Abstract
Background
We model the use of public transit to reach grocery stores and the use of online delivery services to get groceries, before and during the COVID-19 pandemic among people who used transit regularly prior to the crisis.
Methods
We draw upon a panel survey of pre-pandemic transit riders in Vancouver and Toronto. We conduct multivariable two-step tobit regression models that predict the likelihood of a respondent using transit as their primary mode for getting groceries before the pandemic (step 1) and then during the pandemic (step 2). Models are conducted for two survey waves, May 2020 and March 2021. We also conduct zero-inflated negative binomial regression models predicting the frequency respondents ordered groceries online.
Results
Transit riders over the age of 64 were more likely to use transit to reach groceries before the pandemic and more likely to continue to do so during the pandemic (wave 1, OR, 1.63; CI, 1.24–2.14; wave 2, OR, 1.35; CI, 1.03–1.76). Essential workers were more likely to continue using transit to reach groceries during the pandemic (wave 1, OR, 1.33; CI, 1.24–1.43; wave 2, OR, 1.18; CI, 1.06–1.32). Walking distance to the nearest grocery store was positively associated with using transit to get groceries pre-pandemic (wave 1, OR, 1.02; CI, 1.01–1.03; wave 2, OR, 1.02; CI, 1.01–1.03), and in May 2020 (wave 1, OR 1.01; (1.00–1.02). During the pandemic, people who stopped using transit to get groceries were less likely to have made zero online grocery purchases (wave 1, OR, 0.56; CI, 0.41–0.75; wave 2, OR, 0.62; CI, 0.41–0.94).
Discussion
People still physically commuting to work were more likely to still use transit to get groceries. Among transit riders, older adults and those living far walking distances from grocery stores are more likely to use transit to get groceries. Older transit riders and those with higher incomes were also more likely to use grocery delivery services, while female, Black, and immigrant riders were less likely to do so.
1. Introduction
Researchers have spent decades examining the potential role of food environments in contributing to food insecurity, diet, and nutrition. But the relationship between neighbourhood food environments and diets is mediated by a variety of factors, such as time availability, household income, and transportation resources (Widener, 2018). Lack of access to grocery stores among zero car households at the neighbourhood level, for example, is associated with higher obesity levels among the food-insecure (Christian, 2010). Further, where residents travel or commute to during the day shapes their food options and purchasing decisions (LeDoux and Vojnovic, 2013; Widener et al., 2013). Households on low incomes are more likely to rely on public transit to access food (Cannuscio et al., 2013; Shannon and Christian, 2017), making them less likely to reach food options outside of their neighbourhoods (LeDoux and Vojnovic, 2013). This is due in part to a large disparity in food access between the auto and public transit modes (Widener, 2017), as well as temporal variations in both transit service and grocery store hours (DeJohn et al., 2022; Widener et al., 2017). The COVID-19 pandemic worsened the transportation options for households that rely on public transit for groceries, putting this already constrained access at risk.
Policy responses to the COVID-19 pandemic altered transit service availability and discouraged travel by public transit for all but essential purposes. Transit agencies that depended on fare revenues cut services in response to ridership dropping below 85% of pre-pandemic levels early in the crisis (Liu et al., 2020). Changes in service levels between May 2019 and May 2020 were negatively correlated with neighbourhood incomes in Toronto and Montréal (DeWeese et al., 2020), meaning neighourhoods with lower average incomes lost more services than those with higher average incomes. Many transit riders switched to driving if they owned a car, and switched to walking and getting rides from others if they did not own a car (Loa et al., 2021; Palm et al., 2021). Trips to grocery stores declined during this period, though the decline was lower in geographies with greater densities of grocery retailers (Hamidi and Zandiatashbar, 2021).
The pandemic, and societal responses to it, increased food insecurity predominantly through increases in poverty levels and unemployment (Fang et al., 2022; Gundersen et al., 2021). However, several studies have noted that the loss of public transit as a safe and viable mode of travel during the pandemic may have contributed to increased food insecurity among households on low-incomes (Dubowitz et al., 2021; Niles et al., 2020). Households experiencing food insecurity will sometimes travel greater distances to purchase more affordable food, and this strategy was threatened by the pandemic's impact on transit (Kinsey et al., 2020). One survey of pre-pandemic transit riders in Toronto and Vancouver found that women and people with disabilities had greater difficulty getting groceries while avoiding transit (Palm et al., 2021). A similar survey in the USA found that women, Latinos, people with disabilities, and people in poverty were more likely to report difficulties getting groceries while avoiding transit (He et al., 2022). These groups are more likely to be transit-reliant, and so these results suggest that transit-reliant riders who avoided the mode during the crisis struggled to adapt. Notably, while transit riders struggled vis a vis transit, they remained less likely to adopt emerging online services compared to people who previously shopped using other modes (Shen et al., 2022).
Use of online grocery services grew rapidly during the pandemic as consumers sought to avoid exposure to the disease (Grashuis et al., 2020; Shamshiripour et al., 2020), including exposure from using public transit. In Canada, over 31% of respondents reported using these services at least once in the first six months of the pandemic for either grocery delivery or curbside pick-up (Music et al., 2022). A national cross-sectional survey in the USA found less than 10% of respondents are likely to be permanent adopters of grocery delivery, however, while another 59% never used these services at all (Wang et al., 2021). A cross-sectional survey in Portland, USA, found income to be the second most important predictor of increased use of delivery services during the pandemic, with higher income households being more likely to increase their use of these services (Figliozzi and Unnikrishnan, 2021a). Service costs likely explain differences in utilization by income, with 47% of Canadians telling pollsters that they did not plan on using online food delivery services due to their fees (Music et al., 2022). Given that transit-reliant grocery shoppers have lower-incomes (Dubowitz et al., 2021; Niles et al., 2020), it is unlikely that they benefited as much from e-delivery services.
Despite the importance of public transit in enabling residents on lower incomes to access groceries (Cannuscio et al., 2013; Shannon and Christian, 2017), no research to date directly examines how transit-reliant riders' use of transit for groceries changed during the pandemic and what factors are associated with those changes. Further, the extent to which transit-reliant riders adopted online delivery services is unresearched, despite the risk that fees may make delivery a prohibitively expensive alternative (Music et al., 2022), and despite evidence that people who used transit to get groceries pre-pandemic are less likely to have adopted online services (Shen et al., 2022). It is important to address these research gaps, as continued uncertainty over the financial viability of transit systems means that these residents' long-term ability to use to transit to get groceries is still at risk (CUTA-ACTU, 2021; Quiroz-Gutierrez, 2021). Major North American cities are already witnessing cuts in transit service due to slow ridership recovery (Balintec, 2023; Guse, 2022; Henriquez, 2023). Whether and how riders who relied on transit to get groceries were able to adapt during the pandemic can assist policymakers in devising strategies to ensure continued food access as transit systems retrench in response to a “new normal.” To provide these insights in light of the reviewed literature, we test the following hypotheses: A) that transit riders’ income and vehicle ownership are negatively associated with their likelihood of continuing to use transit to reach groceries, while their distance to the nearest grocery store is positively associated with this outcome; and B) that income and distance to the nearest store are positively associated with grocery e-delivery use while vehicle ownership is negatively associated with this outcome.
