ABSTRACT
This paper examines the relationship between food retail density and municipal solid waste. We test for correlations between the volume of solid waste and the number of food-at-home retailers (e.g., grocery stores) and food-away-from-home retailers (e.g., restaurants) at the county level in the state of Mississippi over 2007–2012. Since food scraps comprise the largest share of post-recycling municipal solid waste in the United States, we control for the overall level of economic activity to account for other sources of solid waste, as well as demographic factors, county, and time effects. We find that increases in food-at-home retailer density are negatively correlated with solid waste volume. Conversely, we find that increases in the number of food-away-from-home retailers lead to more waste. While we do not explicitly investigate the mechanisms, we discuss possible avenues such as increased food access in the case of food at home, and increased portion sizes and substitutability in the case of food away from home.
Notes
1 For example, according to the 2015 Consumer Expenditure Survey (U.S. Bureau of Labor Statistics [BLS], Citation2017), the average U.S. household spent nearly one-third of the budget on housing, followed by transportation (17%) and then food (12.5%). In terms of managing fresh produce stocks, Buzby, Bentley, Padera, Campuzano, and Ammon (Citation2016) estimate average supermarket losses of fresh produce of 8–12%, whereas estimates at the household level range from 28% (Gunders, Citation2012) to 39% Hoover (Citation2017) of all fruits and vegetables discarded by households.
2 We considered an alternative measure of business tax receipts, but these data were only available for a limited number of years within the panel dataset.
3 As noted by Wooldridge (Citation2010, p. 326), a lack of variation in key regressors can lead to imprecise fixed effect (or first differencing) results.
4 The free disposability assumption implies the household can costlessly dispose of food inputs. Most municipalities charge a fixed cost for landfill services which at the very least implies no marginal cost of disposability.
5 The issue of whether or not to use weights based on population size boils down to heteroskedasticity and precision. We follow the advice of Solon, Haider, and Wooldridge (Citation2015): First, we perform an unweighted regression and estimate the squares of predicted residuals . Then, we regress
on a constant and the inverse of county population size in year
,
. Since the coefficient on
is significant, this indicates heteroskedasticity and the need to use a weighting scheme. In all our specifications, this coefficient on inverse population is significant with a
-value less than 0.01. We therefore use Eq. (4) in Solon et al. (Citation2015, p. 307) to produce our weights, which is the feasible generalized least squares estimator. Unweighted results are presented in .
6 Qi and Roe (Citation2016) find a majority of their U.S. respondents agreed that they waste more food when it is purchased on sale or in large quantities, which is likely more common at supercenters. Our results may reflect this notion.
7 All other entities in the food markets sector represented less than 10% of food waste in each city: college/universities, K-12 schools, hospitality, health care, events/recreation facilities, correction facilities, grocers/markets, food wholesalers/distributors, and food manufacturing/processing.