Introduction
I analyze the interaction between food insecurity across demographics and understand their effect on GPA. We have a lot of data which sets the academic exercise different from others. In this article, I do some exploratory data analysis.
Around 2017, Jen, Rashida and Aydin conducted a comprehensive survey which asked a variety of questions related to food and housing insecurity. The survey also asked the students to answer personal demographic details along with GPA.
Using the survey data, we want to analyze how race, dream / daca status, ESL status, etc affect GPA. But along with the demographics, we also consider variables related to food and housing insecurity. Last, we also analyze whether programs like Cal Fresh help.
Note that I do not consider the CSU campus in the study. This is a drawback, and we may have to address the issue with alternative checks. I could not link the survey to the respondent CSU (Sorry Jen).
Exploratory Data Analysis
Design and sample
The data comprises undergraduate students at all the CSUs. We ignore reported GPA, which is either less than or equal to zero or higher than four. We include demographic information like gender, race, class standing, parents’ education, veteran status, marital status, first generation, DACA, Dreamer. After the filter, our data includes 14923 responses. As far as we know, our study has the biggest sample and hence we can identify correlations between academic performance and various demographic factors.
The main independent variable of interest is the level of food security. The students answer a few hunger related questions. Based on the responses, we divide the students into four groups: (1) high food security, (2) marginal food security, (3) low food security, or (4) “very low” food security.
| FoodSecurity | GPA_mean | count |
|---|---|---|
| Very Low | 3.08 | 3487 |
| Low | 3.15 | 3145 |
| Marginal | 3.22 | 3381 |
| High | 3.29 | 4910 |
Table 1 shows the average GPA of the respondents grouped by food security. The findings of the Table form the basis of the paper. Two observations are in order. First, consistent with prior literature, 23% of students have very low food security and 44% of students are food insecure (they have either very low or low food security). Second, there is a clear relationship between food security and GPA. As students get more food secure, their GPA increases.
Given the relationship between food security and GPA, we next identify the correlation between demographic factors and GPA.
Relationship between Race and Food Security
Before we talk about race, note that it is difficult to define race. For example, naturally, we classify a non-hispanic black student as African American. But, we also classify a hispanic black student as African American, unless the student reports their race as “Other”. But, consistent with the literature, we carve out a white Hispanic category. An appropriate classification of bi-racial students is difficult by definition. If someone was bi racial, then we do not classify them as bi-racial unless they are White / Hispanic.
| Race/FoodSecurity | Very Low | Low | Marginal | High | Total |
|---|---|---|---|---|---|
| African American | 39% (332) | 23% (196) | 17% (148) | 21% (178) | 100% (854) |
| Asian | 18% (457) | 19% (471) | 24% (596) | 40% (1010) | 100% (2534) |
| Hispanic | 25% (1189) | 24% (1154) | 23% (1102) | 27% (1284) | 100% (4729) |
| International | 22% (69) | 17% (55) | 28% (88) | 34% (108) | 100% (320) |
| Other | 27% (129) | 20% (96) | 19% (89) | 35% (167) | 100% (481) |
| Pacific Islander | 28% (108) | 22% (86) | 21% (82) | 29% (113) | 100% (389) |
| White | 21% (895) | 19% (827) | 23% (993) | 38% (1648) | 100% (4363) |
| White Hispanic | 25% (308) | 21% (260) | 23% (283) | 32% (402) | 100% (1253) |
| Total | 23% (3487) | 21% (3145) | 23% (3381) | 33% (4910) | 100% (14923) |
Figure 1: This figure compares the GPA across races
From Table 2, the racial disparities are clear. 39% of African American students have very low food security while 40% of White and Asian students have high food security.
Figure 1 shows a box plot of GPA grouped by race. Two results may be surprising. First, international students seem to have the highest GPA. Second, White, Hispanic and other students seem to have a long left tail — several students report a GPA lower than 1, which seems too low.
