Literature Review

Previous research has analyzed many aspects of the US sugar market such as sugar-sweetened beverage taxes (e.g., Cawley et al., Reference Cawley, Frisvold, Jones and Lensing2021; Lee et al., Reference Lee, Falbe, Schillinger, Basu, McCulloch and Madsen2019; Allcott, Lockwood, and Taubinsky, Reference Allcott, Lockwood and Taubinsky2019), the impact of NAFTA (e.g., Kennedy and Schmitz, Reference Kennedy and Schmitz2009; Knutson, Westhoff, and Sherwell, Reference Knutson, Westhoff and Sherwell2010; Schmitz and Lewis, Reference Schmitz and Lewis2015), the effect of Mexican sugar dumping and countervailing actions on the US sugar market (e.g., Schmitz, Reference Schmitz2018; Lewis and Schmitz, Reference Lewis and Schmitz2015; Zahniser et al., Reference Zahniser, Kennedy, Nigatu and McConnell2016), US genetically modified (GM) sugar beets (e.g., Lewis et al., Reference Lewis, Grebitus and Nayga2016a; Kennedy, Schmitz, and DeLong, Reference Kennedy, Schmitz and DeLong2019; Carter and Schaefer, Reference Carter and Schaefer2019), consumer preferences for sugar (e.g., Lewis, Grebitus, and Nayga, Reference Lewis, Grebitus and Nayga2016a; Lewis, Grebitus, and Nayga, Reference Lewis, Grebitus and Nayga2016b; Lewis, Grebitus, and Nayga, Reference Lewis, Grebitus and Nayga2016c; Lewis and Grebitus, Reference Lewis and Grebitus2016), and how US sugar prices impact sugar-using firm profitability (e.g., Triantis, Reference Triantis2016; Trejo-Pech et al., Reference Trejo-Pech, DeLong, Lambert and Siokos2020).

Cawley et al. (Reference Cawley, Frisvold, Jones and Lensing2021) provide a review of the literature concerning the impact of sugar-sweetened beverage taxes. They also analyzed Nielsen Retail Scanner Data along with hand-collected price data on sugar-sweetened beverages to determine the passthrough effect of a sugar-sweetened beverage tax in Boulder, Colorado. They found that the passthrough effect of the tax was slightly less than 75% when analyzing hand-collected store data, but was 50% when using the scanner data. In general, they found consumers paid most of the tax.

Schmitz and Lewis (Reference Schmitz and Lewis2015) found that NAFTA cost US sugar producers $474 million to $1.3 billion annually due to increased Mexican sugar imports into the US. Lewis and Schmitz (Reference Lewis and Schmitz2015) estimated that if the Mexican government no longer owned and operated 20% of their sugar industry, US sugar producers would gain $525 million to $1.6 billion annually. Schmitz (Reference Schmitz2018) observed that the 2014 Mexican–US sugar suspension agreements would have saved US sugar producers $138 million annually if these agreements had been in effect from 2008 through 2014.

Kennedy, Schmitz, and DeLong (Reference Kennedy, Schmitz and DeLong2019) used a partial equilibrium framework to evaluate the possible demand impact of US sugar beets being GM compared to their non-GM sugarcane counterpart. They found that the negative demand impact from sugar beets being GM could outweigh any supply induced gains from GM sugar beets. They concluded that since the National Bioengineered Food Disclosure Standard does not classify beet sugar as bioengineered, this negative welfare result would not be fully realized for sugar beet producers.

Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020) examined the effect of US sugar policy on sugar-using-firm profitability when considering both US and world sugar prices. They examined how the relative US-to-world sugar price ratio impacted US firms’ return on assets (ROAs) and market-to-book (MTB) ratio. Counterintuitively, they found that as US sugar prices increase relative to world sugar prices, US sugar-using firms became more profitable.

While research has examined several aspects of the US sugar market and US policy, no known research has examined factors associated with SCP prices and whether US sugar prices are associated with SCPs. To contribute to the economics of sugar literature, this research analyzes a panel of SCP prices across various sugar-using firms and examines whether US sugar prices are associated with SCP prices.

Conceptual Framework and Hypothesized Results

The goal of this paper is to determine factors influencing SCP prices with a particular interest on understanding the role of input sugar prices. Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020) was the first known research to examine the relationship between US sugar prices and sugar-using firm performance as measured by firms’ ROAs and MTB ratios. It was found that as the ratio of US wholesale refined beet sugar price to the world refined sugar (US-to-world) price increased, firms’ ROAs increased. However, in their second model specification, there was no statistically significant relationship between the US-to-world sugar price ratio and firms’ MTB ratios.

Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020, page 14) concluded their research with a possible explanation of, “US sugar-using firms must pass on higher costs to consumers when relative prices increase or do not pass on discounts to consumers when relative sugar prices decrease.” We examine this outstanding question using a unique data set of SCP prices as a dependent variable, rather than using firm profits (ROA and MTB ratios) as a dependent variable like Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020), and by including sugar prices as an independent variable. Thus, this research will directly estimate whether higher US sugar prices are passed on to consumers while controlling for other relevant possible determinants of food prices.

Previous literature has examined determinants of food prices (e.g., Richards and Hamilton, Reference Richards and Hamilton2015; Volpe, Risch, and Boland, Reference Volpe, Risch and Boland2017; Hamilton, Liaukonyte, and Richards, Reference Hamilton, Liaukonyte and Richards2020; Nakamura, Reference Nakamura2008; Ghazaryan et al., Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018; Nakamura and Zerom, Reference Nakamura and Zerom2010). Factors theorized to affect food prices include input prices (Nakamura and Zerom, Reference Nakamura and Zerom2010; Volpe, Risch, and Boland, Reference Volpe, Risch and Boland2017; Richards and Hamilton, Reference Richards and Hamilton2015; Hamilton, Liaukonyte, and Richards, Reference Hamilton, Liaukonyte and Richards2020), product characteristics such as size and appearance (Ghazaryan et al., Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018), macroeconomic factors (Richards and Hamilton, Reference Richards and Hamilton2015; Volpe, Risch, and Boland, Reference Volpe, Risch and Boland2017), and market power (Hamilton, Liaukonyte, and Richards, Reference Hamilton, Liaukonyte and Richards2020; Richards, Hamilton, and Yonezawa, Reference Richards, Hamilton and Yonezawa2017). Similarly, it is hypothesized that input prices, product characteristics, macroeconomic factors, and firm size will have an impact on SCP prices. While no located research has examined how firm financial performance affects food prices, this research postulates it will affect SCP prices since Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020) found a relationship between firm profitability measures and sugar prices, and because firms’ market power throughout the food supply chain is an important consideration in explaining food prices (Hamilton, Liaukonyte, and Richards, Reference Hamilton, Liaukonyte and Richards2020; Richards, Hamilton, and Yonezawa, Reference Richards, Hamilton and Yonezawa2017). Only SCP prices from publicly traded sugar-using firms were utilized so firm size and financial performance variables could be included as independent variables in the analysis.

The categories of variables hypothesized to affect SCP prices are SCP characteristics—including sugar price and other characteristics—firm size and financial performance, and macroeconomic conditions. A traditional hedonic pricing model is estimated such that for each uniquely identified SCP, i, in a calendar quarter, t, the SCP price per pound (SCPprice _i,t ) can be explained as a function of the following factors:

(1) $$SCPpric{e_{i,t}} = f(SC{P_{i,t}},{F_{i,t}},M{E_{i,t}}).$$

where SCP _i,t are characteristics relating to the SCP, F _i,t are variables relating to the respective SCP firm, and ME _i,t controls for macroeconomic conditions over the sample’s time period.Footnote ²

Sugar-Containing-Product Characteristics

Input prices (e.g., Volpe, Risch, and Boland, Reference Volpe, Risch and Boland2017; Richards and Hamilton, Reference Richards and Hamilton2015; Hamilton, Liaukonyte, and Richards, Reference Hamilton, Liaukonyte and Richards2020) and product characteristics (e.g., Ghazaryan et al., Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018) are primary determinants of food prices. It is expected that the SCP characteristics that will impact the SCP price per pound are sugar prices (SugarPrice), the percentage of sugar in the SCP (%Sugar), and SCP weight (Weight).

To test the effect of sugar prices on SCP prices, four specifications of sugar prices are examined across four different models: (1) US wholesale refined beet sugar prices, (2) quarterly lagged US wholesale refined beet sugar prices, (3) US-to-world sugar prices, and (4) US-to-world quarterly lagged sugar prices. With respect to all four specifications, it is hypothesized that the sugar prices will either not be significant in the model or else will be significant and positive (Table 1). If sugar prices are significant and positive, this would be consistent with Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020) and the SUA’s (2021) declaration that higher sugar prices caused by US sugar policy are “unfairly” passed on to consumers. On the other hand, since US sugar prices represent such a small fraction of SCP prices (Triantis, Reference Triantis2016; ASA, 2021), it is also possible that they do not have an impact on SCP prices and are not statistically significant in the model. This argument is also consistent with research documenting a weak transmission of coffee bean prices to consumer prices due to a relatively low share of coffee bean cost relative to the overall cost of manufacturing coffee (Bettendorf and Verboven, Reference Bettendorf and Verboven2000).

