- Email: [email protected]
A Plate Waste Evaluation of the Farm to School Program
ARTICLE IN PRESS Research Article A Plate Waste Evaluation of the Farm to School Program Jaclyn D. Kropp, PhD1; Saul J. Abarca-Orozco, PhD2; Glenn D. Israel, PhD3; David C. Diehl, PhD4; Sebastian Galindo-Gonzalez, PhD3; Lauren B. Headrick, MS, RD5; Karla P. Shelnutt, PhD6 ABSTRACT Objective: To investigate the impacts of the Farm to School (FTS) program on the selection and consumption of fruits and vegetables. Design: Plate waste data were recorded using the visual inspection method before and after implementation of the program. Setting: Six elementary schools in Florida: 3 treatment and 3 control schools. Participants: A total of 11,262 meal observations of National School Lunch Program (NSLP) participants in grades 1–5. Intervention: The FTS program, specifically local procurement of NSLP offerings, began in treatment schools in November, 2015 after the researchers collected preintervention data. Main Outcome Measures: The NSLP participants’ selection and consumption of fruits and vegetables. Analysis: Data were analyzed using Mann–Whitney U and proportions tests and difference-indifference regressions. Results: The NSLP participants at the treatment schools consumed, on average, 0.061 (P = .002) more servings of vegetables and 0.055 (P = .05) more servings of fruit after implementation of the FTS program. When school-level fixed effects are included, ordinary least squares and tobit regression results indicated that NSLP participants at the treatment schools respectively consumed 0.107 (P < .001) and 0.086 (P < .001) more servings of vegetables, on average, after implementation of the FTS program. Conclusions and Implications: Local procurement positively affected healthy eating. Key Words: Farm to School, children, fruit consumption, vegetable consumption, school lunch (J Nutr Educ Behav. 2017;■■:■■–■■.) Accepted October 17, 2017.
INTRODUCTION In 2010, the Healthy, Hunger-Free Kids Act (HHFKA) was signed into law with the primary focus of improving nu-
trition for children. As part of this legislation, the US Department of Agriculture (USDA) established a Farm to School (FTS) program to help schools increase the amount of local foods
1
Department of Food and Resource Economics, University of Florida, Gainesville, FL Department of Family, Youth, and Community Sciences, University of Florida, Institute of Food and Agricultural Sciences Extension, Family Nutrition Program, Gainesville, FL 3 Department of Agricultural Education and Communication, Program Development and Evaluation Center, University of Florida, Gainesville, FL 4 Program Planning and Evaluation, University of Florida, Gainesville, FL 5 Family Nutrition Program, University of Florida, Institute of Food and Agricultural Sciences Extension, Gainesville, FL 6 Family and Nutrition Program, Family Youth and Community Sciences Department, University of Florida, Gainesville, FL Conflict of Interest Disclosure: The conflict of interest disclosures can be found online with this article on www.jneb.org. Address for correspondence: Jaclyn D. Kropp, PhD, Department of Food and Resource Economics, University of Florida, PO Box 110240, Gainesville, FL; Phone: (352) 294-7631; Fax: (352) 846-0988; E-mail: [email protected] © 2017 Society for Nutrition Education and Behavior. Published by Elsevier, Inc. All rights reserved. https://doi.org/10.1016/j.jneb.2017.10.005 2
Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
they procure and serve in their cafeterias.1 The FTS programs typically consist of ≥1 of the following activities: local procurement of products served in school cafeterias, handson learning activities such as school gardens, and integrated nutrition education.2 To measure progress of FTS activities, the USDA began conducting an FTS census in 2013. According to the most recent 2015 FTS census, which collected data from the 2013– 2014 school year, more than 42,000 schools in over 5,200 school districts had FTS programs, reaching more than 23.6 million children.3 Approximately 19% of all school districts served at least 1 locally-sourced product daily.2 Although FTS programs help schools meet the updated nutrition standards that resulted from HHFKA, 66% of schools with FTS programs also reported other benefits such as increased participation in the National School Lunch Program (NSLP) and less food waste.4 The FTS programs also support
1
ARTICLE IN PRESS 2
Kropp et al
local economies3 and aim to increase students’ consumption of fruits and vegetables through increased exposure to fresh produce.5 Although FTS programs are expanding at a rapid rate across the country, several prior studies focusing on the effects of FTS programs on promoting healthy eating found mixed results. Joshi et al6 reviewed 11 prior studies on the effects of FTS programs on dietary behaviors and found that 10 reported positive dietary changes; 4 of those studies reported increases in the consumption of fruits and vegetables outside school. However, the studies that were reviewed relied on production records from school cafeterias, selfreported behaviors, and dietary recall, and hence might have been less accurate than studies in which selection and consumption were observed directly. Recent research recommended that FTS programs use plate waste methodologies to measure impacts on fruit and vegetable consumption to decrease misreporting,7 because observing and measuring plate waste is an increasingly common and more accurate method used to investigate the selection and consumption of foods, particularly in school cafeteria settings. Although 3 prior studies used plate waste methods to evaluate their FTS programs,8-10 Yoder et al8 did not include control schools not participating in the FTS program for comparison, Yoder et al9 used only FTS program offerings as a control variable, and Jones et al10 did not compare program effects over time. Therefore, the objective of this study was to examine the effects of serving locally-procured produce as part of the Alachua County Public Schools (ACPS), FL, FTS program on the selection and consumption of fruits and vegetables served as part of the NSLP. This study advances the literature by using preintervention and postintervention plate waste data collected at control schools without FTS activities and treatment schools with FTS programs. Using Mann–Whitney U and proportions tests and differencein-difference regression, the researchers tested the hypotheses that fruit and vegetable selection and consumption would increase at the treatment schools after implementation of the FTS program.
Journal of Nutrition Education and Behavior
METHODS After the researchers obtained approval from the University of Florida’s institutional review board, they collected preintervention data at 6 elementary schools in the ACPS system in October, 2015. Postintervention data were collected in April, 2016 after implementation of the FTS program at the treatment schools. The ACPS system was composed of suburban and rural schools with 22 elementary schools. Data were collected from 3 elementary schools receiving the treatment (FTS program) and 3 control schools (not receiving the FTS program). Although FTS programs frequently consist of local procurement, nutrition education, and school gardens, this analysis focused on procurement of NSLP offerings from local producers. One treatment and 1 control school had school gardens that were operational for >2 years before the study period, but none of the study schools had formal nutrition education programs during the normal school day during the study period. The products grown in the school gardens were for educational purposes only and were not served in the school cafeterias before or during the study period. Treatment schools began receiving FTS products in early November, 2015 shortly after baseline data were collected; however, owing to unseasonably warm weather, regular deliveries did not begin until January, 2016. When FTS products were offered in the cafeterias, these products were promoted using signage, which included the name of the local farm supplying the product. The FTS produce primarily consisted of raw vegetables including leafy greens, cucumbers, and peppers used mainly in NSLP salad offerings. The FTS products were offered approximately 50 days during the study period at each treatment school. All schools in the sample except for 1 of the control schools were Title I schools, which meant that they had a high percentage of children from low-income families.11 Low-income families were more likely to participate in the NSLP and also more likely to have limited access to fresh fruits and vegetables in the home because of the food environment or income constraints.12 During the 2015–2016 school year, only Title I elementary
•
Volume ■■, Number ■■, 2017
schools in the ACPS system had the opportunity to receive FTS products; however, not all Title I elementary schools in the ACPS system participated in the program. Three Title I elementary schools in the ACPS system did not participate because of logistical and distributional issues associated with delivering products to their rural locations; hence, assignment into the FTS program was not random. One of the Title I elementary schools not participating in the FTS program declined to participate in the study; therefore, a non–Title I school was included in the control set of schools (Control 3 in Table 1). Although this school was more affluent (lower proportion of students eligible for free and reducedprice lunch) than the other study schools, the demographic mix was similar (see Table 1). Three days of preintervention and 3 days of postintervention data were collected at each school based on the conclusion of Martin et al13 that measuring plate waste for 3 days was the statistically significant representation of a 5-day week. Data collection occurred on Tuesdays, Wednesdays, and Thursdays to avoid capturing potential weekend effects in which students may have exhibited different selection and consumption behaviors on Fridays and Mondays owing to food insecurity at home over the weekend. Preintervention data collection occurred over a 3-week period in October, 2015, with data collected at 1 treatment and 1 control school each week. The ACPS system set the menu at the district level; thus, data were collected from 1 treatment and 1 control school each week with the same menu. Postintervention data were collected in a similar manner in April, 2016. The district menu followed a 3-week menu cycle; for each school in the study, pre- and postintervention data were collected when the same menu items were offered at the school. Schools that participated in the NSLP were required to offer all 5 meal components (meat/meat alternative, grain, fruit, vegetable, and low-fat milk) each day; for the school to receive federal reimbursement for the meal, a student had to select at least 3 of the 5 offered components. 14 Furthermore, NSLP guidelines required
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
Kropp et al
3
Table 1. Free and Reduced-Price Eligibility and Demographic Information of Students Enrolled in the Study Schools Number of Students
Free/Reduced-Price Eligible
White
Black
Hispanic
Asian
Other
Control 1
501
47%
58%
26%
8%
2%
6%
Control 2
452
58%
50%
33%
9%
1%
7%
Control 3
717
31%
52%
19%
11%
11%
7%
Treatment 1
742
71%
20%
53%
12%
6%
9%
Treatment 2
714
75%
52%
23%
12%
2%
11%
Treatment 3
636
42%
49%
29%
8%
5%
9%
School
Note: School-level demographic information was obtained from the Florida Department of Education and hence these values reflect the demographics of the enrollment. Since not all enrolled students participate in the National School Lunch Program, the demographics of the study participants may be different. Control 3 is the only non-Title I school.
