1. Introduction
Social sustainability has received significantly less attention in the literature and in practice than environmental and economic sustainability despite the fact that it is supposed to be one of the three pillars of sustainability [
1]. While elaborating upon this issue, Murphy [
2] and Nilsson et al. [
3] suggested equity as one of four fundamental considerations of social sustainability. To assert that equity and justice are critical for sustainability, Agyeman [
4] (p. 751) argued that “…inequity and injustice…are bad for the environment and bad for a broadly conceived notion of sustainability”.
Since the mid-1990s, it has become customary for US practitioners and researchers to cite environmental justice as a way to address inequity and injustice in transportation. A strong connection between sustainability and environmental justice has been shown by many, including Agyeman et al. [
5]. The concerns about inequity in US transportation actually began in the 1960s and 1970s, when researchers were more accustomed to using the term transportation disadvantage to describe inequity and injustice. An entire stream of research on transportation disadvantage evolved with the recognition that certain populations could not travel like others because of personal impediments combined with systemic constraints, such as unavailability, unaffordability, and inconvenience of transportation. Another stream of research, commonly known as the spatial mismatch literature, also evolved around that time, where the primary focus was on the impact of lacking transportation on unemployment and poverty.
Almost all past research and even government policies in this country assume that people’s inability to make trips varies by population characteristics. For example, it has been widely held for decades that people with disabilities, people with low income, minority populations, and older adults are unable to travel like others [
6,
7,
8]. Various Presidential Orders and subsequent orders from the US Department of Transportation since the 1990s have also mentioned specific population groups being underserved. However, actual research to demonstrate variations in trip deprivation has been rare in this country. The reason is that until data from the 2022 National Household Travel Survey (NHTS) by the FHWA [
9] were released in November 2023, it was difficult to meaningfully examine trip deprivation for the US population as a whole.
The subject matter of this paper is related to transportation equity analysis. In the US, metropolitan planning organizations (MPOs) are primarily responsible for conducting transportation equity analysis [
10,
11,
12]. When conducting equity analysis, MPOs follow guidelines from the US Department of Transportation that are based on federal laws and regulations. Since the mid-1990s, the primary objective of MPO equity analysis has been to ensure an equal distribution of benefits and burdens of transportation investments among all populations. Like the traditional transportation disadvantage literature, MPOs also focus on specific populations when conducting equity analysis, most commonly low-income people, people with disabilities, racial/ethnic/linguistic minorities, and older adults. Transportation researchers have often expressed concerns about the state of equity analysis by MPOs [
13,
14,
15,
16]. The critics have emphasized the importance of measuring accessibility and suggested approaches like minimum accessibility guarantee. However, the critics of current practices have generally adopted a theoretical approach that is devoid of empirical demonstrations with real data.
This paper is expected to be beneficial to both transportation planners and researchers. Planning agencies and professional planners will benefit from the methodological approach because similar efforts in the US context have not been made in the past. The study can motivate MPOs to collect household survey data resembling the novel NHTS data analyzed in this paper. When data are available, the methods used in this paper can be easily replicated by MPOs when undertaking human services transportation plans to identify disadvantaged populations and their residential location. This study can benefit researchers interested in transportation equity because of its methodological approach as well as its attention to human diversity. This paper presents a compelling case for researchers to pay greater attention to variations in people’s travel barriers when they attempt to promote the accessibility of places.