2. Methods
2.1. Study sample
This study reports on data using waves 1 and 2 of an online, web-based, panel survey (the Public Transit and COVID-19 Survey). Wave 1 recruitment took places through social media advertising and community list serve distributions in Toronto and Vancouver in May 2020 in English and Chinese (Zhang et al., 2020). Advertisements were viewed by over 484,352 Facebook users in Toronto and 242,816 in Vancouver, with 0.013% of views becoming clicks to the survey landing page. During wave 1 data collection, Ontario reported between 125 and 190 new cases of COVID-19 each day among a population around 14.5 million, while British Columbia averaged between 4 and 22 cases reported per day for a population of roughly 5 million (Province of British Columbia, 2020). To participate in the survey, a respondent needed to have used public transit at least once per week prior to the pandemic. Participants were offered an incentive to enter a raffle for one of 20 $50 gift cards. The survey content included questions on the impacts of COVID-19 on respondents’ use of transit, and difficulties reaching destinations during the crisis (withheld for peer review). Despite being collected through social media, the wave 1 sample is able to replicate known relationships between travel behaviour and the built environment as estimated using representative household travel survey data that is the basis of government travel demand modeling in one of the regions (Toronto), but with smaller and weaker coefficients (withheld for peer review).
Consenting participants who provided an email in wave 1 were recontacted via email in March 2021 for a second wave of data collection. The second wave was also developed in Qualtrics and contained 33 items on 5 screens. The content covered travel behaviour during the pandemic, anticipated post-pandemic travel behaviour, and attitudes towards alternatives to transit. The authors used “Prevent Ballot Stuffing,” a Qualtrics option that stops duplicate entries using cookies. A captcha was also required to prevent bot entries. Participants in wave 2 were entered into a raffle to win one of 10 $100 gift cards to a common grocery store chain. Both waves of the study were approved by the University of Toronto Research Ethics Board (#00039306). Both survey instruments are provided in the appendix.
2.2. Measures
2.2.1. Outcomes
The primary outcome is a binary measure of whether a respondent used public transit for grocery trips, where a 1 indicates that the respondent used public transit or paratransit as their primary mode for getting groceries and a 0 indicates they used another mode or did not travel for groceries. Our secondary outcome is a count of the number of times the respondent used online grocery delivery in the prior month, an integer variable that ranges from 0 to 15. Survey language used to define these variables can be found in the appendix.
2.2.2. Predictors
Predictors are derived from the food and transportation literatures. To account for travel behaviour differences in food access (Cannuscio et al., 2013; Shannon and Christian, 2017), predictors include after-tax household income (continuous, Canadian dollars) and vehicle access (categorical), denoted as either having access to a vehicle, owning a vehicle, or having neither. Income in wave 1 was measured as annual 2019 income, but in wave 2 was collected as monthly income from March 2021. Wave 2 models also included a variable on whether the respondent had purchased a car in between waves 1 and 2 (binary for yes). Access to a bicycle (binary) was also included as a control for access to that mode.
Gender (binary for male), race or ethnicity (categorical), immigration history (binary for immigrated within the last 5 years), self-reported physical disability (binary), self-reported dietary restrictions (binary), and self-reported disability status (binary) are included based on prior studies of transit use during COVID-19 (Palm et al., 2021).
As the pandemic greatly reduced individuals' mobility, and our sample contains frequent transit users who are less likely than the general population to own a vehicle, we operationalized spatial food access as the walking distance to the nearest grocery store or greengrocer. These amenities were identified on OpenStreetsMap using R (version 3.6.3). Walking distances were calculated from the respondent's postcode centroid to the nearest grocer using Google Maps and the R package Rgoogleway. Pre-pandemic usage of grocery delivery services was included in both sets of models with the categories: never, less than once a month, 1–2 times per month, and 3 or more times per month.
The pandemic also necessitated inclusion of other variables. To control for fear of COVID-19, we included level of agreement with the statement “I am at risk of severe COVID-19” (binary for agree or strongly agree). To account for dependence on transit, the wave 1 survey asked respondents to count the number of times they would board a transit vehicle in a typical week pre-pandemic (numeric). Additionally, we included whether the respondent identified as an essential worker (binary for yes).
2.2.3. Statistics
For models predicting use of transit to get groceries, the authors conducted a Tobit two-step selection model using the sampleSelection package in R to account for the fact that not everyone in the sample rode transit pre-pandemic. The first step model predicts using public transit to get groceries pre-pandemic (1 = yes). The second step in the model predicts continuing to use public transit to get groceries (1 = yes), conditional on the first model. The model is run twice, first on wave 1 survey data (May 2020) and then on wave 2 survey data (March 2021). The authors modeled the number of online grocery deliveries using zero-inflated negative binomial regression with the pscl package in R because the predictor is a count variable with excessive zeros i.e., about half of respondents received zero e-deliveries. IPAW weights were also constructed for the wave 2 grocery delivery models using the predictors and co-variates of the model to re-weight the data to align with the prior wave's characteristics.
Just 55% of participants in Wave 1 completed the Wave 2 survey. This high level of attrition elevated the risk of attrition bias in the study, which we corrected for using inverse probability attrition weighting (IPAW) (Cole and Hernan, 2008; Wooldridge, 2002). Weighting was conducted using the ipw package in R, using all predictors and the outcome variables to weight the wave 2 sample to account for attrition (i.e., weighting to align with the characteristics of the first sample wave). The final valid sample was 3317 for wave 1 and 1705 for wave 2.
Table 1 provides details on which variables from wave 1 were used to reweight wave 2 data, as well as the distribution of the weights developed for each model. Extreme values in IPAW weighting are a sign of mis-specified weights (Cole and Hernan, 2008), but this problem is not present in our data.
Table 1.
