1. Introduction

Figure 1. Trends in food insecurity in US households, 2001-21 (Source: USDA, Economic Research Service using data from US Department of Commerce, Bureau of the Census, Current Population Survey Food Security Supplements [4]).

Figure 1. Trends in food insecurity in US households, 2001-21 (Source: USDA, Economic Research Service using data from US Department of Commerce, Bureau of the Census, Current Population Survey Food Security Supplements [4]).

Figure 2. Percentage distribution of SNAP participants and transit commuters across regions of the USA.

Figure 2. Percentage distribution of SNAP participants and transit commuters across regions of the USA.

2. Materials and Methods

2.1. Regression Analysis

In this research, food insecurity is the modeled output variable. Food insecurity, as the dependent variable, assumes a binary value with a dichotomous output. When its value is equal to 1, it indicates that food insecurity is 1; when it is equal to 0, it indicates food security. With this assumption, a logistic regression fits the best as the method to understand how various socioeconomic explanatory variables impact food insecurity. Research manuscripts reporting large datasets that are deposited in a publicly available database should specify where the data have been deposited and provide the relevant accession numbers. If the accession numbers have not yet been obtained at the time of submission, please state that they will be provided during review. They must be provided prior to publication.

A brief theoretical background of the logistic regression applicable in statistical modeling for dichotomous outputs with X being a vector of explanatory variables is provided below (source: [22]):

$$\log \left(\frac{\pi }{1-\pi }\right)=\beta X_i$$

(1)

where,

π is the probability of an event, and β are the estimated coefficients. The coefficient estimation is determined using the principle of maximum likelihood technique, with n observations:

$$L\left(\beta |X_i\right)={\displaystyle \prod _{i=1}^n{\pi }^{Y_i}{\left(1-\pi \right)}^{1-Y_i}}$$

(2)

Using natural logarithm on Eq. (2) gives a sum for the likelihood function as,

$$\log \left\{L\left(\beta |X_i\right)\right\}={\displaystyle \sum _{i=1}^n\left[Y_i\log \pi +\left(1-Y_i\right)\log \left(1-\pi \right)\right]}$$

(3)

Maximizing the log-likelihood equation in Eq. (3) by finding the first derivative and equating it to zero, will give the estimation of the coefficients, β. Various statistical software packages can be used to carry-out logistic regression for the theoretical information presented above. However, it must be noted that it is necessary to identify the socioeconomic variables as categorical or continuous variables before performing the regression.

Furthermore, we use the odds ratio to quantify the association between food insecurity and each of the individual variables. The odds ratio in the context of this research is defined as the ratio of the odds of an independent variable that causes food insecurity to the odds of the same variable not impacting food insecurity. Mathematically, the odds ratio is calculated based on information presented in the contingency Table 1 and the formula is:

(Odds ratio) = (X₂₂/ X₂₁)/(X₁₂/ X₁₁)

(4)

where,

X11 = number of times in the data the value of the independent variable is 0, with food security

X12 = number of times in the data the value of the independent variable is 0, with food insecurity

X21 = number of times in the data the value of the independent variable is 1, with food security

X22 = number of times in the data the value of the independent variable is 1, with food insecurity

The significance of the odds ratio is that if it is greater than 1, it ‘indicates a positive association between the independent variable and food insecurity. An odds ratio less than 1 indicates a negative association, while a ratio equal to 1 indicates no association between the independent variable and food insecurity.

Table 1. Contingency table.

Table 1. Contingency table.

		Food insecurity
		0	1
Independent Variable	0	X11	X12
	1	X21	X22

3. Application Example

3.1. Study Region and Data Collection

Data from the 2020 National Health Interview Survey (NHIS) collected by the National Center for Health Statistics (NCHS) of the USA are analyzed in this research. Although the NHIS data primarily focuses on health-related information of the noninstitutionalized population of the US, the survey, for the first time in the year 2020, includes information on those populations that reside near transit while inquiring about their food security status.

