Source: PENNSYLVANIA STATE UNIVERSITY submitted to
ASSESSING HOUSEHOLD-LEVEL FOOD LOSS: AN EFFICIENCY ANALYSIS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1012396
Grant No.
2017-67030-26611
Cumulative Award Amt.
$100,000.00
Proposal No.
2016-11585
Multistate No.
(N/A)
Project Start Date
May 15, 2017
Project End Date
May 14, 2020
Grant Year
2017
Program Code
[A1801]- Exploratory: Exploratory Research
Project Director
Jaenicke, E. C.
Recipient Organization
PENNSYLVANIA STATE UNIVERSITY
408 Old Main
UNIVERSITY PARK,PA 16802-1505
Performing Department
Ag Econ, Sociology, and Educ
Non Technical Summary
The USDA's Economic Research Service, its collaborators, and other researchers have published rigorous research on the massive economic consequences from food loss in the farm-to-table supply chain, yet these efforts have been limited to indirect approximations for specific food categories. Virtually none of this important research quantitatively addresses the role that individual households play in determining the amount of food loss. The goal of the proposed food-loss exploratory research project will depart from existing methodologies for estimating food loss by food categories, and instead estimate household-specific food loss, framed in percentage terms, for a nationally representative sample of U.S. households.For each household in our dataset (USDA-ERS's new FoodAPS dataset), we propose to (i) classify all at-home and away-from-home food purchases into a range of food groups and/or food types; (ii) calculate an aggregate household measure of body mass for all members in the household; (iii) represent the production of household body mass as a function of all the measured food groups and food types; (iv) use well-known efficiency analysis techniques (including linear programming and econometric techniques) to estimate household-level food-loss as the percentage of inefficiency found in the household production of body mass; and (v) use econometric methods to analyze the relationship between these food-loss estimates and a number of household characteristics. Additionally, we will conduct sensitivity analyses for methodological choices made in estimating household-level body-mass production and calibration checks using other data.While the general method of estimating firms' production inefficiencies is well established, it has rarely been applied at the household production level, and never - to our knowledge - for the production of body mass as a function of food purchases. Our food-loss estimates will tell us which households are most efficient at converting food purchases into body mass, and which households are less efficient. Ultimately, our analysis will allow us to comment on whether possible food-loss prevention policies should be aimed at food types (such as fresh or processed foods), food sources (such as certain types of food retailers or certain types of restaurants), or at particular household types (such as households in certain income or education ranges). We will also be able to comment on whether SNAP requirements might be altered to improve a household's food-loss estimate.
Animal Health Component
60%
Research Effort Categories
Basic
20%
Applied
60%
Developmental
20%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
60750103010100%
Knowledge Area
607 - Consumer Economics;

Subject Of Investigation
5010 - Food;