Inverse probability attrition weight variables and distributions.
| Transit for grocery models | Online delivery models | |
|---|---|---|
| Variables used to develop attrition weights |
Use of transit to get groceries in wave 1 and pre-pandemic, gender, age, ethnicity, immigration, disability, dietary need, self-assessed COVID risk, income, walk distance to groceries, bike access, car access, pre-COVID transit boarding frequency, and essential worker status in wave 1. |
Wave 1 online grocery delivery utilization, pre-pandemic grocery delivery utilization, use of transit to get groceries in wave 1 and pre-pandemic, gender, age, ethnicity, immigration, disability, dietary need, self-assessed COVID risk, income, walk distance to groceries, bike access, car access, pre-COVID transit boarding frequency, and essential worker status in wave 1. |
| Distribution of attrition weights | ||
| Min. | 0.717 | 0.543 |
| 25th | 0.858 | 0.837 |
| Median | 0.947 | 0.953 |
| 75th | 0.1085 | 1.111 |
| Max | 2.577 | 2.863 |
3. Results
3.1. Sample characteristics
Summary statistics are provided from both waves of the survey in Table 2 . The sample has a large overrepresentation of females (wave 1, 62% [2057 of 3317]; wave 2 60% [1083 of 1805]), however this to some extent reflects the overrepresentation of women on transit in the larger of the two cities, Toronto, as measured in a representative household travel survey from the region (withheld for peer review). The similarity of the demographics of both samples suggests that the high attrition between waves did not significantly alter the demographic composition of the second wave sample, explaining the stability of the attrition weights.
Table 2.
Characteristics of survey respondents.
| Variable | Wave 1 (n = 3317) | Wave 2 (n = 1705) |
|---|---|---|
| Categorical variables (%) | ||
| City | ||
| Toronto | 52% | 50% |
| Vancouver | 48% | 46% |
| Gender | ||
| Female | 66% | 65% |
| Male | 30% | 31% |
| Non-binary | 3% | 3% |
| Age | ||
| 18-29 | 35% | 35% |
| 30-39 | 27% | 27% |
| 40-49 | 15% | 15% |
| 50-64 | 16% | 16% |
| 65+ | 6% | 6% |
| Ethnicity or Race | ||
| White | 60% | 61% |
| Black | 4% | 3% |
| Indigenous | 4% | 4% |
| Other | 32% | 32% |
| Immigrated last 5 years | 12% | 12% |
| Has specific dietary needs | 8% | 8% |
| Has a disability | 4% | 4% |
| Can access a bicycle | 48% | 48% |
| Vehicle Access | ||
| No car access | 44% | 44% |
| Yes, can access | 24% | 24% |
| Yes, own a car | 32% | 32% |
| Essential worker | 23% | 22% |
| I am risk of severe COVID-19 = yes | 36% | 35% |
| Pre-pandemic grocery delivery usage | ||
| Never | 79% | 79% |
| 1–2 times a month | 6% | 6% |
| 3 or more times a month | 3% | 3% |
| Less than once a month |
11% |
11% |
| Numeric variables (means) | ||
| 2019 annual income ($10,000s CAD) | 6.81 | |
| March 2020 income ($ CAD) | 4938.63 | |
| Weekly Transit boardings pre-pandemic | 10.87 | 10.88 |
| Square root walk distance in metres to nearest grocer (and raw untransformed) |
24.01 (676.87) |
23.99 (676.16) |
| Dependent variables | ||
| Rode transit for groceries pre-pandemic | 27% | 27% |
| Continued to ride groceries pre-pandemic (as % of above number) | 11% (37%) | 9% (34%) |
| Number of grocery deliveries in last 30 days (mean) | 1.00 | 1.18 |
Just over a quarter of the sample reported using public transit as their primary mode of travel for getting groceries before the pandemic (wave 1, 27%, [896 out of 3317]; wave 2, 27%, [ 455 out of 1805]). In the wave 1 reference period of May 2020, 11% [336 out of 3317] of the sample used public transit to reach groceries, representing 37% [336 out of 896] of those who did so pre-pandemic. Zero respondents switched from another mode to public transit for groceries in Wave 1. In the second wave reference period of March 2021, 9% [156 out of 1805] used transit to get groceries, representing 34% [156 out of 455] continuing to use transit for groceries a year into the pandemic. As in the first wave, zero respondents switched from another mode to public transit for groceries in Wave 2.
Roughly four-fifths of respondents had never used an online grocery delivery service before the pandemic (wave 1, 79% [2620 out of 3317]; wave 2, 76% [1372 out of 1805]). Just over one in ten used the services less than monthly (wave 1, 11% [365 of 3317]; wave 2, 11% [199 out of 1805). During the pandemic, however, respondents in the May 2020 wave used online grocery deliveries an average of 1.00 (95% CI, 0.91–1.09) time in the prior 30 days, while respondents participating in the follow up wave in March 2021 used online grocery deliveries an average of 1.18 (95% CI, 1.08–1.29) times in the preceding 30 days. The authors conducted a two-sided t-test of wave 2 respondent's answers to this question from both waves, and found a statistically significant increase in online grocery utilization, with a mean difference of 0.18 (95% C.I., 0.04 to 0.32).
3.2. Transit use to reach groceries before and during the pandemic
Odds ratios generated from the coefficients of the two-step Heckman models on use of transit to reach groceries are presented in Table 3 . The first step selection models predict a respondents' likelihood of using public transit to reach groceries pre-pandemic. Conditional on the first step, the second step outcome models predict a respondents’ likelihood of using public transit to reach groceries during the pandemic. For wave 1, the second step model predicts transit use for groceries in May 2020, while wave 2 predicts transit use for groceries in March 2021.
Table 3.
Odds ratios from two-step Heckman models predicting usage of public transit to get groceries.