all reimbursable meals to contain either a fruit or vegetable.14 Plate waste data were collected from students in first through fifth grades. Because this study focused on determining the effects of the FTS program on selection, Head Start, prekindergarten, and kindergarten students were excluded from the analysis because at the study schools, these younger students were simply given a lunch tray containing all 5 meal components when they went through the lunch line and therefore did not actually select meal components. There are several methods for collecting plate waste data. The quarterwaste method relies on investigators visually examining each tray after consumption and estimating how much of each food item is wasted. This method was validated in a variety of settings, including school cafeterias.15 Hanks et al15 compared the quarter-waste method, half-waste method, and digital photography method for reliability, accuracy, and cost-effectiveness. The quarterwaste method involves trained data collectors rating waste after consumption on a quarter-serving basis, whereas with the half-waste method data collectors rate waste to the nearest halfserving. The digital photography method involves taking photographs of the plate before and after consumption and calculating plate waste by comparing the pair of images. The authors found that the quarter-waste method had the highest interrater and inter-method reliability, whereas digital photography had the lowest; they also found that digital photography was par-
ticularly inaccurate when packaged foods were prevalent, because investigators cannot always determine from a photo whether a package is empty or full.15 Although Martins et al16 found that visual estimation methods overestimate waste by an average of 8 g compared with weighing food before and after consumption, the weighing method is much less efficient for highvolume observations. Because of its reliability and efficiency, the current study employed the quarter-waste method. For each serving of a particular food item taken, data collectors recorded the portion of a standard serving not eaten on a quarter basis using a scale from 0 to 4; 0 indicated that nothing was wasted (the entire portion was consumed), whereas 4 indicated that the entire portion was wasted (the item was not touched). Before the start of lunch, data collectors weighed 3–5 standard servings of each food item to be served in the cafeteria that day. This allowed the data collectors to become familiar with standard servings of each item. The USDA sets NSLP meal pattern requirements by grade. The requirements and standard serving sizes for kindergarten through fifth grade are the same.17 Because this study used data for first- through fifth-graders, standard servings were the same for all meal observations. Each school differed in the methods used to return trays and collect trash at the end of the lunch periods. Thus, unique data collection protocols were developed for each school to ensure that plate waste data were recorded ac-
curately. At each of the study schools, students sat with their classmates (students with the same teacher sat together in the lunchroom) and trash was collected table by table. Thus, in addition to student-level plate waste data, data collectors recorded students’ teachers’ names to link students’ data to their grade-level information. Data collectors worked in teams of 5 to ensure that plate waste data were gathered for all students participating in the NSLP program without delaying the normal trash collection process at the schools; data collectors collected plate waste data as each table was dismissed from the lunchroom. Each individual NSLP meal component was served in a single-serving container or package. When students went through the lunch line, they made selections by choosing the individual meal components from the line. These empty and partly empty containers were used to determine students’ selections and plate waste. Food item data were categorized into the appropriate NSLP categories: meat/meat alternative, grain, fruit, vegetable, and low-fat milk. Because some students purchased >1 meal/d or >1 fruit or vegetable component, the total daily number of servings selected and consumed in each NSLP category was computed from the plate waste data for each NSLP participant.
DATA ANALYSIS Because selection is a discrete variable, the researchers used proportions tests to determine whether there were
ARTICLE IN PRESS 4
Kropp et al
significant increases in the selection of fruits and vegetables after implementation of the FTS program. Mann– Whitney U tests were used to determine whether there were significant increases in the consumption of fruits and vegetables after the FTS program was implemented. Mann– Whitney U tests were employed because skewness and kurtosis tests for normality indicated that the consumption of fruits and vegetables was not normally distributed (P < .001). In addition, the researchers conducted multivariate regression analyses. Selection was modeled using a logistical regression in which the dependent variable took the value of 1 if the item was selected, and 0 otherwise. Marginal effects were calculated from the estimated logistical regression parameters to determine how the independent variables were related to the probability that a fruit (or vege-
Journal of Nutrition Education and Behavior table) was selected. Consumption was modeled using ordinary least squares in which the dependent variable was the number of servings of fruits (or vegetables) consumed. Because many students did not select the fruit (or vegetable), and thus could not consume the fruit (or vegetable), the data contained numerous 0 observations for the selection and consumption of these items. Hence, analyses for consumption were repeated using tobit models, which model the selection and consumption pieces together. Because assignment to the treatment group was not random, a difference-in-difference approach was used to estimate the treatment effects of the FTS program on the selection and consumption of fruits and vegetables. This approach estimated the effects of the FTS program at treatment schools while controlling for potential changes in the control schools. Grade and data col-
•
Volume ■■, Number ■■, 2017
lector controls were included as covariates in all analyses. Data collector controls were included to control for potential interrater differences. To control for unobserved differences between schools that did not vary over the study period, such as differences in other nutrition interventions, including implementation of behavioral nudges, MyPlate posters in the cafeteria, and the presence of school gardens, analyses were repeated including school-level fixed effects. All analyses were conducted using Stata (version 12, StataCorp LP, College Station, TX; 2011).
RESULTS A total of 11,262 student meal observations were collected and analyzed. Table 2 shows the proportion of NSLP meals selected by students that contained a vegetable or fruit, and the
Table 2. Pre-Intervention and Post-Intervention Selection and Consumption of Vegetables and Fruits by Treatment Group Vegetables
Fruits
Selected
Consumed
Selected
Consumed
0.315
0.161
0.777
0.357
(0.010)
(0.008)
(0.009)
(0.010)
Preintervention Control Treatment Z-Score (Treatment—Control)
0.378
0.167
0.830
0.493
(0.008)
(0.006)
(0.006)
(0.009)
4.630
1.079
4.793
8.324
P < .001
P = .28
P < .001
P < .001
0.399
0.202
0.709
0.300
(0.011)
(0.008)
(0.010)
(0.010)
Postintervention Control Treatment Z-Score (Treatment—Control) Preintervention vs. Intervention Control Z-Score (Post—Pre) Treatment Z-Score (Post—Pre)
0.453
0.215
0.831
0.466
(0.008)
(0.007)
(0.006)
(0.009)
3.970
0.987
10.719
11.399
P < .001
P = .32
P < .001
P < .001
5.458
4.017
−4.880
−5.128
P < .001
P < .001
P < .001
P < .001
6.566
5.438
0.083
−2.698
P < .001
P < .001
(P = .93)
P = .007
Note: Selected indicates the proportion of National School Lunch Program meals selected by students that contained the item (vegetable or fruit). Consumed indicates the average servings consumed by students. Proportion tests were used to test for differences in “selected”; Mann-Whitney U test were used to test for differences in “consumed”. Standard errors are in parentheses. Bold indicates significance at P = .05.