Field Of Science
3010 - Economics;
Goals / Objectives
Using data from FoodAPS, we will develop an innovative model based on mature research methods that typically investigate production efficiency analysis. Rather than modeling firm production, however, we will instead model the production of household body mass as a function of a household's measured food-group or food-type purchases. Just as firm-level efficiency analysis estimates an individual firm's inefficiency as the relative distance to the production frontier, we plan to estimate a body-mass production function and measure household food inefficiency (i.e., food loss) as the relative distance to the estimated frontier. Thus, household-level food loss is equated to inefficiency in the production technology that takes food purchases as inputs and transforms them into body mass, which is treated as an output. Once we estimate food loss for each household, we can explore how it varies according to household types. Subsequent policy analysis can then focus either on inputs to the household production function (such as quantities of various food groups) or on differences due to household types (such as SNAP eligibility, education levels, location, etc.). A detailed description of our proposed quantitative methods is discussed in the Approach section.The project has five major goals:1. Develop a theoretical and empirical model for the household production of body mass.2. Clean and manage the FoodAPS data.3. Estimate the model and recover predicted inefficiency results.4. Analyze how household levels of food losses relate to household attributes. And,5. Conduct sensitivity analysis.
Project Methods
Activity 1: Develop a theoretical and empirical model for the household production of body mass.This modeling activity will be carried out in two steps. In the first step, we propose the deterministic portion for the household production function of body mass. This function tells us, from a nutrition science perspective, how various food and nutrient contents are transformed into body mass. At the theoretical level, several qualitative restrictions will be imposed, such as bounding the production function from above to acknowledge our limited capabilities of digesting and converting chemical energy. An empirical challenge here is to adequately specify the food groups that serve as inputs to the body-mass production technology.The second step is to introduce a stochastic component into the deterministic model. It is crucial to model the randomness because there is considerable heterogeneity among people's transformation rates of nutrients into body mass, such as those due to different metabolism rates and unobserved activities such as exercise. Other unobserved or unquantified measures such as family tradition and values may also play important roles in determining food loss that occurs within households. Adding the stochastic component thus allows each household to have their own "best-practice" of converting food and nutrients into body mass.Activity 2: Clean and manage the FoodAPS data. Before estimating the model, the FoodAPS data must be examined, cleaned, and categorized into relevant inputs for the household production function. A preliminary examination of the data tells us that a small number of households reported unreasonably small or large amounts of food acquisition. Several explanations are possible for these potential outliers: (i) these outliers may be reporting errors; (ii) overly large outliers may be purchases planned for delayed consumption, i.e., food purchased in the reporting week but intended for consumption in future weeks; or (iii) outliers may also be the result of hosting and feeding guests. Hence it is necessary to investigate these observations using additional information from the FoodAPS dataset to either account for their behavioral anomalies or rule them out as outliers.Activity 3: Estimate the model and recover predicted inefficiency results. We plan to specify at least two types of household production functions, one that uses categorized food groups as inputs and another that uses categorized nutrient intake levels as inputs. The first may generate results that are most useful for policy analysis, while the second is more closely related to the scientific representation of metabolism. From an empirical verification perspective, a benefit of testing two different input setups is that we can compare the results and indirectly assess how the models performed. Both models will generate household levels of predicted inefficiency, which, in the context of body mass production, translate into predicted household-specific levels of food loss.Activity 4: Analyze how household levels of food losses relate to household attributes. We plan to investigate how household-specific food-loss estimates vary according to various levels of income, education, presence of children, and other household attributes such as SNAP participation, household location, and the household's food environment. Our model makes it possible to conduct policy counterfactuals aimed at specific groups of households. Further refinements such as assigning population weights to each household can also be performed to produce better national representative predictions.Activity 5: We plan to conduct a wide array of sensitivity analyses to test the robustness of our results. First, we plan to re-estimate the household production models for a number of subsamples, e.g., low-income households, urban households, SNAP/WIC households, obese households, and households with known health conditions. Second, we plan to use NHANES data to see if similar results can be obtained from an established dataset.3.2 Methods - Econometric AnalysisActivity 3 requires the estimation of a household production function, and maximum-likelihood techniques are the most widely used methods in production efficiency analysis literature. These techniques are flexible in choosing the probability distributions that best fit the stochastic elements in the model. Like much of the current production-efficiency research, ours specifically introduces two random components. The first, a nonnegative term that is subtracted from the output, captures the food loss in our context. The other random component is a typical white-noise error term that with a mean of zero. Mathematically, our preliminary analysis assumes that follows a half-normal distribution and follows a standard normal distribution. The composite error is therefore .To incorporate heterogeneity, we will formulate the means and variances of the two random variables as functions of household demographics. Another way of achieving group-wise heterogeneity is to sort households into several types and estimate the model separately for each group. The latter approach matches to elements of Activity 5.Activity 4 also requires econometric estimation, but here the second-stage analysis of the household-level food-loss estimates according to household characteristics will rely on ad hoc models that depend on the available household characteristics in the FoodAPS data. More specifically, household-specific food-loss estimates will be regressed on household characteristics to investigate how food loss varies by household type.