| Wave 1 (May 2020) |
Wave 2 (March 2021) |
|||
|---|---|---|---|---|
| Selection model |
Outcome model |
Selection model |
Outcome model |
|
| (Rode for groceries pre-pandemic) | (Rode for groceries May 2020) | (Rode for groceries pre-pandemic) | (Rode for groceries March 2021) | |
| Intercept | 0.35 (0.28, 0.44)*** | 0.63 (0.23, 1.73) | 0.36 (0.26, 0.49)*** | 2.43 (1.02, 5.78)* |
| City is Vancouver (ref: Toronto) | 0.89 (0.78, 1.02)^ | 0.96 (0.87, 1.06) | 0.83 (0.69, 1.00)^ | 1.03 (0.90, 1.18) |
| Gender (ref: male) | ||||
| Female | 1.06 (0.94, 1.19) | 0.95 (0.88, 1.03) | 1.01 (0.85, 1.2) | 0.92 (0.82, 1.03) |
| Non-binary | 1.21 (0.99, 1.47)^ | 1.01 (0.87, 1.17) | 1.06 (0.80, 1.41) | 0.93 (0.76, 1.13) |
| Age (ref: 18–29) | ||||
| 30–39 | 1.01 (0.88, 1.16) | 1.02 (0.93, 1.12) | 0.97 (0.80, 1.17) | 1.15 (1.01, 1.31)* |
| 40–49 | 1.21 (1.03, 1.43)* | 1.17 (1.02, 1.34)* | 1.14 (0.90, 1.45) | 1.28 (1.09, 1.49)** |
| 50–64 | 1.41 (1.20, 1.66)*** | 1.32 (1.10, 1.58)** | 1.3 (1.02, 1.65)* | 1.31 (1.10, 1.57)** |
| 65+ | 1.72 (1.37, 2.17)*** | 1.63 (1.24, 2.14)*** | 1.74 (1.25, 2.41)*** | 1.35 (1.03, 1.76)* |
| Race or Ethnicity (ref: white) | ||||
| Black | 1.19 (0.91, 1.55) | 1.20 (0.99, 1.45)^ | 1.16 (0.80, 1.69) | 1.17 (0.89, 1.53) |
| Indigenous | 1.34 (1.05, 1.71)* | 1.19 (0.98, 1.46)^ | 1.23 (0.86, 1.76) | 1.06 (0.83, 1.36) |
| Person of colour | 1.37 (1.21, 1.54)*** | 1.16 (0.98, 1.37)^ | 1.24 (1.04, 1.48)* | 0.97 (0.83, 1.12) |
| Immigrated within 5 years of 2020 | 1.20 (1.02, 1.41)* | 1.09 (0.97, 1.24) | 1.27 (1.01, 1.61)* | 0.91 (0.76, 1.09) |
| Dietary restrictions (1 = yes) | 1.06 (0.87, 1.28) | 0.98 (0.86, 1.11) | 1.17 (0.89, 1.53) | 0.91 (0.75, 1.10) |
| Has a disability (1 = yes) | 1.15 (0.90, 1.48) | 1.00 (0.85, 1.18) | 1.33 (0.93, 1.90) | 0.82 (0.64, 1.05) |
| Income | ||||
| 2019 annual | 0.95 (0.94, 0.96)*** | 0.97 (0.95, 1.00)* | 0.96 (0.94, 0.98)*** | |
| March 2021 monthly | 1.00 (1.00, 1.00) | |||
| Walk distance to nearest grocery (sqrt of metres) | 1.02 (1.01, 1.03)*** | 1.01 (1.00, 1.02)* | 1.02 (1.01, 1.03)*** | 0.99 (0.98, 1.00) |
| Access to a bike | 0.81 (0.72, 0.90)*** | 0.89 (0.79, 1.01)^ | 0.74 (0.63, 0.86)*** | 1.15 (0.97, 1.37) |
| Vehicle Access (ref: none) | ||||
| Has access, doesn't own | 0.74 (0.65, 0.84)*** | 0.74 (0.63, 0.87)*** | 0.72 (0.60, 0.86)*** | 1.11 (0.93, 1.33) |
| Owns a vehicle | 0.36 (0.31, 0.42)*** | 0.54 (0.32, 0.92)* | 0.33 (0.26, 0.42)*** | 1.33 (0.77, 2.29) |
| Bought a car between waves 1 & 2 (yes = 1) | 0.79 (0.67, 0.93)** | |||
| Weekly transit boardings pre-pandemic | 1.03 (1.02, 1.04)*** | 1.01 (1.00, 1.03) | 1.03 (1.02, 1.04)*** | 1.00 (0.98, 1.01) |
| Pre-pandemic grocery delivery utilization (ref: Never) | ||||
| 3 or more times a month | 0.91 (0.67, 1.23) | 0.93 (0.59, 1.44) | ||
| 1-2 times a month | 1.10 (0.88, 1.37) | 1.09 (0.78, 1.51) | ||
| Less than once a month | 0.91 (0.67, 1.23) | 0.93 (0.59, 1.44) | ||
| Number of grocery deliveries in last 30 day | 0.97 (0.96, 0.99)*** | 0.97 (0.95, 0.99)*** | ||
| Agree that “I am at risk of severe COVID-19" (wave 1) | 0.99 (0.93, 1.05) | 1.03 (0.93, 1.13) | ||
| Essential worker (wave 1) | 1.31 (1.22, 1.40)*** | 1.16 (1.04, 1.30)** | ||
| R-squared | 0.2044 | 0.2032 | ||
***p < .0001, **p < .01, *p < .05, ^ p < .10.
The availability of alternative modes of travel is strongly and negatively associated with using transit to reach groceries before the pandemic. This includes having access to a vehicle (wave 1, OR, 0.74; 95% CI, 0.65–0.84; wave 2, OR, 0.72; 95% CI, 0.60–0.86), owning a vehicle (wave 1, OR, 0.36; 95% CI, 0.31–0.42; wave 2, OR, 0.33; CI, 0.26–0.42), and having access to a bicycle (wave 1, OR, 0.81; CI, 0.72–0.90; wave 2, OR, 0.74; CI, 0.63–0.86). Additionally, walking distance to the nearest grocery store or greengrocer, a proxy for ease of walking for groceries, is significantly and positively associated with using transit to reach groceries pre-pandemic (wave 1, OR, 1.02; CI, 1.01–1.03; wave 2, OR, 1.02; CI, 1.01–1.03). The typical number of weekly transit boardings pre-pandemic, a measure of reliance on transit, is also positively associated with using transit for groceries in that period (wave 1, OR, 1.03; CI, 1.02–1.04; wave 2, OR, 1.03; CI, 1.02–1.04). Few of these factors remain significant in second step models of transit use for groceries during the pandemic. In wave 1 these included having access to a vehicle (wave 1, OR, 0.76; CI, 0.64–0.90), owning a vehicle (wave 1, OR, 0.60; CI, 0.34–1.05), buying a vehicle between waves (wave 2, OR, 0.79; CI, 0.67–0.93), and walking distance to nearest grocery (wave 1, OR, 1.02; CI, 1.00–1.02) while in wave 2, only the indicator of having purchased a vehicle between survey waves was significant (wave 2, OR, 0.79; CI, 0.67–0.94).
Demographic factors associated with a greater likelihood of using public transit to get groceries pre-pandemic included being ages 50–64 (wave 1, OR, 1.41; CI, 1.20–1.66; wave 2, OR, 1.30; CI, 1.02–1.65), being over age 64 (wave 1, OR, 1.72; CI, 1.37–2.17; wave 2, OR, 1.74; CI, 1.25–2.41), being Indigenous (wave 1, OR, 1.34; CI, 1.05–1.72), being a non-Black and non-Indigenous identified racialized person (wave 1, OR, 1.37; CI, 1.21–1.54; wave 2, OR, 1.24; CI, 1.04–1.48), having immigrated within 5 years of the first survey wave (wave 1, OR, 1.20; CI, 1.02–1.41; wave 2, OR, 1.27; CI, 1.01–1.61), and 2019 annual income (wave 1, OR, 0.95; CI, 0.94,-0.96; wave 2, OR, 0.96; CI, 0.94–0.98). Only two of these factors are also significant in the second step models predicting use of transit for groceries during the pandemic, being age 50–64 (wave 1, OR, 1.32; CI, 1.10–1.58; wave 2, OR, 1.31; CI, 1.10–1.57) or age 65 and up (wave 1, OR, 1.63; CI, 1.24–2.14; wave 2, OR, 1.35; CI, 1.03–1.76). Additionally, being aged 40 to 49 is significant and positively associated with transit use during the pandemic in both waves (wave 1, OR, 1.17; CI, 1.02–1.34; wave 2, OR, 1.28; CI, 1.09–1.49). Finally, being an essential worker was significantly and positively associated with using transit to get groceries in the second step of the models (wave 1, OR, 1.31; CI, 1.22–1.40; wave 2, OR, 1.16; CI, 1.04–1.30). Income was negatively associated with continuing to use transit for groceries in May 2020 (wave 1, OR, 0.97; CI, 0.95–1.00) but not in wave 2.