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
Kropp et al
5
Table 3. Difference-in-Difference Treatment Effects of the Farm to School Program on the Selection and Consumption of Vegetables and Fruits (n = 11,626) Vegetables
Selecteda Consumed
b
Consumed if Selected
c
Fruits
Marginal Effect
Standard Error
P-value
Marginal Effect
Standard Error
P-value
−0.039
0.026
.14
0.036
0.022
.09
0.061
0.020
.002
0.055
0.028
.05
0.034
0.018
.06
0.062
0.021
.003
a Estimated using a logit model. bEstimated using OLS. cEstimated using a tobit model; predicted marginal effects conditional on the selection of the item being positive (yj | xj > 0) are reported. Note: Controls for data collectors and student grade-level are included as covariates.
average number of daily servings of these items consumed by NSLP participants at the treatment and control schools. During both the preintervention and postintervention phases of the study, selection of fruits and vegetables was significantly (P < .001) higher at the treatment schools than at the control schools. Also, during both the preintervention and postintervention phases of the study, consumption of fruits was significantly (P < .001) higher at the treatment schools than at the control schools. Selection and consumption of vegetables was significantly (P < .001) higher during the postintervention phase of the study at both treatment and control schools than during the preintervention phase, whereas consumption of fruits was significantly lower at both treatment (P = .007) and control (P < .001) schools during the postintervention phase of the study relative to the preintervention period. Thus, there is evidence that selection and consumption behaviors differed between treatment and control schools serving the same menu items both before and after introduction of the FTS program. Selection and consumption of fruits and vegetables were significantly different at the treatment and control schools before implementation of the FTS program and the treatment was not randomly assigned; thus, the researchers used difference-in-difference models to estimate the treatment effects of the FTS program. Because it can be difficult to interpret the coefficient estimates of a logistical regression, the marginal effects are presented instead of the coefficient estimates for the models pertaining to
the selection of fruits and vegetables in Table 3. As shown in Table 3, the FTS program did not have a significant effect on the selection of vegetables or fruits. However, the OLS analysis indicated that students at the treatment schools consumed, on average, 0.061 more servings of vegetables (P = .002) after implementation of the FTS program. Although this effect was statistically significant and small, it represents an increase of approximately 37% over the average servings of vegetables consumed at the treatment schools before the FTS program was introduced. Furthermore, the OLS analysis pertaining to fruit consumption indicated that students at the treatment schools consumed 0.055 more servings of fruit, on average (P = .05), after implementation of the FTS program. The result was small, but it represents an increase of approximately 11% over the average consumption of the fruit component at treatment schools before the FTS program was introduced. Part of the reason why the treatment effect on the consumption of vegetables using the OLS estimates was so small is that only 45% of selected NSLP meals at the treatment schools contained a vegetable after the FTS program was implemented. Hence, 55% of the consumption observations used in the analysis from those schools took the value of 0 after the program was implemented. Thus, the analysis was repeated using a tobit model. Coefficient estimates of the tobit model represent the linear effect of the covariates on the uncensored latent variable; hence, Table 3 presents the predicted marginal effects
conditional on yj | xj > 0 or on the selection of the item being positive (consumed if selected). The predicted conditional marginal effects from the tobit analysis indicated that students who selected a fruit at the treatment schools consumed 0.062 more servings, on average (P = .003), after the FTS program was implemented. However, the tobit results indicated that the effect of the FTS program on the consumption of vegetables by students who selected the vegetable was not significant. Table 4 presents the results of the analyses when school-level fixed effects that control for school-level unobserved characteristics that did not vary over the study period, such as behavioral nudges, position of lunch items on the lunch line, school gardens, etc. are included as additional covariates. The OLS and tobit results indicate that NSLP participants at the treatment schools consumed 0.107 (P < .001) and 0.086 (P < .001) more servings of vegetables, on average, following the implementation of the FTS program, respectively. The OLS results represent an increase of approximately 65% over the average consumption of vegetables at the treatment schools before the introduction of the FTS program. The effect of the FTS program on fruit consumption is no longer significant for the OLS (P = .29) or the tobit estimations (P = .09).
DISCUSSION This study contributes to the existing literature on FTS programs by analyzing plate waste data to estimate the effects of a new FTS program, specifically the local procurement
ARTICLE IN PRESS 6
•
Journal of Nutrition Education and Behavior
Kropp et al
Volume ■■, Number ■■, 2017
Table 4. Difference-in-Difference Treatment Effects of the Farm to School Program on the Selection and Consumption of Vegetables and Fruits with School-level Fixed Effects (n = 11,626) Vegetables
Selecteda Consumed
b
Consumed if Selected
c
Fruits
Marginal Effect
Standard Error
P-value
Marginal Effect
Standard Error
P-value
0.026
0.030
.39
0.022
0.020
.27
0.107
0.020
<.001
0.030
0.028
.29
0.086
0.018
<.001
0.040
0.024
.09
a Estimated using a logit model. bEstimated using OLS. cEstimated using a tobit model; predicted marginal effects conditional on the selection of the item being positive (yj | xj > 0) are reported. Note: Controls for data collectors, student grade-level, and school-level fixed effects are included as covariates.
aspect of the program, on the selection and consumption of fruits and vegetables. While several previous studies found that FTS programs had positives impacts on students’ consumption of fruits and vegetables,6,8-10 many of these studies relied on school cafeteria production records, selfreported behaviors, and dietary recall data, and hence, might be less accurate than this study that observed selection and consumption directly in school cafeterias. The current study analyzes the selection of fruits and vegetables using logit models and consumption of fruits and vegetables using both OLS and tobit models. While the OLS analyses provide estimates of the average impact of the FTS program on consumption, these estimates include numerous 0 consumption observations due to the low levels of selection. Thus, tobit models are also included to determine the average impact of the FTS program on consumption conditional on the item (fruit or vegetable) being selected. Jones et al10 found that results can vary when school-level controls are included; hence, all analyses were performed with and without the inclusion of school-level fixed effects. The models without school-level fixed effects indicate the FTS program led to increases in the consumption of fruits and vegetables by students at treatment schools following the implementation of the program. Since the majority of the FTS products offered in the treatment schools were vegetables (mostly leafy greens, cucumbers, and peppers), there is some evidence of spillover effects in which signage promoting the FTS vegetables also impacted non-FTS fruit consump-
tion. However, the models that controlled for unobserved schoollevel characteristics, such as signage in school lunchrooms, other nutrition education interventions, school gardens, or nutrition nudging strategies, indicate that the consumption of vegetables, on average, increased at the treatment schools following the introduction of local procurement, but the program did not impact fruit consumption. These results are more consistent with expectations as the majority of FTS items were vegetables. The differences in the results with and without school-level controls highlight the importance of controlling for different school-level effects that may further influence fruit and vegetable consumption. The results of the current study agree with previous studies that used plate waste methodologies to evaluate FTS programs. Yoder et al8 appears to be the first study to use plate waste methodology to evaluate the impact of FTS programs on fruit and vegetable intake of elementary school children over one year. The researchers compared the effect of years of FTS programming at the schools on combined fruit and vegetable consumption but did not estimate consumption of fruits and vegetables separately, so it is difficult to compare current results to the results of the Yoder et al8 study. However, Yoder et al8 reported an increase in student meals containing fruits and vegetables from baseline to post for all schools, and the percentage of trays with no fruit and vegetable disappearance, which served as a proxy for fruit and vegetable consumption, decreased significantly for schools with previous FTS programming compared
to schools with new FTS programs. Based on this previous finding it is possible that fruit and vegetable consumption will increase in the current study’s treatment schools the longer the schools offer the FTS program. Jones et al10 also used photograph plate waste data to estimate servings of fruits and vegetables consumed. The researchers reported that students in schools with FTS programs consumed more vegetables than students at the control schools. However, when the researchers controlled for school, the difference in consumption was no longer significant. In addition, they found students in schools with FTS programs ate less fruit than students in the control schools, the effect of which decreased when controlling for schoollevel factors. The current study found that students in treatment schools ate more vegetables than the students at the control schools even after controlling for school-level fixed effects. In a multi-year (2010–2013), cross-sectional study of Wisconsin elementary schools participating in FTS programs, Yoder et al5 found locallysourced items were wasted more than conventionally-sourced items. However, Yoder et al 9 used FTS program offerings as a control variable to identify the effects the HHFKA of 2010 and only analyzed data from schools with FTS programs. The current study analyzed the impact of the FTS program on selection of all fruits and vegetables regardless of the source of the specific item, making comparisons difficult. Although impacts of the FTS program on consumption in the current study were modest, consuming more vegetables and fruits at each
ARTICLE IN PRESS Journal of Nutrition Education and Behavior meal throughout the year could result in a large impact over time. The magnitudes of the changes in consumption are similar to the results of other studies that analyzed the impacts of creatively naming and branding lunch items.18,19 Hence, it is unclear if the observed changes in consumption behaviors are due to students’ preferences for the FTS products or simply promotion of these products.
