**1. Introduction**

Traffic congestion has long been a serious transportation-related issue that people confront in their daily life in the U.S. [1–3]. Drivers in the U.S. wasted 7 billion hours on roads in 2015 due to delays caused by traffic congestion [4]. Moreover, it is expected that traffic congestion will intensify in the future as more people will move to urban areas [5]. Transportation and health researchers have thus considered traffic congestion exposure as a critical factor that influences individuals' physical and mental health. For example, a number of studies revealed that higher exposure to traffic congestion may be associated with escalated heart rate and blood pressure [6,7], heightened urinary catecholamine (a stress-related hormone) [8], and negative health outcomes [9–11]. In addition to these physical tolls, studies have also shown that exposures to traffic congestion may be linked to psychological stress [12–17] and reduced well-being [18,19]. Furthermore, some studies have argued that longer commuting time, which is worsened by traffic congestion, may harm people's work-family balance [20] or even increase the likelihood of being involved in domestic violence [21].

To accurately assess the effects of traffic congestion exposures on human health, it is important to accurately measure individual exposures to traffic congestion. Specifically, in terms of measuring traffic congestion exposures, most previous studies used area-based aggregate data (e.g., [9,21,22]) or focused only on commuting trips (and thus disregarded non-commuting trips) (e.g., [11,18,19]). We argue that these approaches in previous research may lead to erroneous assessments of individual exposures to traffic congestion, which may in turn lead to erroneous evaluations of the health impacts of traffic congestion because they did not consider individuals' unique activity-travel patterns.

To address the limitations of previous studies, this research examines whether individual exposures to traffic congestion are significantly different between assessments obtained with and without considering individuals' activity-travel patterns in addition to commuting trips. We used crowdsourced real-time traffic congestion data and the activity-travel data of 250 individuals in Los Angeles to compare these two assessments of individual exposures to traffic congestion. The results indicate that ignoring individuals' activity-travel patterns may lead to inaccurate assessments of their exposures to traffic congestion.

### **2. Limitations of Previous Studies**

Previous studies on traffic congestion have several limitations, some of which this study seeks to address. First, existing research used data aggregated over areas such as metropolitan areas or zip (postal) code areas. For example, Levy et al. [9] assessed the health impact of PM2.5 exposure associated with traffic congestion on mortality and monetized the value of mortality risk in 83 metropolitan areas in the U.S. By using zip code areas in the Los Angeles Metropolitan Area, Beland and Brent [21] noted that traffic congestion may lead to a higher risk of being involved in domestic violence. Brauer et al. [22] examined people's traffic-related air pollution "exposure zones" (defined as a 500-m buffer zone from highways or a 100-m buffer zone from major arterials) and observed that 32% of people in Canada live in exposure zones.

Although these studies provide useful insights into the health impacts of traffic congestion, their estimations of traffic congestion exposure may be erroneous. Part of the reason for the error is because the units of analysis are areal units for which individual data are aggregated (e.g., buffer zones, zip code areas, or metropolitan areas); and the modifiable areal unit problem (MAUP) may contribute to some of such error. This means that previous studies presumed that individuals in the same areal unit are exposed to the same level of traffic congestion when estimating traffic congestion exposure. However, this assumption is problematic because each individual may have distinctive activity-travel patterns and thus may have different levels of exposure to traffic congestion and experience different health impacts [23,24]. In other words, since individuals have idiosyncratic activity-travel patterns, different individuals may be exposed to different levels of traffic congestion in complex and unique ways even when they live in the same area.

Second, previous research focused only on commuting trips while ignoring other types of trips, where individuals may also be exposed to significant traffic congestion. For instance, Olsson et al. [25] found that commuting satisfaction may affect overall happiness based on a survey of 713 commuters. Using a 23-year longitudinal dataset of 2736 commuters, Sandow et al. [11] showed that there may be gender differences in mortality risks due to longer commuting. Despite the meaningful results of these previous studies, focusing exclusively on commuting trips while not considering other components of individuals' activity-travel patterns may lead to erroneous estimations of their exposure to traffic congestion for the following two reasons.

First, commuting trips account for only a small proportion of individuals' total travel. Although commuting trips seem to constitute the most significant portion of our trips at first glance, almost 70% of trips in the U.S. consists of non-work trips according to the 2009 National Household Travel Survey [26]. Thus, considering only commuting trips may result in erroneous estimations of traffic congestion exposure because people may also experience traffic congestion when undertaking other types of trips, which also comprise their travel and are overlooked by previous studies.

