'The Impact of Waiting Periods on Food Subsidy Effectiveness, 2018.' (AsPredicted #9427)
Author(s) Andy Brownback (University of Arkansas) - ABrownback@walton.uark.edu Alex Imas (Carnegie Mellon University) - alex.imas@chicagobooth.edu Michael Kuhn (University of Oregon) - mkuhn@uoregon.edu
Pre-registered on 2018/03/26 - 10:18 AM (PT)
1) Have any data been collected for this study already? It's complicated. We have already collected some data but explain in Question 8 why readers may consider this a valid pre-registration nevertheless.
2) What's the main question being asked or hypothesis being tested in this study? How do subsidies for healthy food affect food purchasing behavior? Motivated by Imas, Kuhn and Mironova (2018), what are the impacts of waiting periods and commitment on how consumers choose between food subsidies? How is food purchasing behavior differentially affected by allowing individuals to choose food subsides vs. giving them predetermined subsidies?
Subjects in our study will be given discounts on grocery purchases. They can choose between discounts on fruits & vegetables (healthy purchases) and discounts on baked goods (unhealthy purchases). We will experimentally test the impact of making the subsidy choice A) with a waiting period—a delay between learning about the choice and the ability to making the choice, B) with commitment—where binding choices are made before a shopping trip takes place, both relative to individuals making the choice without a waiting period or commitment.
We will also experimentally vary whether individuals encounter a choice at all: some will be unconditionally given a subsidy on fruits & vegetables. We will compare how much of the subsidy is used between those who choose their own subsidy and those who do not. Our hypothesis is that individuals who choose their own subsidy will use more of it.
We will also explore the impact of time preferences (as measured in the baseline survey) on the food choices of our subjects. We hypothesize that more impatient subjects will be more likely to choose the subsidy for immediately satisfying, but unhealthy foods, and that they will be more affected by the waiting period and commitment treatments.
3) Describe the key dependent variable(s) specifying how they will be measured. Food subsidy choice is binary: = 1 if the choice is for fruits & vegetables, = 0 if the choice is for baked goods. Subsidy use is the value of goods purchased within the category covered by the subsidy. We will report this both as a dollar amount and as a fraction of overall food spending. We will analyze the overall "nutrition" of a shopping trip, where nutrition is the ratio of spending on fruits & vegetables and meat & dairy to spending on baked goods, snacks and sweetened beverages. Our baseline and endline surveys capture behavior and welfare changes. Behavior changes are captured by shopping receipts from before and after the study. We will analyze these as stated above. Both surveys contain single-day food consumption diaries. We will analyze changes in the food diaries along 3 dimensions: (1) are there fruits & vegetables, (2) the ratio of fruits & vegetables to other goods, (3) overall nutrition. Welfare changes are captured with direct questions. We ask 2 questions on food sufficiency: 1)Which of these statements best describes the food eaten in your food household in the last 30 days? A: “Enough of the kinds of food we want to eat;” “Enough, but not always the kinds of food we want to eat;” “Sometimes not enough to eat;” “Often not enough to eat” 2) In the last 30 days, did you ever worry about whether your food would run out before you got money to buy more? A: “Almost always;” “Most of the time;” “About half of the time;” “Some of the time;” “Almost never.” We ask two questions on healthy foods: 1) In the last 30 days, did you ever feel like your food household couldn't afford to eat well-balanced (healthy) meals because you couldn't afford it? (same answers as before) 2) Do you think your food household eats the right amount of fruits and vegetables? A: “Yes, we eat the right amount;” “No, we should eat more;” “No, we should eat less.” We ask a question on physical health: How would you rate your physical health status? A: “Excellent;” “Very good;” “Good;” “Fair;” “Poor.” We ask a question on cognitive well-being: Thinking about the past couple weeks, do you find that you have difficulty maintaining energy, focus or attention? A: “Almost always;” “Most of the time;” “About half of the time;” “Some of the time;” “Almost never.” We collect self-reports of behavior change: Did you change the foods you ate because of the study (check all that apply)? A: “I ate more fruits & vegetables;” “I ate fewer fruits & vegetables;” “I ate more baked goods;” “I ate fewer baked goods.” We will analyze the within-individual changes in these variables: = 0 if the measure stayed the same, = 1 if the measure improved, = -1 if the measure worsened.