IMPLICATIONS FOR RESEARCH AND PRACTICE Several prior studies have found that nudges such as renaming food items and promoting healthier products in school cafeterias can increase selection18-20; combining these techniques with techniques already used in the district to promote FTS products might increase the selection of FTS products. While the study finds positive impacts on the consumption of vegetables and fruits, the study does not investigate what is driving the observed changes in behaviors. Perhaps these changes are driven by students’ preferences for the FTS products due to the freshness and quality of the FTS products or student nudging by the signage in the school cafeterias promoting the products. However, prior studies found that nudging impacts selection as well as consumption18-20; yet, the lack of significance of the FTS program on selection of fruits and vegetables in this study suggests that nudging may not be driving these results. Further research is necessary to determine what factor or factors are responsible for the observed changes in behaviors. In addition, further research is needed to compare the efficacy and costs of various FTS product promotional strategies to determine the best practices. This study only analyzes the local procurement aspect of FTS program. The effects of local procurement on children’s nutrition behaviors are relatively modest, and future studies need to evaluate the impacts of FTS programs within the larger context of community and school nutrition interventions. Specifically, further studies are necessary to determine the impacts of school gardens and nutrition education components of FTS programs.
•
Volume ■■, Number ■■, 2017
Furthermore, additional research is needed to determine the financial cost of wasted FTS products as well as potential waste reduction due to schools receiving FTS products that are fresher than products received from conventional distributors. Since this study was only conducted in 1 school district in Florida, the results may not be generalizable. More research is needed to determine if other FTS programs have similar effects in other schools throughout the country as the results may depend on the types of products offered. In the study district, the majority of the FTS offerings were vegetables; additional research is needed to determine the effects of the program when local fruits, dairy, and meats are offered. Furthermore, more research is needed to determine why selection and consumption behaviors differed between treatment and control schools before introduction of the FTS program. Possible explanations included differences in the implementation of Smarter Lunchroom strategies, other programs that promote healthy eating such as the USDA’s Fresh Fruit and Vegetable Program, differences in the quality, preparation, and presentation of the produce, and the preferences of the students.
ACKNOWLEDGMENTS This study was supported by the University of Florida/Institute of Food and Agriculture Sciences, Extension Family Nutrition Program funded by the United States Department of Agriculture through the Florida Department of Children and Families (#LF923).
REFERENCES 1. United States Department of Agriculture, Food and Nutrition Service. The Farm to School Census; 2015. https:// farmtoschoolcensus.fns.usda.gov/ about. Accessed September 15, 2017. 2. Ralston K, Beaulieu E, Hyman J, Benson M, Smith M. Daily Access to Local Foods for School Meals: Key Drivers, EIB-168, U.S. Department of Agriculture, Economic Research Service; March 2017. https://www.ers.usda.gov/pub lications/pub-details/?pubid=82944. Accessed August 15, 2017.
Kropp et al
7
3. United States Department of Agriculture, Food and Nutrition Service. Farm to School Program: fact sheet: Research Shows Farm to School Works. Washington DC; 2016. http:// www.fns.usda.gov/sites/default/files/ f2s/FactSheet_Research_Shows_F2S_ Works.pdf. Accessed September 26, 2016. 4. United States Department of Agriculture, Food and Nutrition Service. Schools Serving, Kids Eating Healthier School Meals. Farm to School Census; 2015. https://farmtoschoolcensus.fns.usda.gov/ schools-serving-kids-eating-healthierschool-meals. Accessed September 15, 2017. 5. Izumi BT, Alaimo K, Hamm MW. Farm-to-school programs: perspectives of school food service professionals. J Nutr Educ Behav. 2010;42:83-91. 6. Joshi A, Azuma AM, Feenstra G. Do farm-to-school programs make a difference? Findings and future research needs. J Hunger Environ Nutr. 2008;3:229246. 7. Taylor JC, Johnson RK. Farm to school as a strategy to increase children’s fruit and vegetable consumption in the united states: research and recommendations. Nutr Bulletin. 2013;38:70-79. 8. Yoder ABB, Liebhart JL, McCarty DJ, et al. Farm to elementary school programming increases access to fruits and vegetables and increases their consumption among those with low intake. J Nutr Educ Behav. 2014;46:341349. 9. Yoder ABB, Foecke LL, Schoeller DA. Factors affecting fruit and vegetable school lunch waste in wisconsin elementary schools participating in farm to school programmes. Public Health Nutr. 2015;18:2855-2863. 10. Jones SJ, Childers C, Weaver AT, Ball J. SC farm-to-school programs encourages children to consume vegetables. J Hunger Environ Nutr. 2015;10:511-525. doi:10.1080/19320248.2015.1007259. 11. United States Department of Education. Improving Basic Programs Operated by Local Educational Agencies (Title I, Part A); 2015. http://www2.ed.gov/ programs/titleiparta/index.html. Accessed September 26, 2016. 12. Robinson-O’Brien R, BurgessChampoux T, Haines J, Hannan PJ, Neumark-Sztainer D. Associations between school meals offered through the national school lunch program and the school breakfast program and fruit
ARTICLE IN PRESS 8
Kropp et al
and vegetable intake among ethnically diverse, low-income children. J Sch Health. 2010;80:487-492. 13. Martin CK, Thomson JL, LeBlanc MM, et al. Children in school cafeterias select foods containing more saturated fat and energy than the institute of medicine recommendations. J Nutr. 2010;140:16531660. doi:10.3945/jn.109.119131. 14. United States Department of Agriculture. National Standards in the National School Lunch and School Breakfast Programs; 2012. http://www.fns.usda .gov/school-meals/nutrition-standardsschool-meals. Accessed March 20, 2016.