Despite being an invaluable source of information for understanding how American people travel, until the 2022 round of the survey, the NHTS was not very useful for understanding trip deprivation among American people. The 2009 round of the Survey [
17] included a variable identifying the survey respondents who did not make any trip on the survey’s travel day, but it did not record the reasons for the respondents not making trips. As a result, it was impossible to distinguish the people who did not travel because of free will from those who did not travel because of constraints like having a disability or unavailability of transportation. The 2017 Survey [
18] was an improvement over the 2009 Survey in that it inquired about the reasons for not making trips. However, like the 2009 Survey, the survey question on reasons for not making trips was asked only to those who did not make any trip on a specific day. Although the survey responses were objective in that the data were collected from people who actually did not travel, to draw meaningful conclusions about trip deprivation from a person’s behavior on a single day was difficult. Another difficulty with the 2017 survey question was that it allowed respondents to select 1 of 13 reasons (including “Don’t know” and “I prefer not to answer”), potentially forcing the respondents having multiple reasons to select the one they believed was the most appropriate. This resulted in a very small proportion of the survey respondents selecting transportation as the reason for trip deprivation.
The 2022 NHTS generated a vastly improved dataset compared with the previous rounds of the Survey for examining trip deprivation among American people. That is because the survey included a new special-topic question on ‘equity’, which inquired about taking fewer-than-planned trips in the previous 30 days, followed by another question inquiring about the reasons for taking fewer trips. The question on taking fewer trips was “In the past 30 days, did this person take fewer trips than they had planned on taking for any reason?” When the response was affirmative, the respondents were asked “Which of the following reasons help explain why this person took fewer trips in the past 30 days?” The respondents were given a list of ten reasons, out of which they could select as many as appropriate. The 2022 NHTS also retained the question on the reason for not taking any trips on the travel day meant for the respondents who remained stationary. The number of specified reasons was increased from 13 to 17, but similar to the 2017 Survey, the respondents were allowed to select only one reason. Although the question by itself is not very useful to draw conclusions about trip deprivation in general, because of the addition of the question on taking fewer trips in the past 30 days, it provides an opportunity to compare trip deprivation on 1 day with trip deprivation in 30 days. If trip deprivation is shown to be higher among certain populations by the responses to both questions, the evidence about transportation disadvantage certainly becomes stronger.
Caution is needed when drawing conclusions from survey data about variations in trip deprivation among different population groups. That is particularly the case when respondents are asked about trips they wish to make. To address the issue, as noted by Palm et al. [
19], past studies first inquired about realized trips, followed by an inquiry into additional trips they would have liked to make. However, asking about desired trips generally can also lead to misleading responses, since some people may think about only the most essential trips whereas others may think about the trips they crave [
20]. Fortunately, that issue is irrelevant for the 2022 NHTS data because both questions related to trip deprivation refer to events that have already occurred. The use of the word ‘planned’ instead of ‘desired’ in the question on making fewer trips in the past 30 days also adds an element of reality. Similarly, the responses to the question on trips on the travel day is grounded in reality because it was asked only when respondents did not make any trip.
The impact of travel need on trip deprivation cannot be ignored. Altshuler et al. [
20] argued that variability in people’s travel need adds a layer of complexity when examining people’s inability to travel. A person may not travel because they have no travel need, whereas another person may travel substantially because they have a high travel need. However, one cannot measure travel need based on the number of trips made because some people with higher need may not travel at all because of transportation or other constraints.
A related issue is that a proportion of the people who travel substantially may not only have higher travel need, but they may also experience a higher unmet mobility need, i.e., trip deprivation. Because this issue is particularly pertinent for the 2022 NHTS question on making fewer trips during the past 30 days, it has been explained with the help of a chart in
Figure 1. The chart shows the proportion of respondents taking fewer-than-planned trips during the past 30 days for three types of aggregated reasons by the number of trips those respondents made on the survey’s travel day (a detailed description of the reasons is provided later in the paper, but they are not important for the current discussion). Although, as expected, the proportions are high for the people who made no trip at all on the travel day, they are not always the highest. For example, on the lowest curve, the proportion taking fewer-than-planned trips in 30 days among those who made no trips on the travel day (7.6%) is not higher than the proportions of those who made two or three trips on the travel day (8.1% and 7.8%, respectively). Furthermore, the proportions do not decrease monotonically in any of the three graphs, indicating that despite making more trips, some people may experience greater unmet travel need than people making fewer trips.
