding suburbs	62.4

A binary choice model was specified as a function of the coefficients characterizing CPT’s value and weighting functions. These were estimated with a genetic algorithm (GA) that maximizes a log‐likelihood function by applying a probabilistic search procedure (details are available in Schwanen and Ettema, 2007). The GA approach was employed because of non‐linearity in the parameters to be estimated.

Testing for Multiple Reference Points

This penultimate section discusses the results of the quantitative tests of the extent to which the event‐based and the two clock‐based times function as reference points or boundaries on parents’ time–space of arrival at the nursery. The times used as reference point were manipulated manually; all arrival times in the original data were recoded relative to a single reference point—the clock‐based time at which the children should have been picked up (17.45 hours). The coefficients characterizing CPT’s value and weighting functions were then re‐estimated and compared with those estimated on the data as originally collected. A series of model specifications were fitted to the original data, from which a model with different value functions for fathers and mothers had the best GOF (Schwanen and Ettema, 2007). GAs can produce multiple solutions that are statistically equivalent in terms of GOF and this also occurred for the best model specification on the original data; therefore, two variants of this model are included in Table 2. The coefficients in the original model show that the value function is slightly S‐shaped or almost linear for father but inversely S‐shaped for mothers. Thus, as the extent of lateness is larger, 1 minute of extra lateness bears heavier on mothers’ decisions, but not on those by fathers, which is consistent with the results about gendered moral rationalities emerging from the qualitative phase (see the third section). The loss aversion coefficient λ and the parameter characterizing the weighting function γ are consistent with expectations: losses loom larger than gains—note the correspondence to Small (1982)—and small probabilities are overweighted and larger ones underweighted, respectively. Detailed discussion of the implications of the estimated coefficients is beyond the scope of this paper (but see Schwanen and Ettema, 2007); however, the complete set of estimates reflects the fourfold pattern of risk attitudes found by Tversky and Kahneman (1992), although the differences in attitudes found here are much smaller, which is probably attributable to differences in decision task, experimental design and type of respondents.

Table 2. Estimated coefficients for cumulative prospect theory’s (CPT) value and weighting functions

	Original data^a
	Model 1	Model 2	Manipulated data
α for mothers (shape, value function)	1.09	1.07	1.33
α for fathers (shape, value function)	0.99	0.79	0.99
λ (loss aversion coefficient)	1.36	1.40	3.99
γ (shape, the weighting function)	0.84	0.84	0.83
Goodness of fit^b	0.539	0.539	0.552
^aA genetic algorithm used for the estimation can produce multiple statistically equivalent solutions. The equivalent sets of parameters with the smallest and largest gender difference are presented here (cf. Schwanen and Ettema, 2007)^bIndicator ranges from zero to 1, with zero indicating a perfect fit.

After manipulation of the reference points, the gender difference in the overall shape of the value function as shown by the values for α and the extent of loss aversion as represented by λ become more pronounced, but the coefficient for the transformation of probabilities, γ, remains basically unaltered. What is most important here, however, is that the level of the GOF shifts upward, indicating that the model as a whole performs considerably worse. Thus, the condition with a single reference point does not correspond to parents’ perceptions as closely as condition with multiple reference points; parents do appear to evaluate arrival times against multiple reference points. Further to the qualitative study, the quantitative analysis presented here suggests that the differences between arrival times that are considered appropriate or acceptable (i.e. on time) and those that are inappropriate or unacceptable (i.e. too late) are less singular, stable and unambiguous than common sense might suggest.

Synthesis

This paper has reviewed past attempts to operationalize the ‘when’ dimension of coupling constraints and provided a complementary perspective on how people negotiate such constraints in everyday life. This complementary approach draws on human geography, sociology, and psychology and has developed in tandem with mixed‐method research—interviews and a stated adaptation exercise informed by CPT—of how dual‐earner family households in the Utrecht region of the Netherlands organize and negotiate the chauffeuring of children to/from childcare and primary school on days that both parents engage in paid labour. At the heart of this approach is the concept of the time–space of arrival, or the time span appropriate for arrival at a certain destination.

Various properties of time–spaces of arrivals have been proposed. Time–spaces of arrival are constituted relationally: rather than internal to the human mind, they emerge out of the interactions between people, artefacts and other material items in concrete physical space at specific moments in chronological time. Additionally, they pertain to multiple types of time of which clock‐time is but one; moral and ideological dimensions of time in particular have been shown to be important. Moral notions of appropriateness are, among others, tied to the degree of eurhythmia and arrhythmia (Lefebvre, 2004), or con‐ and discordance with the rhythms of other individuals, objects, nature, and the cosmos, respectively. Such moral notions are also gendered, with mothers by and large more focused on the here and now in which their children are situated and earlier inclined to mobilize their partner if they (fear they) will not reach the nursery on time than fathers. Finally, boundaries between on time or acceptable arrival times and late or unacceptable arrival times are more complex than common sense suggests. Multiple boundaries may coexist for different individuals and even for the same person, although some boundary or boundaries may be more stable than others. Such stability is likely to emerge out of systems of interacting documents, instruments (including clocks and watches), and the disciplining of people. However, disciplinary practices can be—and are—resisted, suggesting that time–spaces of arrival are imbued with at least some indeterminacy and openness. Shifts in boundaries may be irregular and even exceptional, but they do give parents (and others) the elbow room to sort out frictions between commitments both practically and mentally.