Relationship between Parental Education, Race and Food Security
Given the link between race and food security, it is important to understand the impact of parental education. Table 3 shows the relationship between parental education and food security. The results are not surprising. As parents get more education, the students seem to become more food secure. For example, 46% of respondents whose parents have a graduate degree have high food security while only 23% of respondents whose parents are high school dropouts have high food security.
| ParentalEducation/FoodSecurity | Very Low | Low | Marginal | High | Total |
|---|---|---|---|---|---|
| LessHS | 27% (492) | 27% (481) | 23% (414) | 23% (425) | 100% (1812) |
| HS | 24% (639) | 22% (586) | 24% (629) | 29% (763) | 100% (2617) |
| SomeColl | 26% (1020) | 21% (831) | 22% (870) | 31% (1208) | 100% (3929) |
| AssocDeg | 25% (516) | 22% (455) | 22% (437) | 30% (616) | 100% (2024) |
| BacDeg | 18% (539) | 19% (548) | 23% (673) | 40% (1160) | 100% (2920) |
| GradDeg | 17% (281) | 15% (244) | 22% (358) | 46% (738) | 100% (1621) |
| Total | 23% (3487) | 21% (3145) | 23% (3381) | 33% (4910) | 100% (14923) |
Figure 2: This figure compares the GPA across Parental Education grouped by Food Security
Figure 2 shows the GPA by parental education grouped by race. Again, the results with parental education hold. Across all races, as parents get more education, the student GPA increases — the pink horizontal line representing the median is above the red line.
Relationship between Daca / Dreamer Status, Race and Food Security
The left panel of Table 4 shows the relationship between Daca status, race, and food security. The left panel shows the results for Daca recipients. Out of the 260 Daca respondents, 200 are Hispanic and 40 are White Hispanic. For Daca recipients, food security is not an issue. For example, among Hispanic students, more Non-daca students are less food secure than Daca students.
|
|
|
|
Table 5 shows the relationship between Dream status, race and food security. The patterns are the same as the previous Daca table. The left panel shows the results for Dreamers. Out of the 263 Dreamers, 194 are Hispanic and 32 are White Hispanic. Again, For Dreamers, food security is not an issue. For example, among Hispanic students, more Non-daca students are less food secure than Daca students.
Relationship between Food security and other demographic variables
Figure 3 shows how GPA interacts with marital status for varying levels of food security. Married students have a higher GPA. The graph says that for marital status; we need to consider a binary variable: married or unmarried. We also checked to see if marital status and race had a differential impact on GPA. The results are consistent across races and we do not report the results for brevity.
Figure 3: This figure compares the GPA across marital status grouped by Food Security
Figure 4 shows how factors like transfer, EOP, ESL, and First Gen status affect the GPA. From Panel A, only transfer students with very low food security perform worse than their counterpart. In Panel B, the first gen status has the biggest interaction with GPA. Clearly, non-first-generation students have a higher GPA. EOP status and ESL status do not seem to affect GPA, at least via food security.
Figure 4: This figure compares the GPA across various demographics grouped by Food Security
|
|
Because of the impact of first gen, Table 6 further describes the students by race. An obvious pattern emerges. Across all races (except international), first gen students are more likely to have very low food security. Pratish did a quick literature search, and he did not find any articles concerning first gen students.
Figure 5 shows the impact of other demographic variables. Panels A and B show that Veteran Status and Active Duty status do not affect GPA across the four food security categories.
The results of Panels C and D are very interesting. Panel C shows students enrolled in the work-study program do not perform better than others. This result is surprising, as I had expected a positive result. Panel D shows that students experiencing homelessness have a lower GPA across all food security categories. The results are very pronounced for the bottom two levels of food security. Homelessness and food security are correlated.
Panel E shows that disabled students do not underperform their peers, which is also surprising. Last Panel F shows citizens do not have a higher GPA than non-citizens.
Figure 5: This figure compares the GPA across various demographics grouped by Food Security
|
|
Table 7 shows the interaction between homeless, food security and work study. It does not seem that the workstudy program is effective. More students who workstudy also suffer from very low food security and they are homeless. We need to think through this a bit more. Obviously, there is selection bias here but it is interesting to see if the workstudy program is well targeted.
Does food assistance help?
Given the food security issues, it is prudent to think whether students who receive food help perform better. We defined assistance as someone who has either used cal fresh, Ebit or CF assist programs. I count a student who is currently using it or has used it in the past as someone who has been assisted. A student who has heard of the program but has never used it is considered to have not been assisted.