A detailed discussion on travel need has been provided in
Section 2.3. Suffice to say at this stage that the logit models predicting trip deprivation for various reasons account for variations in travel need through the integration of a latent variable representing travel need. That variable was obtained through a confirmatory factor analysis (CFA) which used variables that are theoretically likely to affect travel need but are not likely to be affected by travel need or actual travel.
3. Data and Methods
3.1. Data Description
Although the 2022 NHTS Version 1.0 data were released in November 2023, this paper uses slightly updated data from Version 2.0, released in February 2024. The national sample dataset includes information for 16,997 persons aged 5 years or older (age 5+) belonging to 7893 households. People living in group quarters are excluded. The dataset also includes a file containing detailed information for 31,074 trips made by the persons in the person file. The national sample dataset from the 2022 Survey is significantly smaller than the 2009 and 2017 Surveys. For example, the national samples in the 2009 and 2017 Surveys included 25,510 and 26,099 households compared with 7893 households in the 2022 Survey.
The analyses in this paper are conducted mostly by using the information from the person file, where many of the household characteristics are attached to the persons. However, a few variables were also extracted from the household file because they were not included in the person file. Each file includes weights that correct sampling bias and inflate the data to the US national population. Although the person file includes three weights, this paper uses the standard 7-day weight (named WTPERFIN by the NHTS), which was also included in the previous rounds of the survey. Because direct application of that weight leads to almost all variables being significant in multivariate models due to the inflation of the sample, a revised weight was developed following Deka and Fei [
56]. That weight is essentially a scaled-down version of the original weight that corrects the sampling bias but does not inflate the sample to national population. The new weight, referred to as the revised weight hereafter, was calculated for all observations by multiplying WTPERFIN by the ratio of sample size (n) to population size (N).
Although NHTS data are available for people aged 5+ for most variables, the lower limits are not identical for all variables. In the 2022 dataset, the reasons for not making any trip on travel day are available for all people aged 5+, whereas the data on making fewer-than-planned trips in the past 30 days are available for people aged 16+. Consistent with census data, worker status is also available for people aged 16+. For greater uniformity, all analyses in this paper are conducted by restricting the dataset to people aged 18+ because people 18 and over are generally considered adults. The restriction of the survey sample from 5+ to 18+ decreased the sample size from 16,997 to 13,956.
The top part of
Table 1 shows how many respondents took fewer-than-planned trips in the past 30 days. As shown in the table, whether a person took fewer trips is known for 13,921 of the 13,956 respondents. Out of them, 3970 (28.4%) reported making fewer trips, and 9951 (71.3%) reported not making fewer trips. It could not be ascertained whether the remaining 36 respondents took fewer trips (0.3%).
As shown in the top part of
Table 2, a total of 10,341 (74.1%) out of the 13,956 respondents made at least one trip on the travel day, and 3615 (25.9%) made no trips. However, among the people who did not make any trip, the reasons for not making trips are available for only 2127 people (15.2% of 13,956). For the remaining 1488 (10.7%), data on reasons could not be ascertained from the survey responses, or the respondents did not know or preferred not to respond.
The responses of the 13,956 people aged 18+ to the questions inquiring about the reasons for making fewer trips in the past 30 days are presented in
Table 1. The responses to the question on the reasons for not making any trip on the travel day are presented in
Table 2. Both sets of responses are derived by applying the revised weight, for which the percentages are the same as they would be if the original weight attached to the dataset were used, but the number of respondents reflect the sample rather than the population. It should be noted that the percentages for the reasons for making fewer trips in the past 30 days are derived from a question that allowed for the selection of multiple reasons, whereas the percentages for the reasons for not making any trip on the travel day are derived from a question that allowed only a single selection. Hence, the percentages in the two tables cannot be compared, but the percentages for reasons within the same table are comparable.