Based on the survey documentation, it has been pointed out that, in 2020, due to the COVID-19 pandemic, data collection procedures were disrupted. From April to June, all interviews were conducted by telephone only. From July to December, interviews were attempted by telephone first, with follow-ups to complete interviews by personal visit. In this way, 31,568 sample adult interviews were carried out [23]. The information from these many sample interviews is publicly available for researchers. Our primary focus was to extract only the relevant information on food (in)security, residing status of respondents if located near transit and their socioeconomic characteristics. There were 15,786 respondents who lived within walking distance (or close to a transit stop) at the time of the interview.

The 2020 National Health Interview Survey (NHIS) data on food insecurity and associated explanatory variables were at the spatial scale of counties spanning four regions of aggregated states, namely the Northeast, Midwest, South, and Western. Further, the counties within these aggregated states were classified based on the 2013 NCHS Urban–Rural Classification Scheme [24] . This classification scheme is based on the county population. It has six divisions: four metropolitans (large central metro, large fringe metro, medium metro, and small metro) and two nonmetropolitan (micropolitan and noncore).

A county that is a large central metro is defined as in metropolitan statistical area (MSA) of 1 million or more population that with the entire population of the largest principal city of the MSA. A large fringe metro has a population size between 1 million or more, which does not qualify as large central metro county. A medium metro county in MSAs has a population of 250,000 to 999,999, whereas a small metro county in MSAs has a population of less than 250,000. The nonmetropolitan categories of micropolitan and noncore are the respective counties in micropolitan statistical areas and the nonmetropolitan counties that did not qualify as micropolitan.

However, the survey combines the data from medium and small metro county respondents into one, and the same was done for the micropolitan and noncore nonmetropolitan counties. Thus, instead of six classification counties of the 2013 NCHS Urban-Rural Classification Scheme, the survey data has information on four levels of county classification: large central metro, large fringe metro, medium and small metro, and nonmetropolitan.

The association between the likelihood of facing food insecurity by the survey respondents residing close to transit and the socioeconomic variables of age, income, education level, and dependency on food stamps were estimated for significance using logistic regression. Note that the socioeconomic variables, as well as the food insecurity, were treated as dichotomous variables. Table 2 shows each variable definition used in the logistic regression. It is noted that the age of 38.6 years was identified as a classification level based on the 2020 median age of the US population. The purpose of keeping the median age as a classification level was to understand if there was any significance in food insecurity below this age among the respondents.

Further, the income levels were classified based on the poverty and median household incomes of $21,960 and $67,521, respectively. With an income that is below the poverty benchmark, it would be expected that food insecurity might be a concern among those living near transit. Like median age, median income was a natural choice as a classification level in understanding food insecurity.

Education as the explanatory variable was classified based on the respondents’ qualifications in holding a degree, i.e., if the respondent had a degree, was it an associate degree, a bachelor’s, or a higher educational degree. The purpose was to understand if food security was at all impacted among the respondents with their level of educational attainment.

The last variable tested for impact on food insecurity is the respondents’ utilizing food stamps in the previous twelve months. Availing food stamps or the SNAP by respondents would not mean food insecurity occurs. Since, SNAP is intended to eliminate food insecurity that stems from worrying if the food would not last, run out, or the respondent has to cut the size of meals or skip meals.

Table 2. Variables definition.

Table 2. Variables definition.