Journal of Nutrition Education and Behavior 15. Hanks AS, Wansink B, Just DR. Reliability and accuracy of real-time visualization techniques for measuring school cafeteria tray waste: validating the quarter-waste method. J Acad Nutr Diet. 2014;114:470-474. 16. Martins ML, Cunha LM, Rodrigues SSP, Rocha A. Determination of plate waste in primary school lunches by weighing and visual estimation methods: a validation study. Waste Manag. 2014;34: 1362-1368. 17. United States Department of Agriculture. Food and Nutrition Service. Final Rule Nutrition Standards in the National School Lunch and School Breakfast
•
Volume ■■, Number ■■, 2017
Programs; 2012. https://www.fns .usda.gov/sites/default/files/dietary specs.pdf. Accessed May 15, 2017. 18. Wansink B, Just DR, Payne CR, Klinger MZ. Attractive names sustain increased vegetable intake in schools. Prev Med. 2012;55:330-332. 19. Wansink B, Just DR, Payne CR. Can branding improve school lunches? Arch Pediatr Adolesc Med. 2012;166:967-968. doi:10.1001/archpediatrics.2012.999. Accessed May 17, 2017. 20. Thorndike AN, Riis J, Sonnenberg LM, Levy DE. Traffic-light labels and choice architecture: promoting healthy food choices. Am J Prev Med. 2014;46:143-149.
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
CONFLICT OF INTEREST The authors have not stated any conflicts of interest.
•
Volume ■■, Number ■■, 2017
Kropp et al
8.e1
INTRODUCTION In 2010, the Healthy, Hunger-Free Kids Act (HHFKA) was signed into law with the primary focus of improving nu-
trition for children. As part of this legislation, the US Department of Agriculture (USDA) established a Farm to School (FTS) program to help schools increase the amount of local foods
1
Department of Food and Resource Economics, University of Florida, Gainesville, FL Department of Family, Youth, and Community Sciences, University of Florida, Institute of Food and Agricultural Sciences Extension, Family Nutrition Program, Gainesville, FL 3 Department of Agricultural Education and Communication, Program Development and Evaluation Center, University of Florida, Gainesville, FL 4 Program Planning and Evaluation, University of Florida, Gainesville, FL 5 Family Nutrition Program, University of Florida, Institute of Food and Agricultural Sciences Extension, Gainesville, FL 6 Family and Nutrition Program, Family Youth and Community Sciences Department, University of Florida, Gainesville, FL Conflict of Interest Disclosure: The conflict of interest disclosures can be found online with this article on www.jneb.org. Address for correspondence: Jaclyn D. Kropp, PhD, Department of Food and Resource Economics, University of Florida, PO Box 110240, Gainesville, FL; Phone: (352) 294-7631; Fax: (352) 846-0988; E-mail: [email protected] © 2017 Society for Nutrition Education and Behavior. Published by Elsevier, Inc. All rights reserved. https://doi.org/10.1016/j.jneb.2017.10.005 2
Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
they procure and serve in their cafeterias.1 The FTS programs typically consist of ≥1 of the following activities: local procurement of products served in school cafeterias, handson learning activities such as school gardens, and integrated nutrition education.2 To measure progress of FTS activities, the USDA began conducting an FTS census in 2013. According to the most recent 2015 FTS census, which collected data from the 2013– 2014 school year, more than 42,000 schools in over 5,200 school districts had FTS programs, reaching more than 23.6 million children.3 Approximately 19% of all school districts served at least 1 locally-sourced product daily.2 Although FTS programs help schools meet the updated nutrition standards that resulted from HHFKA, 66% of schools with FTS programs also reported other benefits such as increased participation in the National School Lunch Program (NSLP) and less food waste.4 The FTS programs also support
1
ARTICLE IN PRESS 2
Kropp et al
local economies3 and aim to increase students’ consumption of fruits and vegetables through increased exposure to fresh produce.5 Although FTS programs are expanding at a rapid rate across the country, several prior studies focusing on the effects of FTS programs on promoting healthy eating found mixed results. Joshi et al6 reviewed 11 prior studies on the effects of FTS programs on dietary behaviors and found that 10 reported positive dietary changes; 4 of those studies reported increases in the consumption of fruits and vegetables outside school. However, the studies that were reviewed relied on production records from school cafeterias, selfreported behaviors, and dietary recall, and hence might have been less accurate than studies in which selection and consumption were observed directly. Recent research recommended that FTS programs use plate waste methodologies to measure impacts on fruit and vegetable consumption to decrease misreporting,7 because observing and measuring plate waste is an increasingly common and more accurate method used to investigate the selection and consumption of foods, particularly in school cafeteria settings. Although 3 prior studies used plate waste methods to evaluate their FTS programs,8-10 Yoder et al8 did not include control schools not participating in the FTS program for comparison, Yoder et al9 used only FTS program offerings as a control variable, and Jones et al10 did not compare program effects over time. Therefore, the objective of this study was to examine the effects of serving locally-procured produce as part of the Alachua County Public Schools (ACPS), FL, FTS program on the selection and consumption of fruits and vegetables served as part of the NSLP. This study advances the literature by using preintervention and postintervention plate waste data collected at control schools without FTS activities and treatment schools with FTS programs. Using Mann–Whitney U and proportions tests and differencein-difference regression, the researchers tested the hypotheses that fruit and vegetable selection and consumption would increase at the treatment schools after implementation of the FTS program.
Journal of Nutrition Education and Behavior
METHODS After the researchers obtained approval from the University of Florida’s institutional review board, they collected preintervention data at 6 elementary schools in the ACPS system in October, 2015. Postintervention data were collected in April, 2016 after implementation of the FTS program at the treatment schools. The ACPS system was composed of suburban and rural schools with 22 elementary schools. Data were collected from 3 elementary schools receiving the treatment (FTS program) and 3 control schools (not receiving the FTS program). Although FTS programs frequently consist of local procurement, nutrition education, and school gardens, this analysis focused on procurement of NSLP offerings from local producers. One treatment and 1 control school had school gardens that were operational for >2 years before the study period, but none of the study schools had formal nutrition education programs during the normal school day during the study period. The products grown in the school gardens were for educational purposes only and were not served in the school cafeterias before or during the study period. Treatment schools began receiving FTS products in early November, 2015 shortly after baseline data were collected; however, owing to unseasonably warm weather, regular deliveries did not begin until January, 2016. When FTS products were offered in the cafeterias, these products were promoted using signage, which included the name of the local farm supplying the product. The FTS produce primarily consisted of raw vegetables including leafy greens, cucumbers, and peppers used mainly in NSLP salad offerings. The FTS products were offered approximately 50 days during the study period at each treatment school. All schools in the sample except for 1 of the control schools were Title I schools, which meant that they had a high percentage of children from low-income families.11 Low-income families were more likely to participate in the NSLP and also more likely to have limited access to fresh fruits and vegetables in the home because of the food environment or income constraints.12 During the 2015–2016 school year, only Title I elementary
•
Volume ■■, Number ■■, 2017
schools in the ACPS system had the opportunity to receive FTS products; however, not all Title I elementary schools in the ACPS system participated in the program. Three Title I elementary schools in the ACPS system did not participate because of logistical and distributional issues associated with delivering products to their rural locations; hence, assignment into the FTS program was not random. One of the Title I elementary schools not participating in the FTS program declined to participate in the study; therefore, a non–Title I school was included in the control set of schools (Control 3 in Table 1). Although this school was more affluent (lower proportion of students eligible for free and reducedprice lunch) than the other study schools, the demographic mix was similar (see Table 1). Three days of preintervention and 3 days of postintervention data were collected at each school based on the conclusion of Martin et al13 that measuring plate waste for 3 days was the statistically significant representation of a 5-day week. Data collection occurred on Tuesdays, Wednesdays, and Thursdays to avoid capturing potential weekend effects in which students may have exhibited different selection and consumption behaviors on Fridays and Mondays owing to food insecurity at home over the weekend. Preintervention data collection occurred over a 3-week period in October, 2015, with data collected at 1 treatment and 1 control school each week. The ACPS system set the menu at the district level; thus, data were collected from 1 treatment and 1 control school each week with the same menu. Postintervention data were collected in a similar manner in April, 2016. The district menu followed a 3-week menu cycle; for each school in the study, pre- and postintervention data were collected when the same menu items were offered at the school. Schools that participated in the NSLP were required to offer all 5 meal components (meat/meat alternative, grain, fruit, vegetable, and low-fat milk) each day; for the school to receive federal reimbursement for the meal, a student had to select at least 3 of the 5 offered components. 14 Furthermore, NSLP guidelines required
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
Kropp et al
3
Table 1. Free and Reduced-Price Eligibility and Demographic Information of Students Enrolled in the Study Schools Number of Students
Free/Reduced-Price Eligible
White
Black
Hispanic
Asian
Other
Control 1
501
47%
58%
26%
8%
2%
6%
Control 2
452
58%
50%
33%
9%
1%
7%
Control 3
717
31%
52%
19%
11%
11%
7%
Treatment 1
742
71%
20%
53%
12%
6%
9%
Treatment 2
714
75%
52%
23%
12%
2%
11%
Treatment 3
636
42%
49%
29%
8%
5%
9%
School
Note: School-level demographic information was obtained from the Florida Department of Education and hence these values reflect the demographics of the enrollment. Since not all enrolled students participate in the National School Lunch Program, the demographics of the study participants may be different. Control 3 is the only non-Title I school.