Figure 6: This figure compares the GPA across students receiving food assistance grouped by Food Security
| FirstGen/AssistanceFlag | Assistance Not Used | Assistance Used | Total |
|---|---|---|---|
| FirstGen | 72% (4493) | 28% (1766) | 100% (6259) |
| NonFirstGen | 81% (6993) | 19% (1671) | 100% (8664) |
| Total | 77% (11486) | 23% (3437) | 100% (14923) |
Table 8 shows whether First-generation students receive assistance. Clearly, most first gen students do not get any assistance. We also checked whether the assistance variable correlated with race. From the data, more African Americans as a proportion received assistance than other races. In the same spirit, divorced students receive more assistance as a percentage than other groups. There was not much difference in marital status or transfer student status.
Regression Model that considers all the factors together
Table 9 brings all the demographic factors together. The Table gives fur columns representing different factors. Column 1 only considers food security as an independent factor. A coefficient of 0.204 for high food security means that students with high food security have a GPA that is 0.204 points higher than students with very low food security. That is, the results are relative to very low food security.
Column 2 considers race as an independent factor and the results are relative to GPA of African American students. White students have a GPA that is 0.271 points higher than their black counterparts. Column 3 shows that freshman have the highest GPA relative. Frankly this result was surprising as the result is counter to my own experience.
Column 4 considers all the models. The variables Workstudy, first gen and assistance are not significant any more. Being homeless decreases the GPA by 0.1 point. All other variables and their magnitudes are as expected.
| GPA | ||||
| (1) | (2) | (3) | (4) | |
| Constant | 3.084*** | 3.050*** | 3.234*** | 2.984*** |
| (0.008) | (0.016) | (0.010) | (0.023) | |
| FoodSecurityLow | 0.067*** | 0.053*** | ||
| (0.012) | (0.011) | |||
| FoodSecurityMarginal | 0.140*** | 0.107*** | ||
| (0.012) | (0.011) | |||
| FoodSecurityHigh | 0.204*** | 0.150*** | ||
| (0.011) | (0.011) | |||
| RaceAsian | 0.188*** | 0.151*** | ||
| (0.019) | (0.019) | |||
| RaceHispanic | 0.025 | 0.032* | ||
| (0.018) | (0.018) | |||
| RaceInternational | 0.288*** | 0.247*** | ||
| (0.031) | (0.031) | |||
| RaceOther | 0.188*** | 0.169*** | ||
| (0.027) | (0.026) | |||
| RacePacific Islander | 0.182*** | 0.157*** | ||
| (0.029) | (0.028) | |||
| RaceWhite | 0.271*** | 0.226*** | ||
| (0.018) | (0.018) | |||
| RaceWhite Hispanic | 0.128*** | 0.118*** | ||
| (0.021) | (0.021) | |||
| ClassStandingSophomore | -0.054*** | -0.043*** | ||
| (0.015) | (0.014) | |||
| ClassStandingJunior | -0.054*** | -0.042*** | ||
| (0.012) | (0.012) | |||
| ClassStandingSenior | -0.034*** | -0.034*** | ||
| (0.012) | (0.011) | |||
| GenderMale | -0.086*** | |||
| (0.009) | ||||
| ParentalEducationHS | 0.041*** | |||
| (0.014) | ||||
| ParentalEducationSomeColl | 0.022 | |||
| (0.014) | ||||
| ParentalEducationAssocDeg | 0.067*** | |||
| (0.016) | ||||
| ParentalEducationBacDeg | 0.127*** | |||
| (0.017) | ||||
| ParentalEducationGradDeg | 0.149*** | |||
| (0.019) | ||||
| HomelessYes | -0.100*** | |||
| (0.017) | ||||
| FirstGenNonFirstGen | -0.005 | |||
| (0.010) | ||||
| MarriedFlag | 0.060*** | |||
| (0.017) | ||||
| AssistanceFlagAssistance Used | -0.014 | |||
| (0.009) | ||||
| WorkStudyWorkstudy | 0.005 | |||
| (0.009) | ||||
| N | 14,923 | 14,923 | 14,923 | 14,923 |
| R2 | 0.027 | 0.048 | 0.002 | 0.087 |
| Adjusted R2 | 0.027 | 0.048 | 0.001 | 0.086 |
| Notes: | ***Significant at the 1 percent level. | |||
| **Significant at the 5 percent level. | ||||
| *Significant at the 10 percent level. | ||||