Notably, pre-pandemic use of online grocery delivery services is not significantly associated with using transit to get groceries in the period. In contrast, the number of times a respondent used a grocery delivery service in the prior 30 days is significantly and negatively associated with using transit to get groceries during the pandemic (wave 1, OR, 0.97; CI, 0.96–0.99; wave 2, OR, 0.97; CI, 0.95–0.99). This suggests that individuals were substituting transit trips with delivery services. The next section extends these findings by detailing which transit riders used delivery services more intensively.
3.3. Use of online delivery services during the pandemic
Zero-inflated negative binomial regression models predicting the number of times a respondent ordered groceries online in the prior month are presented in Table 4 . Zero-inflation was necessary due to a high prevalence of zeros in the dataset. These models have two components: logistic regression models predicting the likelihood that a respondent reported zero online grocery purchases, and negative-binomial count models predicting the number of such purchases conditional on the zero-inflated results. Negative binomial models can be interpreted using incident risk ratio (IRRs) which the authors extracted by exponentiating the model coefficients. These are presented in Table 4, which also presents the odds ratios (ORs) for the zero-inflation models. Goodness of fit is measured through the Akaike Information Criterion (AIC), which suggests the model on wave 2 data fits the data better than the wave 1 model.
Table 4.
Predictors of transit riders’ use of online grocery delivery services.
| Wave 1 (May 2020) |
Wave 2 (March 2021) |
|||
|---|---|---|---|---|
| Zero-inflation model (ORs) | Count model (IRRs) | Zero-inflation model (ORs) | Count model (IRRs) | |
| Intercept | 3.65 (2.27, 5.88)*** | 1.83 (1.39, 2.40)*** | 1.13 (0.60, 2.16) | 1.61 (1.13, 2.29)** |
| City is Vancouver (ref: Toronto) | 1.33 (1.01, 1.74)* | 0.92 (0.79, 1.07) | 1.61 (1.11, 2.32)* | 0.79 (0.64, 0.96)* |
| Gender (ref: male) | ||||
| Female | 0.78 (0.60, 1.00)* | 0.98 (0.85, 1.13) | 0.67 (0.48, 0.93)* | 0.90 (0.74, 1.09) |
| Non-binary | 0.60 (0.39, 0.91)* | 0.86 (0.68, 1.09) | 0.86 (0.47, 1.58) | 0.90 (0.63, 1.28) |
| Age (ref: 18–29) | ||||
| 30–39 | 0.76 (0.57, 1.01)^ | 1.04 (0.89, 1.22) | 0.67 (0.45, 0.98)* | 1.07 (0.87, 1.32) |
| 40–49 | 0.75 (0.53, 1.06) | 1.08 (0.90, 1.30) | 0.67 (0.42, 1.05)^ | 1.27 (0.99, 1.62)^ |
| 50–64 | 1.42 (1.00, 2.02)^ | 1.39 (1.13, 1.71)** | 1.12 (0.71, 1.76) | 1.33 (1.02, 1.74)* |
| 65+ | 1.46 (0.89, 2.40) | 1.37 (1.02, 1.84)* | 1.36 (0.73, 2.53) | 1.42 (0.97, 2.07)^ |
| Race or Ethnicity (ref: white) | ||||
| Black | 2.31 (1.18, 4.49)* | 1.33 (0.96, 1.85)^ | 0.99 (0.39, 2.54) | 0.57 (0.35, 0.90)* |
| Indigenous | 1.29 (0.73, 2.27) | 1.26 (0.92, 1.73) | 0.86 (0.44, 1.70) | 1.23 (0.82, 1.84) |
| Other person of colour | 1.38 (1.07, 1.79)* | 1.26 (1.1, 1.45)*** | 1.06 (0.75, 1.49) | 1.07 (0.88, 1.29) |
| Immigrated within 5 years of 2020 | 0.87 (0.59, 1.27) | 1.00 (0.83, 1.22) | 0.46 (0.26, 0.81)** | 0.93 (0.71, 1.21) |
| Dietary restrictions (1 = yes) | 0.52 (0.34, 0.82)** | 0.80 (0.64, 1.00)* | 0.57 (0.33, 0.99)* | 0.97 (0.72, 1.31) |
| Has a disability (1 = yes) | 0.74 (0.41, 1.33) | 0.98 (0.75, 1.28) | 1.82 (0.84, 3.96) | 1.58 (1.07, 2.32)* |
| Agree “I am at severe risk of COVID-19” (wave 1) | 0.72 (0.57, 0.92)** | 1.09 (0.96, 1.23) | 1.07 (0.78, 1.48) | 1.23 (1.03, 1.47)* |
| Income | ||||
| Annual Income 2019 | 0.96 (0.93, 0.99)** | 1.02 (1.01, 1.04)** | ||
| Monthly Income March 2021 | 1.03 (0.98, 1.09) | 1.03 (1.00, 1.06)^ | ||
| Walk distance to nearest grocery (sqrt of metres) | 1.00 (0.98, 1.01) | 1.00 (0.99, 1.00) | 1.00 (0.99, 1.01) | 1.00 (1.00, 1.01) |
| Access to a bike | 1.11 (0.88, 1.4) | 0.99 (0.87, 1.12) | 1.39 (1.03, 1.89)* | 1.04 (0.88, 1.23) |
| Vehicle Access (ref: none) | ||||
| Has access, doesn't own | 1.07 (0.80, 1.45) | 1.02 (0.88, 1.20) | 0.72 (0.48, 1.08) | 0.89 (0.72, 1.11) |
| Owns a vehicle | 1.01 (0.75, 1.36) | 1.00 (0.85, 1.18) | 0.93 (0.64, 1.36) | 1.00 (0.81, 1.23) |
| Weekly transit boardings pre-pandemic | 1.02 (1.00, 1.03)* | 1.01 (1.00, 1.02)* | 1.04 (1.02, 1.06)*** | 1.02 (1, 1.03)** |
| Essential worker (wave 1) | 1.93 (1.43, 2.6)*** | 0.83 (0.69, 0.99)* | 1.68 (1.17, 2.40)** | 1.14 (0.92, 1.41) |
| Pre-pandemic frequency of grocery delivery (reference: never) | ||||
| 1–2 times a month | 0.01 (0.00, 0.15)** | 1.00 (0.85, 1.19) | 0.03 (0.00, 0.18)*** | 1.11 (0.88, 1.4) |
| 3 or more times a month | 0.03 (0.01, 0.08)*** | 2.01 (1.65, 2.44)*** | 0.06 (0.01, 0.27)*** | 1.27 (0.94, 1.72) |
| Less than once a month | 0.05 (0.03, 0.11)*** | 0.87 (0.74, 1.01)^ | 0.26 (0.15, 0.44)*** | 1.01 (0.81, 1.25) |
| Transit use for groceries (ref: did not use before or during pandemic) | ||||
| Kept riding transit during pandemic | 1.73 (1.08, 2.76)* | 1.06 (0.79, 1.40) | 0.83 (0.47, 1.46) | 0.83 (0.6, 1.15) |
| Stopped riding transit during pandemic | 0.56 (0.41, 0.75)*** | 1.16 (1.00, 1.36)^ | 0.62 (0.41, 0.94)* | 1.20 (0.97, 1.47)^ |