all reimbursable meals to contain either a fruit or vegetable.14 Plate waste data were collected from students in first through fifth grades. Because this study focused on determining the effects of the FTS program on selection, Head Start, prekindergarten, and kindergarten students were excluded from the analysis because at the study schools, these younger students were simply given a lunch tray containing all 5 meal components when they went through the lunch line and therefore did not actually select meal components. There are several methods for collecting plate waste data. The quarterwaste method relies on investigators visually examining each tray after consumption and estimating how much of each food item is wasted. This method was validated in a variety of settings, including school cafeterias.15 Hanks et al15 compared the quarter-waste method, half-waste method, and digital photography method for reliability, accuracy, and cost-effectiveness. The quarterwaste method involves trained data collectors rating waste after consumption on a quarter-serving basis, whereas with the half-waste method data collectors rate waste to the nearest halfserving. The digital photography method involves taking photographs of the plate before and after consumption and calculating plate waste by comparing the pair of images. The authors found that the quarter-waste method had the highest interrater and inter-method reliability, whereas digital photography had the lowest; they also found that digital photography was par-
ticularly inaccurate when packaged foods were prevalent, because investigators cannot always determine from a photo whether a package is empty or full.15 Although Martins et al16 found that visual estimation methods overestimate waste by an average of 8 g compared with weighing food before and after consumption, the weighing method is much less efficient for highvolume observations. Because of its reliability and efficiency, the current study employed the quarter-waste method. For each serving of a particular food item taken, data collectors recorded the portion of a standard serving not eaten on a quarter basis using a scale from 0 to 4; 0 indicated that nothing was wasted (the entire portion was consumed), whereas 4 indicated that the entire portion was wasted (the item was not touched). Before the start of lunch, data collectors weighed 3–5 standard servings of each food item to be served in the cafeteria that day. This allowed the data collectors to become familiar with standard servings of each item. The USDA sets NSLP meal pattern requirements by grade. The requirements and standard serving sizes for kindergarten through fifth grade are the same.17 Because this study used data for first- through fifth-graders, standard servings were the same for all meal observations. Each school differed in the methods used to return trays and collect trash at the end of the lunch periods. Thus, unique data collection protocols were developed for each school to ensure that plate waste data were recorded ac-
curately. At each of the study schools, students sat with their classmates (students with the same teacher sat together in the lunchroom) and trash was collected table by table. Thus, in addition to student-level plate waste data, data collectors recorded students’ teachers’ names to link students’ data to their grade-level information. Data collectors worked in teams of 5 to ensure that plate waste data were gathered for all students participating in the NSLP program without delaying the normal trash collection process at the schools; data collectors collected plate waste data as each table was dismissed from the lunchroom. Each individual NSLP meal component was served in a single-serving container or package. When students went through the lunch line, they made selections by choosing the individual meal components from the line. These empty and partly empty containers were used to determine students’ selections and plate waste. Food item data were categorized into the appropriate NSLP categories: meat/meat alternative, grain, fruit, vegetable, and low-fat milk. Because some students purchased >1 meal/d or >1 fruit or vegetable component, the total daily number of servings selected and consumed in each NSLP category was computed from the plate waste data for each NSLP participant.
DATA ANALYSIS Because selection is a discrete variable, the researchers used proportions tests to determine whether there were
ARTICLE IN PRESS 4
Kropp et al
significant increases in the selection of fruits and vegetables after implementation of the FTS program. Mann– Whitney U tests were used to determine whether there were significant increases in the consumption of fruits and vegetables after the FTS program was implemented. Mann– Whitney U tests were employed because skewness and kurtosis tests for normality indicated that the consumption of fruits and vegetables was not normally distributed (P < .001). In addition, the researchers conducted multivariate regression analyses. Selection was modeled using a logistical regression in which the dependent variable took the value of 1 if the item was selected, and 0 otherwise. Marginal effects were calculated from the estimated logistical regression parameters to determine how the independent variables were related to the probability that a fruit (or vege-
Journal of Nutrition Education and Behavior table) was selected. Consumption was modeled using ordinary least squares in which the dependent variable was the number of servings of fruits (or vegetables) consumed. Because many students did not select the fruit (or vegetable), and thus could not consume the fruit (or vegetable), the data contained numerous 0 observations for the selection and consumption of these items. Hence, analyses for consumption were repeated using tobit models, which model the selection and consumption pieces together. Because assignment to the treatment group was not random, a difference-in-difference approach was used to estimate the treatment effects of the FTS program on the selection and consumption of fruits and vegetables. This approach estimated the effects of the FTS program at treatment schools while controlling for potential changes in the control schools. Grade and data col-
•
Volume ■■, Number ■■, 2017
lector controls were included as covariates in all analyses. Data collector controls were included to control for potential interrater differences. To control for unobserved differences between schools that did not vary over the study period, such as differences in other nutrition interventions, including implementation of behavioral nudges, MyPlate posters in the cafeteria, and the presence of school gardens, analyses were repeated including school-level fixed effects. All analyses were conducted using Stata (version 12, StataCorp LP, College Station, TX; 2011).
RESULTS A total of 11,262 student meal observations were collected and analyzed. Table 2 shows the proportion of NSLP meals selected by students that contained a vegetable or fruit, and the
Table 2. Pre-Intervention and Post-Intervention Selection and Consumption of Vegetables and Fruits by Treatment Group Vegetables
Fruits
Selected
Consumed
Selected
Consumed
0.315
0.161
0.777
0.357
(0.010)
(0.008)
(0.009)
(0.010)
Preintervention Control Treatment Z-Score (Treatment—Control)
0.378
0.167
0.830
0.493
(0.008)
(0.006)
(0.006)
(0.009)
4.630
1.079
4.793
8.324
P < .001
P = .28
P < .001
P < .001
0.399
0.202
0.709
0.300
(0.011)
(0.008)
(0.010)
(0.010)
Postintervention Control Treatment Z-Score (Treatment—Control) Preintervention vs. Intervention Control Z-Score (Post—Pre) Treatment Z-Score (Post—Pre)
0.453
0.215
0.831
0.466
(0.008)
(0.007)
(0.006)
(0.009)
3.970
0.987
10.719
11.399
P < .001
P = .32
P < .001
P < .001
5.458
4.017
−4.880
−5.128
P < .001
P < .001
P < .001
P < .001
6.566
5.438
0.083
−2.698
P < .001
P < .001
(P = .93)
P = .007
Note: Selected indicates the proportion of National School Lunch Program meals selected by students that contained the item (vegetable or fruit). Consumed indicates the average servings consumed by students. Proportion tests were used to test for differences in “selected”; Mann-Whitney U test were used to test for differences in “consumed”. Standard errors are in parentheses. Bold indicates significance at P = .05.