Progress 05/15/17 to 05/14/20

Outputs
Target Audience:This project investigates household-level food waste using new empirical methods. Throughout the life of this project the target audience for this project has been researchers in the agricultural and applied economics profession, including those at universities and U.S. government agencies. Additionally, when our results have broader implications, we will also target policy makers and other stakeholders in the food and agricultural fields. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Under the supervision of Jaenicke, one Ph.D. students, Yang Yu (2020), contributed mightily to this research project. This project exposed this student to state-of-the art research training that featured weekly conferences with the faculty supervisor. Yu presented work at several professional meetings and annual research conferences. In addition, Yu will join the agricultural economics faculty at Montana State University in July 2020. Overall, the project was instrumental in Yang Yu's professional development. How have the results been disseminated to communities of interest?Results have been disseminated to target audiences by publishing in peer-reviewed journals (including the published papers listed above). In addition, the food-waste results were picked up by several media sources, including but not limited to the following: U.S. News & World Report, 1/23/20, (https://www.usnews.com/news/healthiest-communities/articles/2020-01-23/americans-waste-240-billion-in-food-each-year-study-says) AgDaily, 1/23/20, https://www.agdaily.com/news/study-u-s-households-waste-nearly-third-food/ CNBC - Squawk Alley, 1/23/20, (20-second clip) What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? The study team's accomplishments for final reporting period ending in May 2020 center on the two peer-reviewed publications that came from this grant. The Yu and Jaenicke (2020) AJAE article accomplishes all the stated goals. More specifically, we use household-level USDA FoodAPS data to estimate a stochastic production frontier that converts acquired food inputs into metabolic energy and estimates the inefficiency in this relationship across households. We then convert this inefficiency estimate into an estimate of household-specific food waste. On average, households waste about 31 percent of the food they acquire. More importantly, we can determine how household characteristics vary with food waste. We find that households with higher incomes, higher self-stated diet quality, and higher-self-stated food security all waste more food. The Yu and Jaenicke (2020) Food Policy article investigates policy-induced food waste for milk as a case study. We find that that a 2012 policy change in New York City that removed a stringent 9-day sell-by date for milk (effectively increasing the sell-by period by 5 or 6 days) decreased milk sales by 10 percent. Using a theoretical model, we also show that this reduction in sales translates to at least an equivalent reduction in milk-related food waste.

Publications

  • Type: Journal Articles Status: Published Year Published: 2020 Citation: Yu, Y., and E.C, Jaenicke. 2020. The Effect of Sell-by Dates on Purchase Volume and Food Waste. Food Policy, forthcoming. https://doi.org/10.1016/j.foodpol.2020.101879.
  • Type: Journal Articles Status: Published Year Published: 2020 Citation: Yu, Y., and Jaenicke, E. C. (2020). Estimating food waste as household production inefficiency. American Journal of Agricultural Economics, 102(2), 525-547. Available online: https://onlinelibrary.wiley.com/doi/10.1002/ajae.12036.
  • Type: Theses/Dissertations Status: Submitted Year Published: 2020 Citation: Yang Yu, Ph.D. 2020. Essays on Food Waste and Consumer Demand Analysis.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2020 Citation: Jaenicke, E.C. Household-level Food Waste: Estimation, Behavioral Insights, and Next Steps, Invited Presentation. Department of Agricultural Economics and Rural Sociology, Auburn University, Feb. 24, 2020.


Progress 05/15/18 to 05/14/19

Outputs
Target Audience:During the past year, the study team was able to complete its estimation of household-specific levels of food waste based on USDA-ERS's FoodAPS. Three related models, estimated as a robustness check, all show that the average levels of household-specific food waste are 31-32 percent. Moreover, houses with higher income, healthier diets, and greater household food security waste more; SNAP and WIC participation, all else equal, are associated with less food waste. In addition, the study team invested one very important potential cause of food waste, namely wasted induced by "sell by" dates. Using both household-level and store-level food purchase data, the study team investigated a 2010 policy change in the sell-by date for milk sold in New York City. The old policy of 9 days after the pasteurization date was eliminated and the industry standard of 14-15 days was adopted. Using a difference-in-differences approach, the study team finds that the policy change caused a 10% (or more) decline in milk sales. Using a simple microeconomic model of utility maximization, this sales drop equates to at least as much decline in food waste. Both sets of results have been presented at several seminars at Penn State and other universities. In addition, two manuscripts have been submitted to peer-reviewed journals. Changes/Problems:The grant has received a one-year no-cost extension. What opportunities for training and professional development has the project provided?The grant provides funding for one PhD student, and that student has (a) conducted empirical analyses, written first drafts of research manuscripts, (c) presented research at annual conferences, and (d) taken the lead on the journal submission process. How have the results been disseminated to communities of interest?Results have been disseminated via seminars and presentaitons out our own institution, at other colleges and universities, and at national conferences. In addtion, working papers are posted at Social Science Research Network's (SSRN's) web page:https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=429112. What do you plan to do during the next reporting period to accomplish the goals?Next steps include (a) moving two completed manuscripts through the publication process at peer-reviewed journals, (b) attempted to extend our main results to estimate food group-specific and household-specific measures of food waste, and (c) conducting an detailed analysis of households that waste the most and the least food.

