Our Students are Hungry: Food Insecurity Impacts More Than We Knew, Part 3 - Reaching Out to Help College Students: Evidence from a Randomized Assignment Protocol Using Email and Texting Systems to Inform Students of Various Campus Supports
This is the last part of a three-part series on my team’s research on students’ food insecurity.
In the first part, Dan Fitzpatrick and I wrote about the factors that correlate with increased food insecurity and how food insecurity links with first-year students’ non-cognitive attributes, first-semester performance, and fall-to-spring persistence. In part two, we engaged a structural equation model — a causal methodology — to understand the direct, indirect, and total influence of food insecurity on first-year students’ non-cognitive attributes, first-year performance, and fall-to-fall persistence. Feel free to revisit those pieces; however, in summary, we find that the impacts of food insecurity on students’ non-cognitive attributes, social adjustment, and performance and persistence are more complex than we previously understood. Both pieces ended with us calling for stakeholders to design interventions that include immediately gauging and addressing students’ food insecurity.
Here, we highlight our findings from a random assignment nudging protocol. This post will be rather quick, as the outcomes of random assignments either suggest the intervention worked or not. For students whom we identified as at high risk of not persisting past their first year at this regional, public institution , we generated a broad ‘nudging’ intervention to encourage participants to take advantage of a variety of campus resources; messages were timed for when they would be useful based on our knowledge of the academic calendar. Students were randomly assigned to three groups : one group got email messages, the other text messages, with a third control group. Like similar studies employing non-narrowed nudging, our treatment effects were null regarding performance and persistence measurements. Overall, it seems as if nudging is not a good mechanism to deal with multiple activities or global academic outcomes.
Yet, we had one significant (and unexpected) finding — in that our email treatment negatively influenced food insecurity, helping students become more food secure. The treatment coefficient is -.44, this magnitude is almost the same size as moving from half-way from one category of security to another; (1) Food Secure, (2) Marginal Security, (3) Low Security; (4) Very Low Security. Plainly said, as the higher number represents less security, the negative coefficient represents movement towards being more food secure .
Unfortunately, although our messaging encouraged students to visit the healthcare center, Invisible Need, and financial aid; because we do not have visitation data, we do not know how (via what behaviors) our email intervention affected food insecurity. We only observed that emailing students our messages produced an effect on the outcome.
So What? Broad nudging interventions are probably not useful tools in moving student performance and persistence. Yet, emailing students and encouraging them to seek out specific campus supports like going to a health-care center, visiting Invisible Need, visiting with peer mentors, or connecting with financial aid may be one way that stakeholders could ease students’ food insecurity. Furthermore, nudging more narrowly focused only on helping ease food insecurity may produce larger effects (and effects for students receiving text-messaging nudges).
Overall Conclusions and Implications
The conversation and understanding of college students’ food insecurity have become part of national academic and public discourse — see the Hope Center for some recent studies. Although higher education researchers and stakeholders have begun to pay closer attention to students’ basic needs, we still have a long way to go in understanding who comes to college hungry, how food insecurity impacts students’ experiences, and how to best serve students with unmet basic needs. Altogether, our studies add a bit more to the unfolding conversations.
In part one, we identified that multi-racial, international, transfer, LGBTQ, and first-generation student were linked with coming to college with increased food insecurity. Each of these findings makes sense given prior studies on the disadvantages these groups of students tend to face and given emergent research on college student food insecurity. We also found that food insecurity is correlated with students’ non-cognitive attributes (like psychological distress and amotivation) and first-semester performance. Therefore, a low-cost and easily implemented policy would be to deploy the 6-item USDA Food Security Scale and identify the exact students with the highest levels of food insecurity, to immediately provide targeted services to food-insecure students. The survey could be included with financial aid award packages, when students place their deposits, or during orientation. The earlier an institution can identify students’ food insecurity, the quicker campuses can intentionally promote visitations to campus supports, assist students with enrolling in SNAP, and seek additional assistance beyond campus.
In part two, we highlighted that the direct and indirect (and total) influence of food insecurity on students’ first-year experience, performance, and persistence is more complex than we previously understood. Including food insecurity into SEM models, we found a strong, indirect influence on first-year college GPA — which otherwise would not be detected in linear regressions. Our findings reinforce the need to specifically capture students’ food insecurity and adapt to food-insecure students’ needs. Interventions developed to address non-cognitive attributes or performance without easing food insecurity are unlikely to succeed.
Finally, in this last piece, we found that one possible way to ease students’ food insecurity is through nudging using campus emails. Therefore, a more narrowly focused intervention might be yet more effective.
1. 42% of students came to the regional public institution with food insecurity.
2. Selected groups of underserved students (like being first-generation, LGBTQ, multi-racial, and transfer students) correlate with higher food insecurity.
3. Food insecurity positively links with financial stress, psychological distress, and amotivation. Food insecurity negatively links with first-semester GPA and first-semester credits earned.
4. Structural Equation Modeling analysis suggests that food insecurity directly decreases social adjustment and indirectly decreases both college GPA and intention to persist.
5. Food insecurity should be addressed as a necessary starting condition for any interventions for first-year students aimed at bolstering non-cognitive attributes, social adjustment, performance, and persistence.
6. Nudging students to make use of existing but under-used resources via university email accounts may be helpful in easing students’ food insecurity.
 At this public, regional institution, students from schools with a high share of students eligible for free-and-reduced lunch (FRL) experience more academic challenges than their peers. The average FRL of incoming students is 24%; with every increase of 19% in high school FRL, students are at 13x higher odds of experiencing academic probation, 7x higher odds of dismissal, and less than one-fifth the odds of earning a degree (Authors, 2019). In the Fall 2018 semester, we identified all freshmen from high schools with at least 50% FRL (U.S. Department of Education, n.d.), giving us N=568 students, 13% of the incoming student population. External evaluators randomly assigned students into three groups, (1) Email communications (n=189), (2) Text communications (n=193), and (3) Control (n=186).
 Once these students from high-FRL schools were identified, an email was sent to their university accounts and text message via Remind to supplied phone numbers for an IRB approved “opt-out” communications. The communication informed students if they remained in the study an immediate gift-card and a future lottery-style was available. Zero students responded to the opt-out in the 3-day timeframe. After the sample was identified, we engaged external consultants at Abt for randomization. Randomized occurred on an individual level (not within groups) based upon gender, race, high school academic achievement measurement (combining GPA and standardized test scores), and high school free-and-reduced lunch percentage.
 Due to missing data, the analyses in the provided table (including on food insecurity) were generated using a chain imputation technique by our external consultants at Abt. We are comfortable with presenting the findings but want to bring attention to the cautious language we use to frame this finding and confirmatory testing is required.