| Log(theta) | 2.28 (1.8, 2.87)*** | 2.20 (1.59, 3.04)*** | ||
| AIC | 7129.7 | 4331.5 | ||
***p < .0001, **p < .01, *p < .05, ^ p < .10.
Shifts in transit usage for grocery shopping are strongly associated with use of online grocery deliveries. People who stopped using transit to get groceries during the pandemic were significantly less likely to have ordered zero online grocery deliveries compared to those who never used transit to get groceries (wave 1, OR, 0.56; CI, 0.41–0.75; wave 2, OR, 0.62; CI, 0.41–0.94). These former riders also made more online grocery orders than this reference group conditional on the zero-inflated model (wave 1, IRR, 1.16; CI, 1.00–1.36; wave 2, IRR, 1.20; CI, 0.97–1.47). In contrast, people who continued to use transit to reach groceries were more likely to have made zero grocery deliveries in May 2020 (wave 1, OR, 1.73; CI, 1.08–2.76), but results for this group were otherwise insignificant.
Access to alternative modes is not significantly associated with grocery delivery rates in either survey wave, while demographic factors and transit use factors are. Factors that are positively associated with not having ordered zero online grocery deliveries include being an essential worker (wave 1, OR, 1.93; CI, 1.43–2.6); wave 2, OR, 1.68; CI, 1.17–2.40), living in Vancouver (wave 1, OR, 1.33; CI, 1.01–1.74; wave 2, OR, 1.61; CI, 1.11–2.32), and the weekly number of transit boardings pre-pandemic (wave 1, OR, 1.02; CI, 1.00–1.03; wave 2, OR, 1.04; CI, 1.02–1.06). Demographic factors that are negatively associated with zero online grocery purchases include identifying as female (wave 1, OR, 0.78; CI, 0.60–1.00; wave 2, OR, 0.67; CI, 0.48–0.93), and having dietary restrictions (wave 1, OR, 0.52; CI, 0.34–0.82; wave 2, OR, 0.57; CI, 0.33–0.99). Conditional on these results, the count models offer fewer significant demographic associations. Significant positive associates include being ages 50–64 (wave 1, IRR, 1.39; CI, 1.13–1.71; wave 2, IRR, 1.33; CI, 1.02–1.74), over age 65 (wave 1, IRR, 1.37; CI, 1.02–1.84; wave 2, IRR, 1.42; CI, 0.97–2.07), and the number of pre-COVID transit boardings (wave 1, IRR, 1.01; CI, 1.00–1.02; wave 2, IRR, 1.02; CI, 1.00–1.03). Pre-pandemic income is negatively associated with ordering zero grocery deliveries (wave 1, OR, 0.72; CI, 0.57–0.92) and positively associated with the number of grocery deliveries ordered (wave 1, IRR, 1.02; CI, 1.01–1.04). However, March 2021 monthly income was not significantly associated with grocery delivery counts in the second survey wave.
Finally, pre-pandemic usage of online grocery delivery services is negatively associated with not using the services during the pandemic, including using the service less than once a month pre-pandemic (wave 1, OR, 0.05; CI, 0.03–0.11; wave 2, OR, 0.26; CI, 0.15–0.44), 1–2 times a month pre-pandemic (wave 1, OR, 0.01; CI, 0.00–0.15; wave 2, OR, 0.03; CI, 0.00–0.18), and three or more times a month pre-pandemic (wave 1, OR, 0.03; CI, 0.01–0.08; wave 2, OR, 0.06; CI, 0.01–0.27). Conditional on the zero-inflated models, only one of these variables is significantly and positively associated with counts of online grocery deliveries during the pandemic: people who used the services three or more times pre-pandemic (wave 1, IRR, 2.01; CI, 1.65–2.44).
4. Discussion
Our survey offers three overarching sets of findings. First, riders without vehicles and those with lower incomes were more likely to use transit to get groceries before and during the pandemic, confirming hypothesis A. Second, the impact of significant socio-demographic predictors on use of transit to get groceries weakened in the second year of the pandemic. Third, transit-reliant riders were less likely to substitute public transit trips to grocery stores with online deliveries, while other riders were more likely to do so. We confirmed that income correlated positively with use of grocery e-delivery as stated in hypothesis B, but we did not find a significant association for distance to the nearest grocery store.
Regarding our first set of findings, regular transit users over the age of 50 were significantly more likely to use the mode to get groceries. A rider's income was negatively associated with their likelihood of using transit to get groceries, in line with prior research (Cannuscio et al., 2013; Dubowitz et al., 2021; Shannon and Christian, 2017) and theory (Niles et al., 2020). Pre-pandemic frequency of transit use is positively associated with using the mode to get groceries, suggesting that riders who depend on transit generally were also more likely to use it to get groceries pre-pandemic. Many other demographic factors were insignificant, including gender, city, disability status, self-assessed risk of a severe case of COVID-19, and dietary restrictions.
Transit riders who owned or had access to vehicles were significantly less likely to use transit to get groceries pre-pandemic. Those who bought a car during the pandemic were also less likely to use transit to get groceries. The challenges of walking long distances with groceries (Palm et al., 2021), may explain why longer walking distances to the nearest grocery store or greengrocer were positively associated with a respondent using transit to get groceries pre pandemic and May 2020, holding constant the availability of other transportation resources. Our results suggest that contexts inducing riders to use transit to get groceries include lacking a vehicle and living an inconvenient walking distance from the nearest store, especially for transit riders over the age of 50. Our second step results on transit use to get groceries during the pandemic add to existing evidence that daily commute tours and travel-activity patterns influence consumers’ food-related travel (LeDoux and Vojnovic, 2013; Widener et al., 2013). Within our sample, those who still had to physically commute to work (essential workers) were significantly more likely to still use transit to get groceries. This theme carries over to our grocery delivery models, wherein essential workers were significantly less likely to have used online grocery delivery at all in the preceding month.