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
•
Volume ■■, Number ■■, 2017
Kropp et al
5
Table 3. Difference-in-Difference Treatment Effects of the Farm to School Program on the Selection and Consumption of Vegetables and Fruits (n = 11,626) Vegetables
Selecteda Consumed
b
Consumed if Selected
c
Fruits
Marginal Effect
Standard Error
P-value
Marginal Effect
Standard Error
P-value
−0.039
0.026
.14
0.036
0.022
.09
0.061
0.020
.002
0.055
0.028
.05
0.034
0.018
.06
0.062
0.021
.003
a Estimated using a logit model. bEstimated using OLS. cEstimated using a tobit model; predicted marginal effects conditional on the selection of the item being positive (yj | xj > 0) are reported. Note: Controls for data collectors and student grade-level are included as covariates.
average number of daily servings of these items consumed by NSLP participants at the treatment and control schools. During both the preintervention and postintervention phases of the study, selection of fruits and vegetables was significantly (P < .001) higher at the treatment schools than at the control schools. Also, during both the preintervention and postintervention phases of the study, consumption of fruits was significantly (P < .001) higher at the treatment schools than at the control schools. Selection and consumption of vegetables was significantly (P < .001) higher during the postintervention phase of the study at both treatment and control schools than during the preintervention phase, whereas consumption of fruits was significantly lower at both treatment (P = .007) and control (P < .001) schools during the postintervention phase of the study relative to the preintervention period. Thus, there is evidence that selection and consumption behaviors differed between treatment and control schools serving the same menu items both before and after introduction of the FTS program. Selection and consumption of fruits and vegetables were significantly different at the treatment and control schools before implementation of the FTS program and the treatment was not randomly assigned; thus, the researchers used difference-in-difference models to estimate the treatment effects of the FTS program. Because it can be difficult to interpret the coefficient estimates of a logistical regression, the marginal effects are presented instead of the coefficient estimates for the models pertaining to
the selection of fruits and vegetables in Table 3. As shown in Table 3, the FTS program did not have a significant effect on the selection of vegetables or fruits. However, the OLS analysis indicated that students at the treatment schools consumed, on average, 0.061 more servings of vegetables (P = .002) after implementation of the FTS program. Although this effect was statistically significant and small, it represents an increase of approximately 37% over the average servings of vegetables consumed at the treatment schools before the FTS program was introduced. Furthermore, the OLS analysis pertaining to fruit consumption indicated that students at the treatment schools consumed 0.055 more servings of fruit, on average (P = .05), after implementation of the FTS program. The result was small, but it represents an increase of approximately 11% over the average consumption of the fruit component at treatment schools before the FTS program was introduced. Part of the reason why the treatment effect on the consumption of vegetables using the OLS estimates was so small is that only 45% of selected NSLP meals at the treatment schools contained a vegetable after the FTS program was implemented. Hence, 55% of the consumption observations used in the analysis from those schools took the value of 0 after the program was implemented. Thus, the analysis was repeated using a tobit model. Coefficient estimates of the tobit model represent the linear effect of the covariates on the uncensored latent variable; hence, Table 3 presents the predicted marginal effects
conditional on yj | xj > 0 or on the selection of the item being positive (consumed if selected). The predicted conditional marginal effects from the tobit analysis indicated that students who selected a fruit at the treatment schools consumed 0.062 more servings, on average (P = .003), after the FTS program was implemented. However, the tobit results indicated that the effect of the FTS program on the consumption of vegetables by students who selected the vegetable was not significant. Table 4 presents the results of the analyses when school-level fixed effects that control for school-level unobserved characteristics that did not vary over the study period, such as behavioral nudges, position of lunch items on the lunch line, school gardens, etc. are included as additional covariates. The OLS and tobit results indicate that NSLP participants at the treatment schools consumed 0.107 (P < .001) and 0.086 (P < .001) more servings of vegetables, on average, following the implementation of the FTS program, respectively. The OLS results represent an increase of approximately 65% over the average consumption of vegetables at the treatment schools before the introduction of the FTS program. The effect of the FTS program on fruit consumption is no longer significant for the OLS (P = .29) or the tobit estimations (P = .09).
DISCUSSION This study contributes to the existing literature on FTS programs by analyzing plate waste data to estimate the effects of a new FTS program, specifically the local procurement
ARTICLE IN PRESS 6
•
Journal of Nutrition Education and Behavior
Kropp et al
Volume ■■, Number ■■, 2017
Table 4. Difference-in-Difference Treatment Effects of the Farm to School Program on the Selection and Consumption of Vegetables and Fruits with School-level Fixed Effects (n = 11,626) Vegetables
Selecteda Consumed
b
Consumed if Selected
c
Fruits
Marginal Effect
Standard Error
P-value
Marginal Effect
Standard Error
P-value
0.026
0.030
.39
0.022
0.020
.27
0.107
0.020
<.001
0.030
0.028
.29
0.086
0.018
<.001
0.040
0.024
.09
a Estimated using a logit model. bEstimated using OLS. cEstimated using a tobit model; predicted marginal effects conditional on the selection of the item being positive (yj | xj > 0) are reported. Note: Controls for data collectors, student grade-level, and school-level fixed effects are included as covariates.
aspect of the program, on the selection and consumption of fruits and vegetables. While several previous studies found that FTS programs had positives impacts on students’ consumption of fruits and vegetables,6,8-10 many of these studies relied on school cafeteria production records, selfreported behaviors, and dietary recall data, and hence, might be less accurate than this study that observed selection and consumption directly in school cafeterias. The current study analyzes the selection of fruits and vegetables using logit models and consumption of fruits and vegetables using both OLS and tobit models. While the OLS analyses provide estimates of the average impact of the FTS program on consumption, these estimates include numerous 0 consumption observations due to the low levels of selection. Thus, tobit models are also included to determine the average impact of the FTS program on consumption conditional on the item (fruit or vegetable) being selected. Jones et al10 found that results can vary when school-level controls are included; hence, all analyses were performed with and without the inclusion of school-level fixed effects. The models without school-level fixed effects indicate the FTS program led to increases in the consumption of fruits and vegetables by students at treatment schools following the implementation of the program. Since the majority of the FTS products offered in the treatment schools were vegetables (mostly leafy greens, cucumbers, and peppers), there is some evidence of spillover effects in which signage promoting the FTS vegetables also impacted non-FTS fruit consump-
tion. However, the models that controlled for unobserved schoollevel characteristics, such as signage in school lunchrooms, other nutrition education interventions, school gardens, or nutrition nudging strategies, indicate that the consumption of vegetables, on average, increased at the treatment schools following the introduction of local procurement, but the program did not impact fruit consumption. These results are more consistent with expectations as the majority of FTS items were vegetables. The differences in the results with and without school-level controls highlight the importance of controlling for different school-level effects that may further influence fruit and vegetable consumption. The results of the current study agree with previous studies that used plate waste methodologies to evaluate FTS programs. Yoder et al8 appears to be the first study to use plate waste methodology to evaluate the impact of FTS programs on fruit and vegetable intake of elementary school children over one year. The researchers compared the effect of years of FTS programming at the schools on combined fruit and vegetable consumption but did not estimate consumption of fruits and vegetables separately, so it is difficult to compare current results to the results of the Yoder et al8 study. However, Yoder et al8 reported an increase in student meals containing fruits and vegetables from baseline to post for all schools, and the percentage of trays with no fruit and vegetable disappearance, which served as a proxy for fruit and vegetable consumption, decreased significantly for schools with previous FTS programming compared
to schools with new FTS programs. Based on this previous finding it is possible that fruit and vegetable consumption will increase in the current study’s treatment schools the longer the schools offer the FTS program. Jones et al10 also used photograph plate waste data to estimate servings of fruits and vegetables consumed. The researchers reported that students in schools with FTS programs consumed more vegetables than students at the control schools. However, when the researchers controlled for school, the difference in consumption was no longer significant. In addition, they found students in schools with FTS programs ate less fruit than students in the control schools, the effect of which decreased when controlling for schoollevel factors. The current study found that students in treatment schools ate more vegetables than the students at the control schools even after controlling for school-level fixed effects. In a multi-year (2010–2013), cross-sectional study of Wisconsin elementary schools participating in FTS programs, Yoder et al5 found locallysourced items were wasted more than conventionally-sourced items. However, Yoder et al 9 used FTS program offerings as a control variable to identify the effects the HHFKA of 2010 and only analyzed data from schools with FTS programs. The current study analyzed the impact of the FTS program on selection of all fruits and vegetables regardless of the source of the specific item, making comparisons difficult. Although impacts of the FTS program on consumption in the current study were modest, consuming more vegetables and fruits at each
ARTICLE IN PRESS Journal of Nutrition Education and Behavior meal throughout the year could result in a large impact over time. The magnitudes of the changes in consumption are similar to the results of other studies that analyzed the impacts of creatively naming and branding lunch items.18,19 Hence, it is unclear if the observed changes in consumption behaviors are due to students’ preferences for the FTS products or simply promotion of these products.