Variable	Notation	Coding
Male	Male	1 = Male, 0 otherwise
Age ≤ 38.6 (USA 2020 Median age: 38.6α)	Age_386	1 if Age ≤ 38.6, 0 otherwise
Age in-between 38.6 and 50	Age_38650	1 if 38.6 < Age ≤ 50, 0 otherwise
Age in-between 50-65	Age_5065	1 if 50 < Age ≤ 65, 0 otherwise
Age > 65	Age_65	1 if Age > 65, 0 otherwise
Income ≤ $21,960 (below poverty with a mean family size of 3β)	Inc_p	1 if Income ≤ $21,960, 0 otherwise
Income $21,960 to $67,521 (Median household income in the USA was $67,521 in 2020γ)	Inc_pm	1 if $21,960 < Income ≤ $67,521, 0 otherwise
Income > $67,521	Inc_m	1 if Income > $67,521, 0 otherwise
Education level with no degree	Edu_nd	1 if Education with no degree, 0 otherwise
Education level with associate degree and aboveθ but below bachelor level	Edu_ad	1 if Education with a degree below undergraduate, 0 otherwise
Education level with bachelors and above degree	Edu_bd	1 if Education with undergraduate and above degree, 0 otherwise
Received food stamp in the last 12 months	Food_S	1 if received food stamp in last 12 months, 0 otherwise

^αSource: Statistica.com, 2021 [25]; ^βSource: Statistica.com 2021 [27]; ^γUS Department of Health and Human Services, 2021 [26]; ^θAssociate degree in occupational, technical, vocational program, academic program etc.

A total of sixteen spatially distinct scenarios covering almost the entire US were analyzed. These scenarios were from four of the 2013 NCHS Urban–Rural Classification Scheme counties times the four aggregated states, Northeast, Midwest, South, and West.

Table 3 shows the descriptive statistics for the explanatory variables. The table was prepared for those respondents living close to transit.

A very high percentage of almost 94% respondents living near transit and in the large central metro counties of the Northeast states indicated food insecurity. The percentage of respondents living near transit and in the nonmetropolitan counties of the South states had the lowest (8%) food insecurity. Overall, for the respondents living near a transit and in the nonmetropolitan counties of the Midwest and South states, the percentages of explanatory variables were found to be lower than in the other two regions of the aggregated states. The percentage distributions of all the explanatory variables were the largest for the large central metro counties across all the four regions of the aggregated states.

The maps in Figs. 3-5 show the spatial variations in the percentages of the food insecurity and the thirteen explanatory variables. The map of food insecurity, the gender variables (males and females, age of respondents between 38.5 to 50 years and 50 years to 65 years, income greater than $67.5k and education level bachelors and above appear to have a similar color shade – although differing in their respective percentages.

Table 3. Descriptive statistics of the dependent and explanatory variables (shown as percentages)

Table 3. Descriptive statistics of the dependent and explanatory variables (shown as percentages)

Variable	Aggregated States
	Northeast States				Midwest States				Southern States				Western States
	Large central metro	Large fringe metro	Medium and small metro	Nonmetropolitan	Large central metro	Large fringe metro	Medium and small metro	Nonmetropolitan	Large central metro	Large fringe metro	Medium and small metro	Nonmetropolitan	Large central metro	Large fringe metro	Medium and small metro	Nonmetropolitan
Food Insecurity (consists of aggregated affirmative responses to - ‘worry food would run out’, ‘food didn't last’, ‘couldn't afford to eat balanced meals’, and ‘cut the size of meals or skip meals’)	94	73	59	36	83	52	57	24	70	53	41	8	89	78	76	45
Male	38	25	21	11	34	16	23	7	28	19	14	3	41	35	31	16
Female	45	28	22	16	39	19	24	8	32	21	18	4	42	40	34	16
Age ≤ 38.6 (USA 2020 Median age: 38.6)	25	15	11	5	26	10	18	3	23	13	9	2	28	20	20	9
Age in-between 38.6 and 50	14	10	7	5	13	6	8	2	10	7	5	1	16	14	12	6
Age in-between 50-65	21	14	12	9	18	9	11	4	14	10	8	1	20	18	16	8
Age ≥ 65	23	14	13	8	17	8	10	6	14	11	10	2	19	22	17	9
Income ≤ $21,960 (below poverty with a mean family size of 3)	17	5	6	6	9	4	10	4	11	4	7	2	9	6	8	5
Income $21,960 to $67,521 (Median household income in the USA was $67,521 in 2020)	29	17	19	10	28	13	20	7	22	15	14	3	27	24	27	13
Income ≥ $67,521	36	31	19	11	36	18	17	5	27	21	11	1	47	44	30	14
Education level with no degree	36	21	19	14	29	15	23	8	25	17	16	4	34	29	32	16
Education level with associate degree and above but below bachelor level	7	6	5	3	7	5	6	3	6	5	5	1	10	11	9	4
Education level with bachelors and above degree	39	25	19	10	37	15	17	5	29	19	11	2	40	34	23	11
Received food stamp in the last 12 months (2019)	15	9	11	19	10	9	14	14	12	9	15	24	8	7	12	17