Table 1 shows that 10 reasons were given to the responses to select from when they were asked about not making trips in the past 30 days. The reasons numbered 2 through 6 in the table more clearly relate to perpetual transportation constraints (lack of safety, cleanliness, reliability, availability, and affordability) than the other reasons. Reason 7, especially health problems, can also be perceived as a transportation constraint, because both legally and morally, people with disabilities are considered transportation-disadvantaged. Reason 8—not having time to travel—is ambiguous and hence difficult to interpret. For example, one person may have no time to travel because they plan to spend the day sunbathing, and another person may not have time because they need to take care of a sick parent. Reason 9—“concerns related to COVID-19”—was certainly important during and immediately after the pandemic, but it is not a perpetual transportation issue like reliability or affordability. Reason 10, “another reason” (i.e., something else), is obviously too ambiguous, and it is impossible to know what types of constraints the respondents had, if any. Finally, Reason 1—having home deliveries—is difficult to interpret as a constraint because if a respondent selecting that reason acquired something from somewhere, the home delivery relinquished the person’s need to travel. Based on the above observations, it was decided that three variables could be generated for multivariate analysis, the first representing all respondents who selected any of the ten reasons in
Table 1, the second representing reasons 2 to 6 (transportation reasons only), and the third representing reasons 2 to 7 (transportation and health reasons). In the first case, out of the 13,921 people with valid data, 3970 made fewer trips for any of the ten reasons in
Table 1, whereas the other 9951 did not make fewer trips. In the second case, 1072 people selected one or more of reasons 2 to 6, whereas the other 12,849 did not select any of those reasons. In the third case, 1614 people selected one or more of reasons 2 to 7, while the other 12,301 did not select any of those reasons. Note that aggregating the number of respondents in
Table 1 will not generate these numbers because a single respondent could select more than one response (e.g., a respondent selecting reasons 2, 4, and 5 is still one person).
In
Table 2, where the responses to the question on not making any trip on the travel day are shown, only reason 10—“No transportation available”—directly relates to transportation. However, only 56 respondents selected that option. Reasons 3 and 4, representing caretaking and disability, can also be considered long-lasting constraints based on moral grounds. Being personally sick or quarantining (reason 1) and being hospitalized or being otherwise confined (reason 12) are certainly constraints, but whether those reasons reflect long-term or short-term ailments is difficult to know. Regarding reason 7, household chores can be more easily conceived as a constraint than projects. Household projects are difficult to interpret as a constraint because the opportunity costs of the projects are not known. For example, if a person’s project is renovation of their dwelling, it is difficult to interpret the project as a constraint. Additionally, a connection between personal transportation and such projects is difficult to comprehend. All other reasons in
Table 2 are also difficult to interpret as constraints. Based on the implications of all the reasons in the context of transportation, it was decided that reason 10 (“No transportation available”) could be a variable by itself for further analysis even though only 56 respondents selected that reason. Another variable of interest could be one that combines reason 10, reason 3 (“Caretaking”), and reason 4 (“Disabled or home-bound”). A total of 219 respondents selected one of those three reasons. Although reason 1 (“Personally sick or quarantining”) and reason 12 (“Hospitalized or otherwise confined”) also reflect constraints, they were not included with the assumption that the constraints are temporary. Because these two reasons reflect constraints, but it is not known whether they reflect long-term or short-term ailments, they were excluded from analysis (similar to those who refused to respond or did not know the answer). Thus, the total number of observations excluded from the analysis of people making no trips on travel day for transportation and/or disability and caretaking reasons is 1667.
3.2. Methods
Three binary logit models were used to compare the 2022 NHTS respondents who took fewer-than-planned trips in 30 days with others. The first compared the respondents who took fewer trips for transportation reasons only (reasons 2 to 6 in
Table 1), the second compared the respondents who took fewer trips because of transportation or health reasons (reasons 2 to 7 in
Table 1), and the third compared all respondents who took fewer trips for any reason (reasons 1 to 10 in
Table 1). In all cases, the stated categories were coded 1, whereas others were coded 0, meaning that the models predicted the likelihood of a respondent belonging to the stated categories.