The paper has touched upon some more general points regarding travel demand analysis that go beyond its substantive focus on parents’ coping with coupling constraints. First, the paper has tried to show that time should not be equated to clock‐time. Otherwise, failure to capture the richness of (spatio)temporal influences on travel decisions in general and/or for certain social groups—mothers in the current example—in particular is likely to result. Second, the author believes that the centrality of social, moral and emotional factors extends beyond travellers’ coping with arrival times, and is equally important to mode, destination and other choices. While interest in social influences on travel decisions currently proliferates (e.g. Larsen et al., 2006; Páez and Scott, 2007), attention for the experiential side of travel behaviour or the influence of culturally defined norms and values has so far been very modest. Finally, therefore, travel behaviour studies have much to gain from engaging with less familiar theoretical traditions. The field has diversified enormously in the past ten to 15 years with researchers importing ideas about heuristics, learning, cognition, etc. The work reported here provides preliminary evidence that CPT and various approaches within social theory can also be blended profitably into travel behaviour studies. CPT holds the promise of providing insights about coping with uncertainty that are more representative of how people actually make choices, among others because it considers both cognitive limitations on decision‐making and situation dependence. Social theory can be used to understand the relational character of phenomena, the multiplicity and instability of seemingly distinct analytical categories, and the myriad of ways that taken‐for‐granted power differences operate and are being (re)produced. These traditions should be seen as complementary to existing paradigms within travel behaviour studies, able to answer different policy‐relevant research questions.

The last remark begs the question whether the proposed perspective on coping with coupling constraints is actually capable of providing new, policy‐relevant insights. The answer is probably affirmative. Before commencing the reported study, the author considered the extension of the clock‐based opening hours of childcare services as being the most effective strategy for easing some of the conflicts between paid employment and chauffeuring children to/from childcare. However, the relevance of social comparison and the gendered moral rationalities suggested by the empirical analysis have tempered that optimism substantially. A suite of additional measures to reduce parental concerns about children being left behind alone and their well‐being are also required (e.g. meals being served to the children at the nursery) for extended openings hours truly to make a difference to parents’—and especially mothers’—activity schedules. And even then the effects may be rather small, because parents might well prefer their own care‐giving or that by next of kin (Portegijs et al., 2006). Nonetheless, much more theoretical and empirical work is required before these and related ways of mediating travel choices through policies can be implemented.

Acknowledgements

The help of Dick Ettema in preparing the estimation and the comments and suggestions of the anonymous reviewers is gratefully acknowledged. The research reported here was made possible by the Netherlands Organization of Scientific Research (NWO) (Grant No. 154‐03‐058).

Notes

1. Interviewees’ original names have been replaced to ensure their anonymity.

2. Beyond these issues, the cultural specificity of the mechanisms captured in such scheduling models as proposed by Small and others should not be overlooked because “[t]iming and timeliness are defined differently in various cultures and under different historical circumstances” (Hareven, 1991, p. 169, quoted in Glennie and Thrift, 1996, p. 288; also Adam, 1990; and Glennie and Thrift, 2005). There is increasing evidence that Anglo‐Germanic conceptions about time‐keeping may not hold elsewhere, which does not invalidate those scheduling models but rather calls for research into their transferability across geographical contexts.

3. Essentialism is the tendency to attribute inherent or intrinsic properties to certain states or categories. These properties are believed to be invariant across times and spaces. Essentialist thinking has been attacked vigorously in contemporary social theory under the influence of such thinkers as Foucault and Derrida.

4. To a certain extent these gender differences can be captured in the schedule delay models based on Small (1982) if a sex‐specific term were included in the utility function and even better if a heterogeneity within each of the sexes was allowed (e.g. by using a mixed‐logit model [Hensher and Greene, 2003] for departure time choice). However, this way of incorporating variations between, and within, men and women in timing decisions does not address the difference between masculine and feminine time.

5. Von Neumann and Morgenstern (1944) provided EUT as an axiomatic theory of choices made without advance knowledge of the consequences (risky choices). While useful as a normative framework, EUT’s value as a descriptor of how people actually make choices has long since been questioned. Various generalized utility theories have been proposed that have relaxed or dropped one or more of EUT’s axioms to arrive at theories that correspond more closely with actual choices. CPT is the most comprehensive of these efforts.

6. Suppose the wages of all employees in a firm are increased by 10%, but one employee gets a 5% raise. In absolute terms and according to EUT the latter is a positive outcome; however, in relative terms and in CPT this would be coded as a loss rather than a gain as it is lower than the reference level of 10%.

7. Senbil and Kitamura (2004) employed a broadly similar approach in their study of departure time choices for home‐to‐work trips using revealed‐behaviour data from a bedroom community in the Osaka‐Kyoto area of Japan.