Second, and more specifically, individuals also experience traffic congestion when undertaking trips during non-peak hours. At first sight, it sounds reasonable that commuting trips in peak hours (e.g., 7–9 A.M., 3–6 P.M.) are the only time when individuals are exposed to traffic congestion. However, this may not be true for large cities, where serious traffic congestion occurs almost all day long in certain road segments (e.g., [27–29]). This suggests that individuals are exposed to traffic congestion in complex ways in their daily life. They may be exposed to different levels of traffic congestion when undertaking not only commuting trips during peak hours but also other types of travels during non-peak hours.

We thus argue that, for these two reasons, the approaches used in previous studies on individual exposure to traffic congestion and its health impact may exacerbate the uncertain geographic context problem (UGCoP). The UGCoP is a critical methodological issue, and one of the ways in which it may be encountered is when people's exposure to environmental contexts or risk factors (e.g., traffic congestion and air pollution) is inaccurately estimated as a result of ignoring their activity-travel patterns [23,24]. Recent studies have shown that using inaccurately estimated exposures to environmental pollutants may lead to serious inferential errors or misleading results when examining the health impacts of such exposures [30–35].

Further, these approaches in previous studies may aggravate the neighborhood effect averaging problem (NEAP) that arises when individuals' unique activity-travel patterns are ignored [36]. The NEAP suggests that when the probability distribution of residence-based exposures approximates a bell-shaped distribution, individual exposures tend to converge toward the average if individuals' activity-travel patterns are considered when compared to exposures obtained when such patterns are not considered. As one of the manifestations of the UGCoP, the NEAP thus suggests that ignoring individuals' activity-travel patterns may lead to erroneous exposure estimations because of neighborhood effect averaging [36]. Eventually, this may also result in inferential errors or misleading results when researchers investigate the health effects of these exposures [23,24].

In traffic congestion exposure and health studies, these two methodological issues—the UGCoP and the NEAP—arise because of the following two reasons. First, the issues are caused by the spatiotemporal variations in traffic congestion intensities. Although at first glance, the levels of traffic congestion seem to be homogeneous over space and time in large metropolitan areas, this may not be true. For example, intensities of traffic congestion in non-peak hours may not be as severe as those in peak hours (i.e., temporal variations). Also, even for the same time of day, people may be exposed to different traffic congestion intensities based on where trips occur (i.e., spatial variations). Second, the issues arise because individuals are undertaking various types of trips rather than undertaking commuting trips only. Therefore, individuals may be exposed to traffic congestion in complex spatiotemporal ways when they are undertaking commuting trips and other trips [23,37,38].

To sum up, the approaches in previous research examining the effects of traffic congestion exposure on human health largely ignore the UGCoP and the NEAP. This is because both the spatiotemporal variations in traffic congestion and individuals' unique activity-travel patterns increase contextual complexities when measuring individual exposures to traffic congestion. These complexities may lead to erroneous assessments of individual exposures to traffic congestion if their activity-travel patterns are overlooked. Eventually, using inaccurately estimated traffic congestion exposure may exacerbate inferential errors or lead to misleading results when investigating the health effects of traffic congestion exposure.

In light of the relative ignorance of critical methodological issues like the UGCoP and the NEAP in past research on traffic congestion exposure and health, this research seeks to fill this gap. Specifically, this research examines whether individual exposure to traffic congestion is significantly underestimated when individuals' travels beside commuting trips are ignored. To achieve this research goal, we seek to answer the following three questions in this paper by utilizing crowdsourced real-time traffic congestion data and activity-travel data of 250 individuals in Los Angeles, California.

The first question is: Do spatiotemporal variations exist in traffic congestion intensities? In other words, we will investigate whether traffic congestion intensities are different over space and time. Spatiotemporal variations in traffic congestion intensities indicate that two approaches adopted by previous studies (i.e., using data aggregated over areas like census tracts and focusing on commuting trips only) may exacerbate the UGCoP and the NEAP, which is problematic.

The second question is: Will exposures to traffic congestion based on commute-only assessments be significantly lower than those obtained from assessments that also consider individuals' activity-travel patterns in addition to commuting trips? In other words, we will compare individuals' exposure to traffic congestion obtained from two assessments: one that only considers commuting trips and one that also considers individuals' activity-travel patterns in addition to considering commuting trips. If we find that exposures to traffic congestion based on commute-only assessments are significantly lower than those obtained from assessments that also consider individuals' activity-travel patterns in addition to commuting trips, more attention is needed to address the UGCoP and the NEAP in future research on traffic congestion exposure and health.

Lastly, the third question is: How are individuals uniquely exposed to traffic congestion as a result of their activity-travel patterns? In other words, can we observe individuals' idiosyncratic activity-travel patterns and their associated exposures to traffic congestion? Answering this question will help researchers develop an in-depth understanding of the unique characteristics of individuals' activity-travel patterns and how such unique characteristics may affect the accurate assessment of traffic congestion exposure.

### **3. Study Area and Data**