Regarding our second overarching finding, demographic and transportation variables proved less reliable in predicting use of transit to get groceries in 2021 versus 2020. Vehicle ownership, walking distance to the nearest grocer, bicycle ownership, and income significantly predicted continued transit ridership to groceries in May 2020, but not March 2021. Being over the age of 65 remained a significant and positive predictor of continuing to use transit to get groceries, but the magnitude of the coefficient shrunk. These shifts may reflect riders’ continued adaptation to the crisis, especially given the significant increase in use of online delivery services between survey waves.
Concerning our third overarching finding, online grocery delivery became a substitute for transit use among less transit-reliant riders. Those who had been using transit to get groceries pre-pandemic, but who reporting switching modes during the crisis, were more likely to order online groceries compared to those who never used transit for groceries at all. Further, each additional online grocery delivery in the past 30 days was negatively associated with a respondents' odds of continuing to use transit in both 2020 and 2021. Distance to the nearest grocery store also had no relationship to use of e-delivery services, suggesting that spatial barriers to groceries were not a factor in e-delivery uptake. People without alternatives to transit, as well as those who had higher levels of pre-pandemic transit use, did not adopt grocery e-delivery as much as those with alternatives. We thus recommend against policymakers promoting private, unsubsidized e-delivery programs in transit-reliant communities until further research can shed light on why these riders did not adopt this alternative. Instead, our results point to the need to assess programs that blend food security programs with transit, such as LA Metro's on-demand food delivery service (LA Metro, 2020).
We also add to emerging evidence that income was positively associated with utilization of these services during the pandemic (Figliozzi and Unnikrishnan, 2021a). Our results suggest riders over 50 used online grocery delivery services more frequently than people ages 18–30, aligning with other recent work on the topic within the pandemic context suggesting that concerns about health and safety have reversed the negative impact of age on use of these services (Figliozzi and Unnikrishnan, 2021b). We also find evidence that riders with dietary restrictions were more likely to try these services, though they did not have higher counts of e-deliveries conditional on having used the service at all. Finally, Black and other persons of colour were significantly less likely to have tried these services in May 2020, though this association did not hold significance in wave 2. Relatedly, immigrants were less likely to have used e-delivery, but this association was only significant in wave 2.
Research examining the role of public transit in getting groceries is uncommon, with studies on the topic looking cross-sectionally at the entire population, and not specifically transit users themselves. This study suggests that within the transit-riding population, riders over the age of 50, those without vehicles, and those with longer walking distances to the nearest grocery store were more likely to use the mode to get their groceries before the pandemic. As countries begin to exit the crisis, our results demonstrate the importance of transit planners considering grocery access when planning transit systems in lower density, suburban areas, particularly those with lower-than-average incomes and vehicle ownership rates. We recommend that when agencies make service planning decisions, planners preserve transit access to grocery stores in neighborhoods that are not within walkable distances to grocery stores, and that contain higher-than-average proportions of zero car households.
4.1. Limitations
Several limitations should be considered when engaging with our results. The data were self-reported, and baseline (pre-pandemic) data were retrospective by a couple of months. Further, our large convenience sample cannot be used to estimate the extent of these dynamics in the general population. Our sampling approach was necessary due to the need to rapidly gather data without endangering researcher or subject health in the pandemic's early months.
5. Conclusions
We conducted two waves of surveys of people who used transit regularly pre-pandemic in two major Canadian cities. Our results paint a nuanced picture of reliance on transit to reach groceries before and during the COVID-19 pandemic. Among this sample, age and distance to the nearest grocery store were positively associated with using transit to get groceries, while income and vehicle access were negatively associated with this choice. Conditional on these results, age and being an essential worker were positively associated with using transit to get groceries during the pandemic in both survey waves. Essential workers used online grocery delivery services less frequently than other riders. Riders who switched to a non-transit mode to get groceries during the pandemic were more likely to use an online delivery service than those who never used transit for groceries pre-pandemic.
Financial disclosure
The Authors did not receive any specific funding for this work.
Declaration of competing interest
The authors have no competing interests to declare.
Acknowledgements
We would like to thank TTCRiders, and Shelagh Pizey-Allen in particular, for inspiring this project. We would like to acknowledge Jeff Allen and Bochu Liu for support in appending the data with additional built form covariates and reviewing the survey instrument. We also acknowledge Yixue Zhang for providing translations of the survey instrument into Mandarin and Cantonese. This work was supported by the University of Toronto School of Cities and the Canada Research Chair in Transportation and Health. The lead author is supported by funding from the Social Science and Humanities Research Council's partnership grant: Mobilizing Justice: Towards evidence-based transportation equity policy.
Footnotes
Supplementary data related to this article can be found at https://doi.org/10.1016/j.jth.2023.101623.