Impacts
What was accomplished under these goals? Goals 1-5 have been accomplished in the sense that a manuscript containing all these elements is under review. Depending on peer review comments, some elements of these goals may have to be revised. The study team is currently expanding the set of goals.

Publications

  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2018 Citation: Consumer-level Food Waste: Household Estimates, and Effects from Extended "Sell-by" Dates. Invited Presentation. Macalester College Economics Seminar, Saint Paul, MN, Oct. 4, 2018.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2018 Citation: Estimating Food Waste at the Household Level and (ii) Food Waste Due to Sell-by Dates. Invited Presentation. Toulouse School of Economics, Food Economics Group, Toulouse, France. May 24, 2018.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2018 Citation: The Effect of Sell-By Dates on Purchase Volume and Food Waste: A Case of New York Citys Sell-By Regulation of Milk. Conference Presentation. Selected Presentation at the Agricultural and Applied Economics Association (AAEA) conference, August 5-7, 2018, Washington, DC
  • Type: Journal Articles Status: Under Review Year Published: 2018 Citation: Yu, Y., & Jaenicke, E. C. Estimating Household Food Waste Using Food Acquisition Data. American Journal of Agricultural Economics.
  • Type: Journal Articles Status: Under Review Year Published: 2018 Citation: Yu, Y., & Jaenicke, E. C. The Effect of Sell-by Dates on Purchase Volume and Food Waste.


Progress 05/15/17 to 05/14/18

Outputs
Target Audience:This project uses food purchase data from USDA's FoodAPS to investigate and estimate food waste conceptually as an inefficiency term in a household production process. The target audience for this project is researchers in the agricultural and applied economics profession, including those at universities and U.S. government agencies. Additionally, when our results have broader implications, we will also target policy makers and other stakeholders in the food and health field. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?These results have been presented at academic conferences and one seminar. What do you plan to do during the next reporting period to accomplish the goals?Next steps include (a) refining our household production model to align it more closely with nutrition science, (b) conducting additional robustness checks on the model specification and data, and (c) investigating subsample results

Impacts
What was accomplished under these goals? The study team has completed our initial analysis using new FoodAPS data from USDA-Economic Research Service. More specifically, we estimated food waste at the individual household level indirectly using a stochastic production frontier approach. The estimated average percentage of waste is 32.4%. However, by taking proxied physical activities into consideration and utilizing limited information maximum likelihood estimation, our estimate for average household food waste is reduced to 27.6%. We are also able to explore the relationship between food waste and important demographic variables, and we find that households with higher levels of income and food security waste more food. So do households that report healthier diets, presumably because these households purchase larger amounts of perishable fruits and vegetables. Lastly, participants if food assistance programs, such as SNAP and WIC, waste less.

Publications

  • Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Presentations: Invited Jaenicke, E.C., with Y. Yu. Estimating Food Waste at the Household Level. University of Minnesota, Department of Applied Economics, October 25, 2017. Presentations: Annual Conferences and Professional Meetings Yu, Y., and E.C. Jaenicke. Assessing Household Food Loss: An Efficiency Analysis. Selected Paper at the Northeastern Agricultural and Resource Economics Association (NAREA) conference, June 11-12, 2017, Arlington, VA. Selected Poster at the Agricultural and Applied Economics Association (AAEA) conference, July 30-August 1, 2017, Chicago, IL.