4) How many and which conditions will participants be assigned to? There are 5 conditions each varying the nature and existence of food subsidies. Subjects are assigned using block-randomization following the baseline survey. Randomization is blocked based on participation in SNAP, and a stated desire to eat more fruits & vegetables, to ensure these characteristics are balanced across treatments.
Control condition: Subjects do not receive food subsidies. They participate in the baseline and endline and their shopping trips in between are observed. This group is used the show a lack of time trends between baseline and endline (or to adjust the main results if there are time trends) and the lack of experimenter observation effects (or to adjust the main results if there are experimenter observation effects).
No-choice condition: Subjects receive a 30% subsidy on fruits & vegetables (up to a $10 value) in 4 grocery shopping trips between the baseline and endline surveys.
Immediate condition: Subjects choose between a 30% subsidy on fruits & vegetables or a 30% subsidy on baked good (both up to a $10 value) in 4 grocery shopping trips between the baseline and endline surveys. They make this choice in the grocery store during their shopping trip.
Commitment condition: Subjects choose between a 30% subsidy on fruits & vegetables or a 30% subsidy on baked good (both up to a $10 value) in 4 grocery shopping trips between the baseline and endline surveys. They make this binding choice between 4 and 48 hours prior to their grocery shopping trip.
Waiting period condition: Subjects choose between a 30% subsidy on fruits & vegetables or a 30% subsidy on baked good (both up to a $10 value) in 4 grocery shopping trips between the baseline and endline surveys. They make this choice in the grocery store during their shopping trip but are notified and reminded of this upcoming choice 4 to 48 hours prior to their grocery shopping trip.
5) Specify exactly which analyses you will conduct to examine the main question/hypothesis. Food subsidy choice: we will estimate a linear random effects regression of the binary variable for choosing fruit & vegetable subsidies on treatment indicators. Standard errors will be clustered at the individual level. We are very interested in whether the impact of waiting periods and commitment are related to individual time preferences. In the baseline survey, we obtain measures of individual discounting (1-12 scale), present-bias (-11 to 11 scale) and impulsivity (-11 to 11 scale). We will estimate a version of the random effects model with these interaction terms.
Subsidy use: we will estimate a linear random effects regression of money spent in the subsidy category on treatment indicators. As stated earlier, we will also do this with money spent in the subsidy category as a fraction of overall food spending, and the overall nutrition of a shopping trip. We will control for the individual level of spending on the subsidized goods observed prior to treatment in the baseline.
Each condition runs for 4 weeks. Because conditions repeat each week, the first exposure may produce different outcomes than subsequent exposures. As such, we will also study just the first exposure separately from all others in the above specifications, dropping the random effects from the model.
Behavior and welfare changes: these differences between the baseline and endline survey will be regressed on treatment indicators with OLS regressions.
6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations. There are specific rules that subjects need to follow about when grocery shopping trips occur relative to choices. Individuals who fail to follow those rules once per shopping trip are given an opportunity to make up that shopping trip. If they fail to follow those rules again, they are dropped from the study. This will be performed blindly to any data those subjects have provided.
Individuals who fail to take pictures of their receipts at the store (based on the geo-tag of their receipt) will be dropped from the analysis. Two such failures will result in their expulsion from the study.
Since there is a chance that subjects may not complete all 4 weeks of the study, we will also consider the main analyses both including and excluding the subjects who fail to complete the entire study.
7) How many observations will be collected or what will determine sample size? No need to justify decision, but be precise about exactly how the number will be determined. Our goal is for 750 subjects (150 in each of the 5 treatments) to complete the baseline survey. We will screen substantially more subjects and invite 750 eligible to complete the baseline. If individuals fail to complete the baseline, we will backfill with more eligible subjects. We will backfill subjects to ensure that 750 subjects complete the baseline and each treatment has 150 subjects. Since subjects don’t learn of their treatment until the week 1 survey, we do not have to worry about attrition based on treatment assignment at this point.
8) Anything else you would like to pre-register? (e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?) We wish to explore whether the impact of our conditions differs based on household income and whether there are kids in the household.
When analyzing heterogeneous treatment effects by time preference measures, we may also characterize our time preference measures as structural utility parameters rather than the raw choice data.
We plan to study differential attrition by treatment, and to analyze compliance and completion over time.
We collected a pilot sample of 50 individuals to ensure our technical procedures were successful. These data will not be included in our main analysis.