Table 1.

Variables, definitions, and hypothesized signs

^a EPSthresholdt = 1 if EPS_t−1/EPS_t−2 ≤ 1.01, 0 otherwise.

^b Firm market value size, or firm market valuation, is defined as total assets − common equity + market capitalization, where market capitalization equals the stock price multiplied by the number of shares outstanding during the fiscal quarter.

^c Prices were adjusted for inflation using the Consumer Price Index.

Notes: NS stands for not significant. NA stands for not applicable.

Table 2.

Summary statistics of sugar-containing-product (SCP) prices, SCP characteristics, firm financial performance, and recessions

Notes: Observations = 379.

Similar to Ghazaryan et al. (Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018), food product characteristics are also included in the model. The estimated coefficient on the percentage of the product’s weight that is sugar (%Sugar) will indicate whether products of a higher sugar content are priced higher than products that have a lower sugar content. The %Sugar is also a control variable given there is a diverse panel of SCPs considered in the analysis. It is uncertain whether more highly sweetened products are priced higher or lower (Table 1).

Product weight has been shown to affect pricing decisions of foods (Ghazaryan et al., Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018). To incentivize sales volume, manufacturers may increase a product’s package size, which decreases the SCP price per unit, and in consequence, makes their products economically more attractive. It has also been suggested that sugar-using firms may increase price per unit by repacking candy into smaller bars sold at similar prices (Trejo-Pech et al., Reference Trejo-Pech, DeLong, Lambert and Siokos2020), a phenomenon known as “shrinkflation” (National Public Radio, 2021). Decreasing the size of the candy bars without decreasing their price proportionally would have the impact of increasing the SCP price per pound. From a more general standpoint, product Weight is also included as a control variable since the data are a panel of SCPs ranging from candy bars to sucrose-sweetened soft drinks and it is likely SCP prices vary by product types which also have different weights. Given these relationships between SCP weight and price per pound, we expect that as the package size increases (decreases), the SCP price per pound will decrease (increase) (Table 1).

Econometric Model

A fixed-effects panel regression was estimated to determine factors influencing SCP prices. The panel consists of a sample of SCPs sold by sugar-using publicly traded firms over time. The dependent variable is SCP price per pound (SCPprice) expressed in 2020 Consumer Price Index (CPI) adjusted dollars (U.S. Bureau of Labor Statistics, 2021). This was calculated by dividing SCP CPI adjusted prices by the total pounds in each product’s package. Expressing SCP price on a per pound basis standardizes prices across product categories and uses the same units for both output and input prices (i.e., sugar prices are by industry convention expressed in $ per pound). This model specification is similar to Ghazaryan et al. (Reference Ghazaryan, Westgren, Parcell and Gedikoglu2018) who estimated a hedonic pricing model with produce prices as the dependent variable and included independent variables related to produce characteristics.

SCP price is modeled as a function of SCP characteristics, firm size, firm past financial performance, and macroeconomic conditions. Equation(2) shows the model:

(2) $$\eqalign{\ln (SCPprice)_{i,t}^{SCP} = {\beta _0} + {\beta _1}\cdot\ln (Weight)_{i,t}^{SCP} + {\beta _2}\cdot\ln (\% Sugar)_{i,t}^{SCP} \cr \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\, + {\beta _3}\cdot\ln (SugarPrice)_{i,t}^{SCP} + {\beta _4}\cdot SalesGrowth_{i,t - 1}^F + {\beta _5}\cdot EPSthreashold_{i,t - 1}^F \cr \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\, + {\beta _6}\cdot LargeMarketVal_{i,t}^F + {\beta _7}\cdot\ln (DispIncome)_{i,t}^{ME} + {\beta _7}\cdot R_{i,t}^{ME} + {a_i} + {e_{i,t}} \cr} $$

where subscript i represents the uniquely identified SCP, t is a calendar quarter, and superscripts represent characteristics associated with the SCP (SCP), SCP firm (F), and macroeconomic conditions (ME). SCP characteristics include SCP weight (Weight), the percentage of sugar in the SCP (%Sugar), and the four considered sugar prices (SugarPrice) which were adjusted for inflation using the CPI. To facilitate the interpretation of results (i.e., have parameters estimated as elasticities), the natural log (ln) of the SCP characteristic variables and disposable income was used in equation(2).

