The independent variables unique to the food desert research are whether the fast food restaurant is: cobranded, freestanding, and/or has a playplace. Restaurants that are cobranded, i.e. combined with at least one other brand, are expected to have higher prices. This is because cobranded restaurants often have to pay higher franchise fees and royalty fees than noncobranded units (Abcede 1994). These higher fees likely translate into higher prices. Nonfreestanding restaurants, such as those in malls and airports, are expected to charge higher prices because of their convenient locations. Similarly, fast food restaurants with playplaces are expected to charge higher prices because of the additional entertainment value the playplaces provides. The expected signs of the remaining independent variables were discussed in the overview of prior studies’ findings in the literature review. Methods Using the phone survey and GIS data described in Table 1, descriptive statistics are used to compare fast food price, restaurant characteristics, costs, demographics, and competition variables between food deserts and non-food desert block groups. Mean comparison tests are also used to determine whether there is a statistically significant difference in the prices charged for each of the restaurants’ top-three best-selling items in food deserts compared to non-food deserts. The findings from the descriptive statistics and mean comparison tests are used to inform the multivariate regression analysis. OLS regression techniques are used to analyze the pricing strategies of the fast food restaurant managers in food deserts. The general form of the model being estimated is detailed in Equation 1. (1) 𝑝𝑟𝑖𝑐𝑒1-9 = (home, sales, 𝑖𝑛𝑐, 𝑝𝑜𝑝, 𝑎𝑔𝑒, 𝑎𝑓𝑟, h𝑖𝑠, 𝑓𝑟𝑒𝑒, 𝑐𝑜, 𝑓𝑟𝑎𝑛, play, ffres, 𝑠𝑑𝑟e𝑠,, city, fdes) 3 . The models dependent variable is the price of the fast food item. A total of nine regressions are estimated, one for each of the prices of the top three selling items at McDonald’s, , j & w kitchen , restaurant cleaning services
Burger King, steam table , pizza crust recipe and Subway. Results Descriptive statistics, mean comparison tests, and regression results are presented for each fast food firm separately in order to determine if they use different pricing strategies in food desert versus non-food desert markets. McDonald’s Descriptive statistics of the data collected on McDonald’s restaurants in both the non-food desert and food desert block groups are shown in Table 2. The McDonald’s restaurants are relatively 3 Independent variable abbreviations are detailed in Table 1. Leschewski and Weatherspoon Volume17 Special Issue A, 2014 2014 International Food and Agribusiness Management Association (IFAMA). All rights reserved. 154 evenly dispersed among food desert and non-food desert block groups, with 37 restaurants located in food deserts and 30 restaurants located in non-food deserts. Table 2. McDonalds Descriptive Statistics Source. 2010 US Census and Original Price Survey The descriptive statistics for the price variables suggests that there is price variation among McDonald’s restaurants located in food deserts versus non-food deserts. For McNuggets, the mean price is lower in food desert restaurants than in non-food desert restaurants, with mean prices of $3.44 and $3.62 and a standard deviation of $.44 and $0.19 respectively. The mean prices for a Big Mac and Medium Fries are the same for both food desert and non-food desert locations. However, Big Mac price and Medium Fries price have standard deviations that differ between food desert and non-food desert locations. The price of a Big Mac has a mean of $3.66 with standard deviations of $0.27 in food deserts and $0.23 in non-food deserts. Similarly, Medium Fries has a me Leschewski and Weatherspoon Volume17 Special Issue A, 2014 2014 International Food and Agribusiness Management Association (IFAMA). All rights reserved. 155 Table 3. Mean Comparison of McDonald's Prices by Product in Food Deserts vs Non-Food Deserts Big Mac Price Observations Mean Std. Err. Std. Dev. Food Desert 37 3.66 0.04 0.27 Non-Food Desert 30 3.66 0.04 0.23 Combined 67 3.66 0.03 0.25 t-statistic: 0.00 Pr(|T|>|t|): 1.00 McNuggets Price Food Desert 37 3.44 0.07 0.44 Non-Food Desert 30 3.62 0.03 0.19 Combined 67 3.52 0.04 0.36 t-statistic: 2.21 Pr(|T|>|t|): .03 Medium Fries Price Food Desert 37 1.63 0.02 0.12 Non-Food Desert 30 1.63 0.03 0.15 Combined 67 1.63 0.02 0.13 t-statistic: -.12 Pr(|T|>|t|): .91 Source. 