Three similar models were used to compare the respondents who did not take any trip on the travel day. The first compared the respondents who did not take any trip because of the unavailability of transportation (reason 10 in
Table 2), the second compared the respondents who did not take any trip because of transportation unavailability, caretaking, or disability (reason 3, 4, or 10 in
Table 2), and all respondents who did not take any trip (reasons 1 to 15 of
Table 2). Similar to the models on taking fewer trips, the three models predicted the likelihood of a respondent belonging to the stated categories.
Three categories of explanatory variables were included in the logit models: person-related, household-related, and geography-related. The inclusion of geographic explanatory variables was important because, on the one hand, they can inform transportation planners about which types of areas deserve more attention for transportation improvements, and on the other, they help to control for geographic variations when examining the relation between people’s characteristics and trip deprivation. Some geographic variables, such as metropolitan area designation and population size, as well as urban–rural distinction, are already included in the 2022 NHTS dataset. A variable for residence type (detached, apartment, etc.) also provides insight into the nature of the residential location of the respondents. However, unlike the 2017 NHTS, even Version 2.0 of the 2022 NHTS, released in March 2024, does not include any variable for population or job density at the census tract or census block group level. The author’s correspondence with an NHTS official revealed that it was still uncertain whether those variables would be integrated with the NHTS data in the future. Based on that understanding, population density at the block group level and job density at the census tract level were imputed from the 2017 NHTS data.
The explanatory variables of the logit models, presented in
Section 4.3, were selected on the basis of the literature review, especially the literature on transportation disadvantage and spatial mismatch as well as government policies. Based on the literature, laws, and regulations, it was hypothesized that the likelihood of being trip-deprived would be higher for the respondents belonging to populations that are generally considered to be transportation-disadvantaged: people with disabilities, people with low income, people belonging to minority races and ethnicities, people speaking non-English languages, unemployed people, older adults, and women. It was also hypothesized that people with higher travel need would be more likely to be trip-deprived. Similarly, single parents and people with children generally were expected to be more trip-deprived because of the need to transport children. People from single-person households were also expected to be more trip-deprived because of the need to single-handedly take care of all household activities. People in rural areas were expected to have a higher likelihood of being trip-deprived than people living in urban areas, but people living in apartments and in areas with high population density in the residential block group were also expected to have a higher likelihood of being trip-deprived because of the socioeconomic differences between cities and suburbs. A variable for rail transit availability in the region and residence in single-family detached home were included in the models to examine if they had any relationship with the predicted variables. A variable indicating whether a person also had forgone trips on the travel day was included to examine if there is a correspondence between people who forgo trips over a period of 30 days and just 1 day. A similar (i.e., positive or negative) significant association would indicate direct correspondence between the two.
The variables on population density and job density for the 2022 households were imputed by using geographic characteristics (including metropolitan area population size, rural–urban classification, availability of rail transit in the region, and census region), as well as personal and household characteristics (including household size, number of workers in household, household income, vehicles in household, and the race of the respondent). The method involved the integration of the 2017 data (with the density variables and the predictor variables) and the 2022 data (with the predictor variable but without the density variables) and subsequently predicting the density variables. Following recommendations in Heymans and Eekhout [
57], the imputed values were obtained by using the expectation maximization method instead of linear regression. With the recognition that the imputed density estimates are not exact, the predicted variables were converted to quartiles for inclusion in the models.
CFA was used to develop a latent variable representing travel need to be included in the logit models. Including a variable for travel need is important because people with higher need may have a greater likelihood of taking fewer-than-planned trips over a period of time. For example, a worker who needs to commute to work and a parent who needs to drop off children at school several days a week may have a higher travel need compared with others, and because of their higher need to make many trips, they may be more likely to forgo some trips. Although it seems intuitive to assume a positive association between travel need and number of trips one makes, need cannot be reasonably estimated from observed trips because some people with high need may not be able to travel more because of constraints. Furthermore, travel need is subjective, and hence abstract.