Figure 3. Spatial distribution of percentages of explanatory variables – a. Food insecurity, b. Male, c. Female, d. Age ≤ 38.6, e. Age in-between 38.6 and 50, and f. Age in-between 50-65.

Figure 3. Spatial distribution of percentages of explanatory variables – a. Food insecurity, b. Male, c. Female, d. Age ≤ 38.6, e. Age in-between 38.6 and 50, and f. Age in-between 50-65.

Figure 4. Spatial distribution of percentages of explanatory variables – g. Age ≥ 65, h. Income ≤ $21,960, i. Income $21,960 to $67,521, j. Income ≥ $67,521, k. Education level with no degree, and l. Education level with associate degree and above but below bachelor level.

Figure 4. Spatial distribution of percentages of explanatory variables – g. Age ≥ 65, h. Income ≤ $21,960, i. Income $21,960 to $67,521, j. Income ≥ $67,521, k. Education level with no degree, and l. Education level with associate degree and above but below bachelor level.

Figure 5. Spatial distribution of percentages of explanatory variables – m. Education level with bachelors and above degree and n. Received food stamp in the last 12 months.

Figure 5. Spatial distribution of percentages of explanatory variables – m. Education level with bachelors and above degree and n. Received food stamp in the last 12 months.

Results and Discussion

Table 4. Coefficient estimates of explanatory variables (values in parenthesis are standard error).

Table 5. Odds ratio for the explanatory variables

Figure 6. Spatial distribution of significant impacts of the explanatory variables (a)-(i) on food insecurity.

Figure 6. Spatial distribution of significant impacts of the explanatory variables (a)-(i) on food insecurity.