SCP firm size and financial performance variables include sales growth (SalesGrowth), the EPS threshold proxy (EPSthreshold), and the variable regarding the size of the firm relative to other firms in the sample (LargeMarketVal). The macroeconomic variables of per capita disposable income (DispIncome) and recessions (Recessions) were also included. DispIncome was adjusted for inflation using the CPI and the natural log of DispIncome was used in the model. Complete variable descriptions appear in Table 1. a _i in equation(2) captures product-specific fixed effects.

From equation(2), the null hypothesis of whether sugar prices affect retail SCP prices $({\beta _3} = 0)$ can be tested. If the null is failed to be rejected, then sugar prices do not have a relationship with SCP prices. Meanwhile, if the null hypothesis is rejected $({\beta _3} \ne 0)$ , then sugar prices do affect SCP prices. Equation(2) is estimated under four specifications of SugarPrice to test the null hypothesis. In the first specification, SugarPrice is the quarterly sugar price, and in the second specification, SugarPrice is the one-quarter lagged sugar price, relative to the time-period SCP prices were collected. This was done to determine whether SCP prices are based on current sugar prices or previous (lagged) sugar prices. If sugar prices do affect SCP prices, it is likely they do not instantly change.Footnote ⁴ Thus, it is suspected that any relationship between sugar prices and SCP prices would be lagged by one-quarter. To keep the notation simple, equation(2) shows only the current price specification.

In the third sugar price model specification, SugarPrice is the US-to-world sugar price ratio and in the fourth specification it is the quarterly lagged US-to-world price ratio. This follows a similar approach to Trejo-Pech et al. (Reference Trejo-Pech, DeLong, Lambert and Siokos2020). The sugar relative price coefficient, β ₃, is expected to be either positive or non-significant. A positive coefficient will imply that US sugar-using firms will increase prices whenever they find themselves in a competitive disadvantage relative to firms that can obtain sugar at world prices. This may occur in the following instances on which the US-to-world price ratios increase: when both US and world sugar prices increase but the US sugar price increases at a higher rate, when both US and world sugar prices decrease with US sugar prices decreasing at a lower rate, and in general when any combination of US and world sugar price changes increase the US to world sugar price ratio. Alternatively, the US-to-world sugar price coefficient will be zero if US sugar prices represent such a small percentage of SCP prices that US sugar firms do not make SCP pricing decisions based on sugar input price changes, regardless the relationship between US and sugar world prices.

To estimate the panel fixed-effects regression in equation(2), the xtreg command in Stata with the fe specification was used (StataCorp, 2017). The regression was estimated accounting for heteroscedasticity and contemporaneous correlation of the error terms through the vce(robust) command in Stata. This specification was used since the results of the xttest3 command (Baum, Reference Baum2001) provided a rejection of the test’s null hypothesis which means the error terms exhibited groupwise heteroscedasticity. The vce(robust) command results in the Huber/White/sandwich variance–covariance estimation (vce) estimator to be calculated for the regression coefficients (StataCorp, 2021). The coldiag2 command in Stata was used to determine if multicollinearity was an issue. If the estimated condition indexes for the variables are less than 30, the regression is considered free of collinearity issues (Belsley, Reference Belsley1991).

Data

Starting in 2008, the ASA created an annual database of SCPs. To qualify as a SCP, the product had to contain sugar (sucrose) as a primary ingredient (be at least the fifth listed ingredient on the product’s label) and the product had to contain at least 10% sugar by product weight. Sugar in this analysis is limited to sucrose; for example, soft drinks in this data set include only those made with real sugar (sucrose) and do not include those sweetened with high fructose corn syrup (HFCS). Annually (typically in June), the ASA collects data from grocery stores in the metropolitan Washington, D.C. area on a variety of SCPs. They collect the SCP prices as well as information from the respective products such as the product’s net weight and information from the nutritional facts panel such as serving size, servings per container, and grams of sugar per serving. Previous research has also used hand-collected data to estimate the passthrough effects of taxes on sugar-sweetened beverages (Cawley et al., Reference Cawley, Frisvold, Jones and Lensing2021).

For this analysis, only publicly traded companies with available financial accounting and stock price data in the Wharton Research Data Services (WRDS) database were analyzed since firm financial performance was controlled for in the analysis. Firm financial data were obtained specifically from WRDS’s Compustat Capital IQ Fundamentals Quarterly module, with the following items used to calculate financial variables in the model: total assets, total revenue, total equity, stock price, number of shares outstanding, and EPS. Thus, there is a panel of data by firm and SCP type ranging from 2008 through 2020 with the exception of 2011 when the ASA did not collect data.