Although it is intuitive to assume that people with higher travel need are more likely to forgo trips over a period of time, it is difficult to hypothesize the same about a person forgoing trips on a particular day. The reason is that one can also hypothesize that the people with higher travel need could be more capable and organized (e.g., a worker must go to work, and a parent must drop off children at school). Thus, the effect of travel need on trip deprivation on a single day seems less predictable than it would be over a period of time.
In the CFA, to develop the variable travel need, only personal and household characteristics were used to generate the factor scores. Variables related to driving, drivers in household, the propensity or frequency of using different travel modes, the propensity of using online shopping, etc., were avoided because they themselves could be the outcome of travel need. Only six variables were included in the CFA: a variable representing the number of children in the household, a variable representing the number of adults in the household, a dummy variable indicating worker status, a dummy variable indicating retired status, a dummy variable indicating older age (age 65+), and a dummy variable representing people with high education (bachelor’s degree or higher). It was hypothesized that the number of children, the number of adults in the household, worker status, and high educational attainment would be positively associated with high travel need, whereas retired status and older age would be negatively associated with high travel need. The CFA was undertaken by the CALIS procedure in SAS 9.4.
Following the practice in a wide body of empirical research, the CFA results were validated based on recommendations by Gefen et al. (2000). The study recommended that Goodness-of-Fit Index (GFI), the adjusted GFI (AGFI), and the Normed Fit Index (NFI) be larger than 0.90 and Root Mean Square Residual (RMR) be smaller than 0.05. The study also recommended that all variables included in the CFA be statistically significant. Based on the study’s recommendation, the standardized coefficients from the model are presented.
Analytical methods also included the computation of variance inflation factors (VIFs) of the explanatory variables to examine multicollinearity. Although Hair Jr. et al. [
57] recommends the examination of multicollinearity especially when the number of observations in the dataset is small, because the latent variable travel need was created and the variables on population and job density were imputed by variables that were also expected to have a separate and independent effect on trip deprivation, multicollinearity among the explanatory variables was checked by computing the VIF of all variables. VIF is not a statistical test, but econometricians have recommended an upper limit. For example, Hair Jr. et al. [
58] Kennedy [
59] and suggest a value of 10, whereas Studenmund [
60] suggests a value of 5. The estimated VIF values are discussed in
Section 4.2.
5. Discussion
Although little could be known in the past about the American people who could not make trips because of transportation and other constraints, the recently released 2022 NHTS data made it possible to examine the characteristics of those people. By analyzing those data, this paper examined the characteristics of US adults who had forgone trips (a) within a 30-day period and (b) on the travel day of the survey. The results of the first set of analysis are presented in
Table 5, whereas the results of the second set are presented in
Table 6. The first two models in each table are more relevant for transportation planners and policy makers because they examine trip deprivation under constraints. The third model is less useful in both tables because many of the cited reasons do not necessarily reflect constraints.
The models predicting the likelihood of taking fewer-than-planned trips in 30 days, presented in
Table 5, are more meaningful than the models predicting people not taking trips on the travel day, presented in
Table 6. The models in
Table 6 are less meaningful because the question that generated the data was asked only to the people who did not travel on a single day. When people are restricted to only those who did not travel on a given day and then asked why they did not travel, the people who make non-discretionary trips (e.g., workers going to work and parents dropping off children at school) are more likely to be excluded from the question on reasons because they are more likely to make daily trips than others. However, when all people are asked about taking fewer-than-planned trips over a period of time—as in the question about forgoing trips in the past 30 days, the respondents are not similarly censored.
