Source: COLORADO STATE UNIVERSITY submitted to NRP
CAUSES AND CONSEQUENCES OF ADULT FOOD INSECURITY IN AT RISK POPULATIONS: PERSISTENCY, MORTALITY, AND SAMPLING
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1031006
Grant No.
2023-67023-40460
Cumulative Award Amt.
$481,673.00
Proposal No.
2022-10666
Multistate No.
(N/A)
Project Start Date
Aug 1, 2023
Project End Date
Jul 31, 2026
Grant Year
2023
Program Code
[A1641]- Agriculture Economics and Rural Communities: Markets and Trade
Recipient Organization
COLORADO STATE UNIVERSITY
(N/A)
FORT COLLINS,CO 80523
Performing Department
(N/A)
Non Technical Summary
Food insecurity (FI) is a significant concern for many US households. The negative health outcomes associated with FI include lower dietary quality, the worsening of both physical and mental health, and higher all-cause mortality, which are likely to disproportionately affect at risk populations. This project proposes three interrelated empirical analyses aiming at measuring: 1) the persistency of adult FI among minorities and underserved populations across multiple time periods and geographies; 2) mortality rates and life expectancy of individuals experiencing adult FI, especially at risk populations; and 3) the existence of a downward bias in the measure of elderly FI incidence. The first objective aims to assess the incidence (and determinants) of persistent food insecurity among U.S. minorities, using matching methods to create pseudo-longitudinal samples from publicly available Current Population Survey - Food Security Supplement (CPS-FSS) data. The second objective combines CPS-FSS data with publicly available Mortality Data of the National Center for Health Statistics, to estimate mortality rates for individuals experiencing FI, and Food Insecurity Adjusted Life Expectancy, with a focus on at-risk populations, and on the potential effectiveness of policies reducing FI. The third objective will document the extent to which the exclusion of institutionalized individuals from the CPS-FSS may lead to undermeasurament of FI in the elderly, by combining Health and Retirement Study data with CPS-FSS.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
70460103010100%
Knowledge Area
704 - Nutrition and Hunger in the Population;

Subject Of Investigation
6010 - Individuals;