Conclusion and Future Research

References

1. 2013 National Center for Health Statistics (NCHS) Urban–Rural Classification Scheme for Counties, https://www.cdc.gov/nchs/data/series/sr_02/sr02_166.pdf (accessed on 4 May 2022).
2. National Health Interview Survey (NHIS), 2020, https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2020/srvydesc-508.pdf (accessed on 1 May 2022).
3. Poverty income threshold of $21,960 for a family size of 3: https://www.statista.com/statistics/183657/average-size-of-a-family-in-the-us/.
4. US Department of Health and Human Services 2021, https://aspe.hhs.gov/topics/poverty-economic-mobility/poverty-guidelines/prior-hhs-poverty-guidelines-federal-register-references/2021-poverty-guidelines.
5. Statistica.com, https://www.statista.com/statistics/241494/median-age-of-the-us-population/ (accessed on 5 January 2021).
6. United States Department of Agriculture (USDA), Food and Nutrition Service, United States Department of Agriculture (USDA), https://www.fns.usda.gov/pd/snap-state-activity-reports (accessed on 22 April 2023).
7. Pruitt, S., Leonard, T., Xuan, L., Amory, R., Higashi, R., Nguyen, O., & Swales, S. (2016). Who is food insecure? Implications for targeted recruitment and outreach, National Health and Nutrition Examination Survey, 2005–2010. Retrieved June 6, 2018. [CrossRef]
8. Larson, J. ; Moseley, W. G. Reaching the limits: a geographic approach for understanding food insecurity and household hunger mitigation strategies in Minneapolis-Saint Paul, USA. GeoJournal, 2012; 77, 1-12. [CrossRef]
9. Feeding America, https://map.feedingamerica.org/?_ga=2.156318203.1397879697.1691855096-248042924.1691855096 (accessed on August 10, 2023).
10. Coleman-Jensen, A., Rabbitt, M. P., Gregory, C. A., & Singh, A. (2022). Household food security in the United States in 2021. Economic Research Report Number 309, US Department of Agriculture, accessed on July 23, 2023, https://www.ers.usda.gov/webdocs/publications/104656/err-309.pdf.
11. International Food Policy Research Institute (IFPRI). https://www.ifpri.org/topic/food-security (accessed on 25 April 2022).
12. HungerMap, World Food Programme (WFP), https://static.hungermapdata.org/insight-reports/2022-04-25/rbbsummary.pdf (accessed on 25 April 2022).
13. Prosekov, A. Y.; Ivanova, S. A. Food security: The challenge of the present. Geoforum, 2018; 91, 73-77. [CrossRef]
14. SNAP Eligibility, Food and Nutrition Service, United States Department of Agriculture, https://www.fns.usda.gov/snap/recipient/eligibility, (accessed on 21 May 2022).
15. Welch, T. F. Equity in transport: The distribution of transit access and connectivity among affordable housing units. Transport policy, 2013; 30, 283-293. [CrossRef]
16. Wang, K.; Woo, M. The relationship between transit rich neighborhoods and transit ridership: Evidence from the decentralization of poverty. Applied Geography, 2017; 86, 183-196. [CrossRef]
17. Liang, C.;Huang, Y.; Yip, T. L.; Li, V. J. . Does rail transit development gentrify neighborhoods? Evidence from Hong Kong. Transportation Research Part A: Policy and Practice, 2022; 155, 354-372. [CrossRef]
18. Derakhti, L.; Baeten, G. Contradictions of transit-oriented development in low-income neighborhoods: The case study of Rosengård in Malmö, Sweden. Urban Science, 2020; 4(2), 20.
19. Gupta, S., “How can Delhi provide affordable housing near transit to its poor”, DownToEarth, https://www.downtoearth.org.in/blog/governance/how-can-delhi-provide-affordable-housing-near-transit-to-its-poor-73619 (accessed on 11 May 2022).
20. Sharp, S. “Affordable Housing Rises Next to Blue Line in Long Beach”, Urbanize Los Angeles, 2018, “Affhttps://la.urbanize.city/post/affordable-housing-rises-next-blue-line-long-beach (accessed on 4 May 2022).
21. Burrows, M.; Burd, C.; McKenzie, B. Commuting by public transportation in the United States: 2019. American Community Survey Reports 2021, 802.
22. Baek, D., 2016. The effect of public transportation accessibility on food insecurity. Eastern Economic Journal, 2016; 42(1), 104-134.
23. Gundersen, C.; Kreider, B.; Pepper, J. The economics of food insecurity in the United States. Applied Economic Perspectives and Policy, 2011; 33(3), 281-303. [CrossRef]
24. Baek, D. (2014). Essays on Poverty and Infant Health. Louisiana State University and Agricultural & Mechanical College.
25. Martinez, J. C.; Clark, J. M.; Gudzune, K. A. Association of personal vehicle access with lifestyle habits and food insecurity among public housing residents. Preventive medicine reports, 2019; 13, 341-345. [CrossRef]
26. Swayne, M. R., & Lowery, B. C. (2021). Integrating transit data and travel time into food security analysis: A case study of San Diego, California. Applied Geography, 131, 102461. [CrossRef]
27. Spencer, M., Petteway, R., Bacetti, L., Barbot, O., 2011. Healthy Baltimore 2015, 1st ed. Baltimore City Health Department, Baltimore City. http://health.baltimorecity.gov/ sites/default/files/HealthyBaltimore2015_Final_Web.pdf, (accessed on 10 August 2023).
28. Patil, G. R. ; Sharma, G. Overweight/obesity relationship with travel patterns, socioeconomic characteristics, and built environment. Journal of Transport & Health, 2021; 22, 101240. [CrossRef]