The NHTS question on the reasons for not traveling on the travel day is still useful because it allows for a comparison of the responses to the responses to the question on taking fewer-than-planned trips in 30 days. That is because a consistency between the responses of the two questions can add credence to the responses to the question about taking fewer trips over a period of time. For example, if people with disabilities are found to have a higher likelihood of forgoing trips in a 30-day period and also on the survey’s travel day, it becomes almost a certainty that people with disabilities forgo trips more than others.
The analyses in this paper demonstrated that the assumptions in the literature on transportation disadvantage and spatial mismatch, as well as the premises of the federal policies, are mostly valid. For example, the higher likelihood of trip deprivation for transportation reasons among people with disabilities, unemployed people, people speaking non-English languages at home, people with low income, and Black people—shown by the first model in
Table 5—confirms that these populations are truly transportation-disadvantaged. In the second model of the table, where health constraints were added to transportation constraints, all of these populations except Black people were again found to be trip-deprived. The negative association between car ownership and the likelihood of trip deprivation in the first two models of
Table 5 shows that the researchers who have attributed variations in car ownership to transportation disadvantage [
20,
64,
65,
66] have a valid argument. The first two models of
Table 5 also show that people from single-person households as well as people from households with children are more likely to be trip-deprived than others.
The populations that were expected to be trip-deprived because of transportation constraints but were not found to be trip-deprived are Hispanics, females, and older adults. The results on Hispanics are perhaps not very surprising because of diversity within the group. The results on females are mixed. The first model of
Table 5 shows females having a higher likelihood of being trip-deprived for transportation reasons, but only at the 10% level of significance. The third model of the same table, which predicts people taking fewer trips for any reason, shows a higher likelihood for females at the 5% level. The third model of
Table 6 shows similar results. These results seem to indicate that women are more likely to forgo trips than men, but the reason for that cannot be clearly attributed to transportation.
The variable older adults, defined as people aged 65+, was not significant in any of the six models in
Table 5 and
Table 6. The variable was not significant even when older adults were defined as people aged 75+ instead of 65+ in a separate model. However, these results do not necessarily refute the fact that many older adults are trip-deprived. A reason for the variable older adults not being significant is the overlap between people who have disabilities and people who are old. At least two US studies [
41,
42] showed that older adults are highly heterogeneous. While many older adults cannot travel like others because of disabilities and low income, others are fully capable of traveling when they wish. The results of this study indicate that public policy and future research on older adults’ transportation barriers should recognize this heterogeneity.
The results also show the importance of considering racial and ethnic minorities to be heterogenous. The variable representing Black people is positively associated with the dependent variable at the 5% level in the first and the third model of
Table 5, whereas the variables representing Asian people and Hispanic people are not significant in any of the three models. The variable representing mixed race and other races is significant with a negative sign in the first two models of
Table 5, indicating that they are less likely to be trip-deprived. This shows, once again, that considering all minorities together as transportation-disadvantaged ignores the fact that the Black population is more disadvantaged than other minorities in terms of trip deprivation.
A number of place-related variables are significant in the models in
Table 5. The variable representing high population density in the neighborhood has a significant and positive association with trip deprivation in all three models of the table, indicating that people living in neighborhoods with high density are more likely to be trip-deprived. That could be because neighborhoods with high population density are often inhabited by low-income, Black, and carless households. The positive relationship of the variable representing people living in apartments in the first model of
Table 5 is consistent with the results on population density because the high-density areas typically contain a larger share of apartments. These results show that transportation policy should pay significantly greater attention to central city areas with high population density.
Although one could be surprised to see that rural residents are not more likely to be trip-deprived than others, it is perhaps not surprising because the comparison category is people living in urban areas, where many people live without cars. Just because trip deprivation of rural residents is not higher, one should not conclude that they do not encounter other transportation barriers. One of the most obvious barriers for rural residents is the distance they have to travel between activity sites.