Field Of Science
3010 - Economics;
Goals / Objectives
The goal of this research is to provide a more robust understanding of FI among at risk groups in the U.S. and to examine how FI may have long-term impacts on the health and well-being of households in such groups. To this end, we propose the following three objectives:Measure the incidence of Persistent Food Insecurity (PFI) among at-risk Americans and analyze its drivers.Estimate mortality rates for individuals experiencing Adult FI and calculate Food Insecurity Adjusted Life Expectancy. Evaluate the potential undermeasurement of elderly FI incidence in the CPS-FSS.
Project Methods
Objective 1.To measure the incidence ofPFIamong the U.S. population over a long time-period and across geographies, we will use the CPS-FSS data to create pseudo longitudinal datasets of households. This will require a two-steps process. First, we will isolate observations across the CPS-FSS years that are candidates to be part of a pseudo-longitudinal sample (PLS), by selecting individuals who have (potentially) aged throughout the different CPS survey years. The second step involves creating balanced panels of households over time from the samples. We plan to use Coarsened Exact Matching (CEM) to create consecutive matched samples of households balanced across observable characteristics (Stuart 2010). CEM matches households forcing the distribution of observed explanatory variables between the treated (in our case, household in yearyof agex) and the control group (household in survey yeary+1 agex+1) to be similar (Iacus et al. 2011). As a result, the distribution of unobservable characteristics that are related to the observable characteristics are forced into similar distributions between groups.The samples obtained after matchingwill be referred to as pseudo-longitudinal samples.For each samplewe will obtain different values ofPFITmeasured at T=(2,..,5), and present summary statistics of its incidence conditional on a series of demographics (e.g. ethnicity and race, gender of the household head, education etc.) also considering the years of surveys included. Then, treating the different PFITas binary dependent variables, we will estimate a model to assess what factors are associated with the probability of experiencing PFI.Objective 2.For objective 2 we will use the CPS-FSS and information on mortality of different population cohorts retrieved from publicly available sources. Detailed information on mortality by age for sub-populations characterized by different socio-demographic characteristics, including race, ethnicity, gender, education is available at the National level from the Mortality statistics of the National Center for Health Statistics (NCHS). The Mortality dataset will be used to obtain age-year-cohort specific mortality statistics (described below). Given that the NCHS restricts the public use of vital statistics to datasets at the national level, and publicly available micro-data files do not include state, county, or larger geographies, information on deaths by age, state of residence, gender, and urban status will be retrieved online from the NCHS, Mortality Data - Underlining Causes of Death Database (MD-UCDD).Because of confidentiality issues, age-year-cohort specific records including less than 10 cases are omitted.In those cases, we will retrieve age adjusted mortality data for 5-year age groups.We will construct a pseudo-panel(see Verbeek and Nikman, 1992; 1993).of households using the CPS-FSS where each observation represents a cohort of individuals with specific characteristics (e.g. gender, race, education level, etc.) at a given agex. Foreach age-cohort pair we will calculate mortality ratesfrom the MD-UCDD data. For each observation of the pseudo panel, we will calculate the incidence ofFI,and estimate and econometric model which coefficient will measure the association between FI rate and mortality rate.The estimated parameters also measuretheceteris paribuschange in mortality rate for those individuals in cohortz, at agex,that went from not experiencing FI to experiencing it. We will use its estimated value to calculate how experiencing FI can affect the Life Expectancy (LE) of individuals in each age-cohort combination to calculate a Food Insecurity Adjusted Life Expectancy (FIALE).Objective 3.CPS-FSS data will be usedwith data from the Health and Retirement Study (HRS) Core data. The HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and it is conducted by the University of Michigan. The HRS Core is a national longitudinal study of the economic, health, marital, and family status, as well as public and private support systems, of older Americans (aged 50 and above). The HRS core data are collected biannually. The first wave of the data dates to 1992.The HRS Core data asks respondents whether they (or their spouse, for those who are married) live in a nursing home or other Long-Term Care (LTC) facilities. For this objective we will limit the analysis to years of the CPS-FSS including information on home ownership collected during the month of December (when the FSS collection takes place) that is, 2016 and prior survey years, and those where the presence of adults with disabilities was recorded (available from 2008 onwards). Thus, the data used for this objective will focus on the 2008-2016 period. The analysis will take place in three steps. First, we will use the HRS core data (2008, 2010, 2012, 2014 and 2016 waves) andan econometric model to estimate the probability of HRS survey respondents will use a LTC facility conditional on a series of observable characteristics.The second step involves creating a dataset of older adult respondents (age 50 and above) in the CPS-FSS who are candidates to move to a LTC facility. To do that, we will adopt the same procedure used to construct PLSs in objective 1, restricted to the 2008-2016 CPS-FSS data.The observations discarded during the eight sequential CEMswill be retained as they will represent respondents that were in the survey in yearybut likely not to have been included in the year y+1. Pooling all the discarded observations across the 22 samples, we will obtain a dataset of "LTC candidates".The third and last step of the analysis consists in combining the estimated parameters obtained in step one using the HRS data with the CPS-FSS "LTC candidates" dataset to get predicted probabilities of residing in a LTC facility for each observation in the "LTC candidates" data.Once the predicted probabilities of LTCs are calculated, we will multiply the predicted LTC probabilities for FI individuals in the "LTC candidates" by one minus their FI adjusted mortality rate obtained from objective 2.We will use these "survival adjusted" probabilities of LTC to determinewhether a positive association exists between the predicted probability of LTC and FI among CPS-FSS respondents who are "LTC candidates". Any evidence of a positive association between LTC adjusted predicted probability and FI will indicate that excluding elderly LTC residents from the CPS-FSS is a contributing factor to the low elderly FI incidence reported in the CPS-FSS.