The complete list of SCPs in the analysis, by firm and by year, appears in Appendix A. The firms in the analysis, and examples of their SCPs in the panel, are the following: Tootsie Roll (Junior Mints), The Hershey Company (e.g., Milk Duds, Hershey Bar, Hershey Bar with Almonds), Unilever (e.g., Klondike bars, Ben and Jerrys ice cream), The J.M. Smucker Company (e.g., Pillsbury cake and frostings), Pinnacle Foods and Conagra (e.g., Duncan Hines frosting and cake mixes), PepsiCo, Inc. (e.g., Pepsi Throwback, Quaker corn bread), Mondelez (e.g., Oreos), Kraft (Jello-O), and General Mills (e.g., Betty Crocker cake mixes and frostings). Appendix A also shows summary statistics by SCP firm (Table A1).

Summary statistics for the variables used in equation(2) appear in Table 2. There are a total of 379 observations in the sample. SCP prices were collected directly from the SCPs appearing in the grocery stores and, for this study, were adjusted for inflation using the CPI series published by the U.S. Bureau of Labor Statistics (2021) with August 2020 as the base year (i.e., the last month SCP prices were collected in the database). SCP price was then standardized by dividing it by the SCP’s net weight (in pounds). This is the dependent variable in the analysis, SCPprice. The average SCP price per pound (SCPprice) was $6.22/pound with a minimum price of $0.55/pound (Pepsi Throwback and Mountain Dew Throwback) and a maximum price of $17.95/pound (Hershey York Peppermint Pattie) (Table 2).

Sugar-Containing-Product Characteristics

The model was estimated using four SugarPrice specifications as shown in Table 1. The US wholesale refined beet sugar prices were considered for analysis, and they were adjusted for inflation using the CPI.Footnote ⁵ The US wholesale refined beet sugar price was utilized since this is the industry conventional benchmark for wholesale refined sugar dating back to 1960 (USDA ERS, 2021b). To construct the US-to-world sugar price ratio, the US wholesale refined beet sugar price was divided by the world wholesale refined sugar price.Footnote ⁶ Both prices were adjusted for inflation using the CPI.

Equation(2) was estimated using four SugarPrice specifications (i.e., US contemporaneous and lagged sugar prices, and US-to-world contemporaneous and lagged price ratios). Both wholesale refined beet sugar prices averaged around 33 cents/pound and ranged from around 23 cents/pound to 44 cents/pound. Both US-to-world sugar price ratios averaged around 1.8 with the lowest being 1.2 and the highest being 2.7. Figure 1 shows the annual relationship of wholesale refined beet sugar prices per pound and SCP prices per pound as well as the percentage of sugar in each product. As shown by Figure 1, sugar prices represent a small percentage of the SCP price.

Figure 1.

Annual SCP prices, sugar prices, and sugar share of SCP weight.

The percentage of sugar (%Sugar) in each SCP was calculated from the weight of sugar in the product provided on the nutritional facts panel (in pounds) divided by the SCP weight provided on the package label (in pounds). The average percentage of sugar in SCPs was 48%, the minimum was 12% (Pepsi Throwback), and the maximum was 99% (Tootsie Roll Junior Mints). The SCP weight (Weight) is equal to the net weight of the SCP, which is found on the package of the SCP. The average weight was 1.17 pounds with the minimum SCP weight being 0.086 pounds (General Mills Betty Crocker muffin and quick bread mix-cinnamon streusel) and the maximum SCP weight being 9 pounds (Pepsi Throwback and Mountain Dew Throwback).

Sugar costs in the package per pound (SugarCostPackage) were also computed by multiplying %Sugar by the quarterly SugarPrice. Footnote ⁷ The average cost of sugar across all SCPs was 15.9 cents/pound, with a maximum of 44 cents/pound (Tootsie Roll Junior Mints) and a minimum of 2.7 cents/pound (Pepsi Throwback) (Table 2). Meanwhile, the average SCP price is $6.22/pound. Thus, the percentage of the cost of sugar across SCPs is only 2.56% of the total SCP price despite the fact that sugar consists of 48% of the SCP’s weight. Figure 2 shows this relationship from 2008 to 2020.

Figure 2.

Cost of sugar in the sugar-containing-product (SCP) versus the SCP price.