The variable representing rail transit in metropolitan areas is significantly and negatively associated with the dependent variables of all three models in
Table 5, indicating that the likelihood of being trip-deprived is lower in metropolitan areas served by rail. However, it is difficult to conclude from this association that rail transit has any effect on reducing trip deprivation because rail transit in US cities is commonly used by people with high income. It is perhaps more reasonable to conclude from the observed association that rail is more likely to be available in areas where people have lower trip deprivation because of affluence.
Although the primary objective of this paper is to identify populations that are more likely to be trip-deprived than others, certain theoretical and practical recommendations can be made for addressing trip deprivation and its effects. In the theoretical domain of transportation equity, researchers lately have been arguing in favor of operationalizing the Capability Approach (CA), a detailed review of which can be found in van Burgsteden et al. [
67]. Most studies seeking the implementation of the CA suggested both land use strategies and transportation strategies. Between the two types of strategies, the latter would be more effective to address trip deprivation because transportation improvements usually have a more direct and immediate effect on mobility than land use improvements. However, the CA also demands that people choose beneficial strategies themselves, meaning that strategies to address trip deprivation should be selected by a bottom-up approach. That would require substantial public engagement.
The initial strategies to address transportation disadvantage in this country evolved around public transportation. For example, the subsidization of transit trips by people with disabilities and older adults was one of the earliest strategies. However, despite making technological progress by providing real-time information and allowing online ticketing, transit agencies still offer services mostly by trains and fixed-route buses in high-density urban areas. Due to the decentralization of activities over the past decades, fixed-route transit cannot effectively serve trip-deprived individuals who live in transit-poor areas or those whose typical destinations are not served by buses and trains. Even within high-density areas, fixed-route transit does little to serve certain traditionally deprived populations like older adults [
41]. The provision of variable-route and door-to-door service with smaller vehicles in addition to fixed-route transit could address the issue somewhat. Other strategies could be a collaboration between transit agencies and ridehailing companies and means-tested subsidization of ridehailing trips. In recent years, researchers have been increasingly arguing in favor of greater car access for underserved populations [
64,
66]. App-based car rental practices could potentially be expanded.
The effects of trip deprivation could be addressed by recent improvements in information and communication technologies (ICTs) as well as app-based transportation technologies [
68]. Greater flexibility for workers to work from home could provide relief to those who have difficulty commuting. Technologies like telemedicine, online shopping, and videoconferencing have already become popular. Ridehailing has also acquired a reasonable share of the travel market. There is an expectation among researchers like Golant [
68] that autonomous vehicles (AVs) will provide another travel option in the future to people who cannot drive. However, modern technologies will be beneficial to people with low income only if their costs are affordable. Expanded subsidization of childcare could provide relief to parents with children who cannot make planned trips because of childcare responsibilities [
69]. Thus, in addition to transportation policies and strategies, modern technologies and government policies could help to address the effects of trip deprivation. That is not to suggest that trip deprivation and its effects could not be addressed by improving conventional fixed-route transit. Certain transit agencies are already undertaking projects like bus network redesign with consideration of the residential location of traditionally underserved populations. Those efforts could be improved through collaborative efforts between MPOs and transit agencies to identify trip-deprived populations. Access to stations/stops could also be improved.
It ought to be noted that the results of this research are pertinent to the US as a whole and may not be directly applicable to a specific metropolitan area. In addition to the national sample, the 2022 NHTS also collected data through oversampling for certain areas. Metropolitan planning organizations (MPOs) in regions for which such data were collected can conduct analyses similar to this research to learn about trip deprivation in their own regions. Other MPOs could include questions on trip deprivation in their own household travel surveys for similar analyses.
The NHTS does not include questions seeking recommendations for addressing trip deprivation. Such questions are not highly valuable in a national survey because the recommendations may vary from one region to another. However, MPOs could include such questions in their own household travel surveys and analyze the responses at the subregional level to determine what types of transportation improvement strategies may be applicable to which part of the region.