Appendix A. Supplementary data
The following is the supplementary data related to this article:
References
- Balintec V. Canadian Broadcast Corporation; 2023. Higher Fares for Less Service? A Breakdown of what the TTC's Budget Means for Riders. [Google Scholar]
- Cannuscio C.C., Tappe K., Hillier A., Buttenheim A., Karpyn A., Glanz K. Urban food environments and residents' shopping behaviors. Am. J. Prev. Med. 2013;45:606–614. doi: 10.1016/j.amepre.2013.06.021. [DOI] [PubMed] [Google Scholar]
- Christian T.J. Grocery store access and the food insecurity–obesity paradox. J. Hunger Environ. Nutr. 2010;5:360–369. [Google Scholar]
- Cole S.R., Hernan M.A. Constructing inverse probability weights for marginal structural models. Am. J. Epidemiol. 2008;168:656–664. doi: 10.1093/aje/kwn164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- CUTA-ACTU . vol. 53. Canadian Urban Transit Association; Toronto, Ontario: 2021. (Why Public Transit Needs Extended Operating Support (Issue Paper No.). [Google Scholar]
- DeJohn A.D., Widener M.J., Shannon J. Transit access to subsidized food stores in the US midwest. Prof. Geogr. 2022;1–14 [Google Scholar]
- DeWeese J., Hawa L., Demyk H., Davey Z., Belikow A., El-geneidy A. 2020. A Tale of 40 Cities: A Preliminary Analysis of Equity Impacts of COVID-19 Service Adjustments across North America. Findings. [DOI] [Google Scholar]
- Dubowitz T., Dastidar M.G., Troxel W.M., Beckman R., Nugroho A., Siddiqi S., Cantor J., Baird M., Richardson A.S., Hunter G.P., Mendoza-Graf A., Collins R.L. Food insecurity in a low-income, predominantly african American cohort following the COVID-19 pandemic. Am. J. Publ. Health. 2021;111:494–497. doi: 10.2105/AJPH.2020.306041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fang D., Thomsen M.R., Nayga R.M., Yang W. Food insecurity during the COVID-19 pandemic: evidence from a survey of low-income Americans. Food Secur. 2022;14:165–183. doi: 10.1007/s12571-021-01189-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Figliozzi M., Unnikrishnan A. Home-deliveries before-during COVID-19 Lockdown: Accessibility, environmental justice, equity, and policy implications. Transport. Res. D. 2021 doi: 10.1016/j.trd.2021.102760. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Figliozzi M., Unnikrishnan A. Exploring the impact of socio-demographic characteristics, health concerns, and product type on home delivery rates and expenditures during a strict COVID-19 lockdown period: a case study from Portland. OR. Transport. Res. A: Pol. Pract. 2021;153:1–19. doi: 10.1016/j.tra.2021.08.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Grashuis J., Skevas T., Segovia M.S. Grocery shopping preferences during the COVID-19 pandemic. Sustainability. 2020;12:5369. [Google Scholar]
- Gundersen C., Hake M., Dewey A., Engelhard E. Food insecurity during COVID‐19. Appl. Econ. Perspect. Pol. 2021;43:153–161. doi: 10.1002/aepp.13100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Guse C. 2022. MTA Quietly Cuts Bus Service across NYC as ‘money Saving Initiative,’ Manager Says in Email — EXCLUSIVE. New York Daily News. [Google Scholar]
- Hamidi S., Zandiatashbar A. Compact development and adherence to stay-at-home order during the COVID-19 pandemic: a longitudinal investigation in the United States. Landsc. Urban Plann. 2021;205 doi: 10.1016/j.landurbplan.2020.103952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- He Q., Rowangould D., Karner A., Palm M., LaRue S. Covid-19 pandemic impacts on essential transit riders: findings from a U.S. Survey. Transport. Res. Part D. 2022;105 doi: 10.31235/osf.io/3km9y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Henriquez G. The wait for a bus in Montreal just got longer. Global News. 2023 https://globalnews.ca/news/9398261/bus-wait-time-montreal-january-2023/ [Google Scholar]
- Kinsey E.W., Kinsey D., Rundle A.G. COVID-19 and food insecurity: an uneven patchwork of responses. J. Urban Health. 2020;97:332–335. doi: 10.1007/s11524-020-00455-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- LeDoux T.F., Vojnovic I. Going outside the neighborhood: the shopping patterns and adaptations of disadvantaged consumers living in the lower eastside neighborhoods of Detroit, Michigan. Health Place. 2013;19:1–14. doi: 10.1016/j.healthplace.2012.09.010. [DOI] [PubMed] [Google Scholar]
- Liu L., Miller H.J., Scheff J. The impacts of COVID-19 pandemic on public transit demand in the United States. PLoS One. 2020;15 doi: 10.1371/journal.pone.0242476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Loa P., Hossain S., Mashrur Sk.Md., Liu Y., Wang K., Ong F., Habib K.N. Exploring the impacts of the COVID-19 pandemic on modality profiles for non-mandatory trips in the Greater Toronto Area. Transport Pol. 2021;110:71–85. doi: 10.1016/j.tranpol.2021.05.028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- LA Metro Metro's Mobility on Demand Pilot Makes Emergency Food Deliveries. 2020. https://thesource.metro.net/2020/06/15/metros-mobility-on-demand-pilot-makes-emergency-food-deliveries/ URL.
- Music J., Charlebois S., Toole V., Large C. Telecommuting and food E-commerce: socially sustainable practices during the COVID-19 pandemic in Canada. Transp. Res. Interdiscip. Perspect. 2022;13 doi: 10.1016/j.trip.2021.100513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Niles M.T., Bertmann F., Belarmino E.H., Wentworth T., Biehl E., Neff R. The early food insecurity impacts of COVID-19. Nutrients. 2020;12:2096. doi: 10.3390/nu12072096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Palm M., Allen J., Liu B., Zhang Y., Widener M., Farber S. Riders who avoided public transit during COVID-19: personal burdens and implications for social equity. J. Am. Plann. Assoc. 2021 doi: 10.1080/01944363.2021.1886974. [DOI] [Google Scholar]
- Province of British Columbia British Columbia COVID-19 Dashboard. 2020. https://experience.arcgis.com/experience/a6f23959a8b14bfa989e3cda29297ded URL.
- Quiroz-Gutierrez M. 2021. It Could Take Nearly a Decade for Public Transit to Return to Pre-pandemic Levels. Fortune. [Google Scholar]
- Shamshiripour A., Rahimi E., Shabanpour R., Mohammadian A., Kouros) How is COVID-19 reshaping activity-travel behavior? Evidence from a comprehensive survey in Chicago. Transp. Res. Interdiscip. Perspect. 2020;7 doi: 10.1016/j.trip.2020.100216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shannon J., Christian W.J. What is the relationship between food shopping and daily mobility? A relational approach to analysis of food access. Geojournal. 2017;82:769–785. doi: 10.1007/s10708-016-9716-0. [DOI] [Google Scholar]
- Shen H., Namdarpour F., Lin J. Investigation of online grocery shopping and delivery preference before, during, and after COVID-19. Transp. Res. Interdiscip. Perspect. 2022;14 doi: 10.1016/j.trip.2022.100580. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang X.C., Kim W., Holguín-Veras J., Schmid J. Adoption of delivery services in light of the COVID pandemic: who and how long? Transport. Res. Pol. Pract. 2021;154:270–286. doi: 10.1016/j.tra.2021.10.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Widener M.J. Comparing measures of accessibility to urban supermarkets for transit and auto users. Prof. Geogr. 2017;69:362–371. doi: 10.1080/00330124.2016.1237293. [DOI] [Google Scholar]
- Widener M.J. Spatial access to food: retiring the food desert metaphor. Physiol. Behav. 2018;193:257–260. doi: 10.1016/j.physbeh.2018.02.032. [DOI] [PubMed] [Google Scholar]
- Widener M.J., Farber S., Neutens T., Horner M.W. Using urban commuting data to calculate a spatiotemporal accessibility measure for food environment studies. Health Place. 2013;21:1–9. doi: 10.1016/j.healthplace.2013.01.004. [DOI] [PubMed] [Google Scholar]
- Widener M.J., Minaker L., Farber S., Allen J., Vitali B., Coleman P.C., Cook B. How do changes in the daily food and transportation environments affect grocery store accessibility? Appl. Geogr. 2017;83:46–62. doi: 10.1016/j.apgeog.2017.03.018. [DOI] [Google Scholar]
- Wooldridge J.M. Inverse probability weighted M-estimators for sample selection, attrition, and stratification. Portuguese Econ. J. 2002;1:117–139. [Google Scholar]
- Zhang Y., Palm M., Scheff J., Farber S., Widener M. 2020. Travel Survey Recruitment through Facebook and Transit App: Lessons from COVID-19. Transport Findings December. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.