2010 US Census and Original Price Survey Mean comparison tests of Big Mac, McNuggets, and Medium Fries prices, shown in Table 3 confirm that only the price of McNuggets varies between food deserts and non-food deserts. The mean comparison test for McNuggets price has a t-statistic of 2.21, implying that McNuggets price differs between food deserts and non-food deserts at the 5% significance level. Although there is price variation for McNuggets, Table 2 shows that all McDonald’s restaurants offer similar amenities, have similar ownership structure, and have similar business approaches. Of the 37 McDonald’s restaurants located in food deserts (non-food deserts), 75% (77%) are franchised, 5% (6%) are cobranded, 95% (90%) are freestanding, and 30% (20%) have a playplace. In contrast, as expected the cost and demographic variables show differences between the two types of locations. Considering the cost variables, the median home price is lower in food desert block groups at $70,528 compared to $110,124 in non-food desert block groups. Inversely, the sales volume at McDonald’s restaurants in food desert block groups is higher at $2,127,000 compared to those in non-food desert block groups, $1,948,000 in 2010. The demographic variables show that per capita income, population, and median age are lower for McDonald’s in food desert block groups versus non-food desert block groups. Conversely, the proportion of the population that is either African-American or Hispanic is higher in food desert block groups compared to non-food desert block groups. The competition variables reveal major differences in the market environment as well with there being more supermarkets, other fast food restaurants, and sit-down restaurants in non-food desert block groups compared to food desert block groups. This is consistent with the definition of a food desert. Leschewski and Weatherspoon Volume17 Special Issue A, 2014 2014 International Food and Agribusiness Management Association (IFAMA). All rights reserved. 156 Table 4. McDonald's OLS Regression Results by Product (N=67) Independent Variables Big Mac Price Chicken Nuggets Price Medium Fry Price Location Food Desert 0.118* -0.066 -0.017 Restaurant Characteristics Franchise 0.028 -0.087 -0.079 Cobranded -0.097 0.183 0.007 Playplace -0.078 -0.003 0.057* Costs Sales Volume -0.001** -0.000 0.000 Demographics Per Capita Income 4.02e-06 8.68e-06 6.06e-06* Population -0.000 -0.000 -0.000 Median Age 0.008* 0.011** -0.001 African-American .042 0.122 0.167** Hispanic .018 0.570 0.358** Competition Other Fast Food .017 0.026 0.013** Sit-Down Restaurants .003 0.006 -0.008** City Dummies Lansing .181* 0.357** -0.046 Flint -0.240*** 0.006 -0.071* Grand Rapids 0.032 0.109 -0.122* Dearborn -0.249* 0.014 0.172* Warren 0.236* 0.100 0.005 Livonia 0..246* 0.282 0.007 R-Squared 0.544 0.309 0.461 Significant at the 90% level, ** Significant at the 95% level, *** Significant at the 99% level The ordinary least squares estimates for Equation 1 with the prices of McDonald’s three most popular food items as dependent variables are shown in Table 4. Variance inflation factors were calculated in order to determine if a multicollinearity problem was present. Three of the independent variables (freestanding, median home price, and supermarkets) had variance inflation factors greater than 10. These variables were removed from the model as they were indicative of a multicollinearity problem (Greene 2003). The Breusch-Pagan test for heteroskedasticity was then used to determine if heteroskedasticity was present in each of the three price models. The test results showed that heteroskedasticity was present in the regression with McNuggets price and Medium Fry price as dependent variables. Robust standard errors were then calculated for all three regressions in order to correct for the presence of heteroskedasticity. The r-squared values of 0.544, 0.309, and 0.461 for the Big Mac price, McNuggets price, and Medium Fries price regressions respectively, show that a significant proportion of the price variation is explained by the independent variables.