IMPLICATIONS FOR RESEARCH AND PRACTICE Several prior studies have found that nudges such as renaming food items and promoting healthier products in school cafeterias can increase selection18-20; combining these techniques with techniques already used in the district to promote FTS products might increase the selection of FTS products. While the study finds positive impacts on the consumption of vegetables and fruits, the study does not investigate what is driving the observed changes in behaviors. Perhaps these changes are driven by students’ preferences for the FTS products due to the freshness and quality of the FTS products or student nudging by the signage in the school cafeterias promoting the products. However, prior studies found that nudging impacts selection as well as consumption18-20; yet, the lack of significance of the FTS program on selection of fruits and vegetables in this study suggests that nudging may not be driving these results. Further research is necessary to determine what factor or factors are responsible for the observed changes in behaviors. In addition, further research is needed to compare the efficacy and costs of various FTS product promotional strategies to determine the best practices. This study only analyzes the local procurement aspect of FTS program. The effects of local procurement on children’s nutrition behaviors are relatively modest, and future studies need to evaluate the impacts of FTS programs within the larger context of community and school nutrition interventions. Specifically, further studies are necessary to determine the impacts of school gardens and nutrition education components of FTS programs.
•
Volume ■■, Number ■■, 2017
Furthermore, additional research is needed to determine the financial cost of wasted FTS products as well as potential waste reduction due to schools receiving FTS products that are fresher than products received from conventional distributors. Since this study was only conducted in 1 school district in Florida, the results may not be generalizable. More research is needed to determine if other FTS programs have similar effects in other schools throughout the country as the results may depend on the types of products offered. In the study district, the majority of the FTS offerings were vegetables; additional research is needed to determine the effects of the program when local fruits, dairy, and meats are offered. Furthermore, more research is needed to determine why selection and consumption behaviors differed between treatment and control schools before introduction of the FTS program. Possible explanations included differences in the implementation of Smarter Lunchroom strategies, other programs that promote healthy eating such as the USDA’s Fresh Fruit and Vegetable Program, differences in the quality, preparation, and presentation of the produce, and the preferences of the students.
ACKNOWLEDGMENTS This study was supported by the University of Florida/Institute of Food and Agriculture Sciences, Extension Family Nutrition Program funded by the United States Department of Agriculture through the Florida Department of Children and Families (#LF923).
REFERENCES 1. United States Department of Agriculture, Food and Nutrition Service. The Farm to School Census; 2015. https:// farmtoschoolcensus.fns.usda.gov/ about. Accessed September 15, 2017. 2. Ralston K, Beaulieu E, Hyman J, Benson M, Smith M. Daily Access to Local Foods for School Meals: Key Drivers, EIB-168, U.S. Department of Agriculture, Economic Research Service; March 2017. https://www.ers.usda.gov/pub lications/pub-details/?pubid=82944. Accessed August 15, 2017.
Kropp et al
7
3. United States Department of Agriculture, Food and Nutrition Service. Farm to School Program: fact sheet: Research Shows Farm to School Works. Washington DC; 2016. http:// www.fns.usda.gov/sites/default/files/ f2s/FactSheet_Research_Shows_F2S_ Works.pdf. Accessed September 26, 2016. 4. United States Department of Agriculture, Food and Nutrition Service. Schools Serving, Kids Eating Healthier School Meals. Farm to School Census; 2015. https://farmtoschoolcensus.fns.usda.gov/ schools-serving-kids-eating-healthierschool-meals. Accessed September 15, 2017. 5. Izumi BT, Alaimo K, Hamm MW. Farm-to-school programs: perspectives of school food service professionals. J Nutr Educ Behav. 2010;42:83-91. 6. Joshi A, Azuma AM, Feenstra G. Do farm-to-school programs make a difference? Findings and future research needs. J Hunger Environ Nutr. 2008;3:229246. 7. Taylor JC, Johnson RK. Farm to school as a strategy to increase children’s fruit and vegetable consumption in the united states: research and recommendations. Nutr Bulletin. 2013;38:70-79. 8. Yoder ABB, Liebhart JL, McCarty DJ, et al. Farm to elementary school programming increases access to fruits and vegetables and increases their consumption among those with low intake. J Nutr Educ Behav. 2014;46:341349. 9. Yoder ABB, Foecke LL, Schoeller DA. Factors affecting fruit and vegetable school lunch waste in wisconsin elementary schools participating in farm to school programmes. Public Health Nutr. 2015;18:2855-2863. 10. Jones SJ, Childers C, Weaver AT, Ball J. SC farm-to-school programs encourages children to consume vegetables. J Hunger Environ Nutr. 2015;10:511-525. doi:10.1080/19320248.2015.1007259. 11. United States Department of Education. Improving Basic Programs Operated by Local Educational Agencies (Title I, Part A); 2015. http://www2.ed.gov/ programs/titleiparta/index.html. Accessed September 26, 2016. 12. Robinson-O’Brien R, BurgessChampoux T, Haines J, Hannan PJ, Neumark-Sztainer D. Associations between school meals offered through the national school lunch program and the school breakfast program and fruit
ARTICLE IN PRESS 8
Kropp et al
and vegetable intake among ethnically diverse, low-income children. J Sch Health. 2010;80:487-492. 13. Martin CK, Thomson JL, LeBlanc MM, et al. Children in school cafeterias select foods containing more saturated fat and energy than the institute of medicine recommendations. J Nutr. 2010;140:16531660. doi:10.3945/jn.109.119131. 14. United States Department of Agriculture. National Standards in the National School Lunch and School Breakfast Programs; 2012. http://www.fns.usda .gov/school-meals/nutrition-standardsschool-meals. Accessed March 20, 2016.
Journal of Nutrition Education and Behavior 15. Hanks AS, Wansink B, Just DR. Reliability and accuracy of real-time visualization techniques for measuring school cafeteria tray waste: validating the quarter-waste method. J Acad Nutr Diet. 2014;114:470-474. 16. Martins ML, Cunha LM, Rodrigues SSP, Rocha A. Determination of plate waste in primary school lunches by weighing and visual estimation methods: a validation study. Waste Manag. 2014;34: 1362-1368. 17. United States Department of Agriculture. Food and Nutrition Service. Final Rule Nutrition Standards in the National School Lunch and School Breakfast
•
Volume ■■, Number ■■, 2017
Programs; 2012. https://www.fns .usda.gov/sites/default/files/dietary specs.pdf. Accessed May 15, 2017. 18. Wansink B, Just DR, Payne CR, Klinger MZ. Attractive names sustain increased vegetable intake in schools. Prev Med. 2012;55:330-332. 19. Wansink B, Just DR, Payne CR. Can branding improve school lunches? Arch Pediatr Adolesc Med. 2012;166:967-968. doi:10.1001/archpediatrics.2012.999. Accessed May 17, 2017. 20. Thorndike AN, Riis J, Sonnenberg LM, Levy DE. Traffic-light labels and choice architecture: promoting healthy food choices. Am J Prev Med. 2014;46:143-149.
ARTICLE IN PRESS Journal of Nutrition Education and Behavior
CONFLICT OF INTEREST The authors have not stated any conflicts of interest.
•
Volume ■■, Number ■■, 2017
Kropp et al
8.e1