Progress 08/01/23 to 07/31/24

Outputs
Target Audience:Results of ongoing research under objective 1 were prepared tobe presentedat the 2024 AAEA meeting in New Orleans, LO. The target audience is that of applied economists, academics, researchers (for federal and state agencies) and graduate students. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Professional development has focused on the training od a graduate student. Beyond increasing the students' proficiency in matching methods and econometric modeling of food insecurity, the student has been tasked to lead manuscript writing and to present preliminary results to a scholarly audience. PI Bonanno has met weekly with the studentand provided the students with specific feedback on presentation effectiveness, framing of the narrative, storytelling and effective data visualization. How have the results been disseminated to communities of interest?Preliminary results of the analysis under objective 1 were presented at the AAEA meeting in New Orleans in July 2024. A working paper with some of the preliminary results was uploaded in AgEcon Search on May 2024. What do you plan to do during the next reporting period to accomplish the goals?In order to finalize the work on objective 1, we will assess the validity of our matching methods, by comparing prevalence of food insecurity and different spells of persistence food insecurity in the pseudo panels for the years 2015-2019, with data from the Panel Study of Income Dynamics (PSID) waves of 2015, 2017 and 2019. Once the method has been validated, we will continue with the data analysisand proceed to write an academic manuscript. Further, we plan to present results of the analysis under objective one to the online brownbag seminar series of the FDA in October 2024. The main focus of the year two activities will be on objective 2. We will continue the collection of detailed mortality rates data from the CDC - wonder platform and model the relationship between food insecurity prevalence and mortality rate. We plan to present preliminary results of this analysis at the 2025 AAEA meeting in Denver. As year two nears its completion we will begin the process of acquiring Health and Retirement Study to be ready and begin the work of objective 3 in year 3.

Impacts
What was accomplished under these goals? The Main focus of the activities for this reporting period has been on objective 1. Specifically, the research team has: Identifiedall households in the Current Population Survey - Food Security Supplements that appear for two consecutive years and evaluated the transition probability of being food secure (or food insecure) over a two- year period.including theCPS-FSS years 2001-2019. Across multipletwo-year periods, roughly80-85% ofhouseholds werefood secure for two consecutive years. of the 10-15% of households who experienced food insecurity at least once, 2/3 showed transitory food insecurity (that is, they were food insecure only for one of the two years),and the remaining one third experiencedfood insecurity inboth years, or they showed persistent food insecurity. For black and white/Hispanic households, the incidence of households being food secure for two years is much lower than for white non-Hispanic(about 70% for both black and white / Hispanic, 88% on average for white non-Hispanic). Similarly,the incidence of experiencing two-year food insecurity spells is more than twice as large for blacks (11.4%) and white-Hispanic (9.4%) than for white non-Hispanic (4.3%). Similar gradients are found with respect of education level of the household head, with the incidence of remaining food secure (food insecure) for two consecutive years increasing (decreasing) with education: from 66.5 (12.2%) for household heads with less than high-school degree, 80.9% (6.72%) those with a High school degree; 84% (5.6%) those with Some college or associate degrees, and 93.8% (1.85%) for bachelor degree or higher. Interestingly, even though we find intertemporal food insecurity incidence (both transitory and persistent) beinghigher during the years of the great recession, we find that, on average, a household experiencing food insecurity in a given year has about 50-50 chance to remain food insecure during the following year, across race / ethnic group, education levels, as well as gender of the household head, and the age cohort they belong to. We have made considerable progress in the creation of pseudo-longitudinal datasets, using CPS-FSS for the 2001-2019 years. To this end, we developed a multi-stage matching process that begins with exact matching of households from consecutive survey years, which do not belong to the subpopulation of returning households, followed by Coarsened Matching, Mahalanobis distance matching and propensity score (nearest neighbor) matching. The characteristics used in the matching process are chosen following the sampling procedure as documented by the BLS and the US Bureau of Census. Specifically, time invariant attributes used to determine the strata of the sample are gender, age, race and ethnicity of the household head, and state of residence. We also use education attainment of the household head which, for the overall range of ages considered (30-69 years old, can be assumed to be invariant). In order to account for 1) changes in sample size of the CPS-FSS over time, 2) differences in the number of households across age groups and across the different survey years, and 3) the Great Recession years, we obtain pseudo-longitudinal datasets for 5, 7, 9, 11, 15 and 19 years. We find that the likelihood of experiencing food insecurity for 5 or more consecutive years is on average very small, across different subsamples of the data. Also, subsampling the data by race/ethnicity, gender of the household head, and their educational attainment, the incidence of showing no instances of food insecurity over the entire period considered varies considerably across subsamples, with households with a white non-Hispanic, highly educated (college or higher) male household head being more than five times more likely to never experiencing food insecurity than households with a female, low educated (less than high school) , black household head. With respect to the other objectives of the grant, data collection for objective 2 has begun. Specifically state- and county-level mortality data by age group, race/ethnicity and gender is being collected for the years 1999-2016 from Center for Disease Control and Prevention - Wonder data platform (Data series: Compressed Mortality).

Publications

  • Type: Conference Papers and Presentations Status: Submitted Year Published: 2024 Citation: A Study of Prevalence and Determinants of Persistent Food Insecurity Using Repeated Two-Year Household Panels