1. Introduction
Although product packaging plays important strategic and tactical roles in marketing [
1,
2], recent publicity about waste derived from packaging has not been favorable. Anthropogenic debris is becoming a larger threat to ocean ecosystems as it is now listed as one of the main causes for the non-natural death of sea animals [
3]. Social network services reveal this new threat through vivid photos of turtles with plastic straws up their nostrils and ever-increasing landfill space used for municipal solid waste.
It is disturbing to bear those phenomena with regard to the magnitude; the human population produces 348 million tons of plastic each year [
4]. Making things worse for high-income and waste-exporting countries, China has banned the import of plastic waste since 2017. The government-level decision is particularly impactful since China imported 45% of the planet’s recyclables [
5]. The aftermath of China’s decision triggered public policy implementations, such as charging a small fee and introducing partial or complete bans for single-use bags in some countries in Europe [
6] and more awareness and behavioral response in the U.S. Notwithstanding this megatrend across the globe, the academic research on eco-friendly consumption in general has been scarce [
7,
8] and the financial efficacy and the mechanisms of eco-packaging sales are largely unknown [
9].
Even if executed, the previous studies excessively investigate eco-packaging in the context of the European market [
7,
10,
11] and firm-level investment sentiments [
12,
13], while the consumer-level purchasing behavior and the research context diversity are important. A number of authors call for more attention on and scrutiny of, including but not limited to, eco-packaging and market reactions, including its impacts on actual product choice [
7,
14] and testing a diverse range of products using a larger sample [
10], such as a frequently purchased product category [
8].
To extend and enrich the prior body of sustainability research focusing on eco-friendly packaging, we will first investigate the actual consumer behavioral reactions to the eco-packaging appeals in individual product-level sales in the context of a U.S. grocery store. A grocery store is selected due to its wide range of product categories and the scanner data availability from which actual purchase behavior can be evaluated. Secondly, we will address whether eco-friendly packaging positively impacts the brand-level sales (brand spillover effects). Lastly, we will respond to the long-standing debate [
9] on the mechanisms of the eco-packaging return (linear return vs. curvilinear return) on brand-level sales. While most research in the context of curvilinear return apply firm-level financial performance (e.g., return on assets), investment reactions (e.g., stock market performance) [
15], and firm valuation [
16], there is, to the best of our knowledge, no empirical research on the product and brand level simultaneously. In sum, we aim to evaluate and elaborate the impacts of eco-friendly packaging on the most basic level of product hierarchy––identified by the Universal Product Code (UPC)—and its impacts on other regular products within the same brand as well as the rate of return mechanism in a fast-moving consumer goods (FMCG) context.
3. Method
3.1. Research Design
The current study aims to evaluate the suggested hypotheses in the context of FMCG, which have a wide range of product categories. This way, we can reduce the social desirability bias which is one of the pervasive problems in ethical consumption studies using survey data [
30]. That is, respondents feel pressure to answer favorably toward eco-consumption against their true wills or actual behaviors [
11]. The availability of the scanner data is possible due to the collaboration of a large supermarket chain in the Southwest region of the U.S. The parent company owns more than 50 stores with different segmentation and targeting strategies. We specifically selected 10 stores due to their large assortments of eco-packaging appeal among their products.
3.2. Data Description
The original data comprise real-time transactions for a 106-week period in 2009–2010. Since the raw data were collected whenever a transaction occurred, appropriate aggregations were necessary. For instance, the data were aggregated by week, by store, and per UPC for Hypothesis 1. Further aggregations were made for Hypothesis 2 and 3 in terms of the brand level. The analysis includes at least three categories in each brand. A weekly increment allows for an optimal period-to-period comparison of sales since a typical promotion cycle is also weekly. For the particular grocery chain from which we obtained the data, the weekly cycle lasts from Wednesday to the following Tuesday. The first few days before week 1, Wednesday and the last few days after week 106, Tuesday were eliminated and so were other incomplete data with entry errors and unknown errors. As a result, the final data set contains a total of 3,525,537 and 1,434,917 unique identifiers for UPC- and brand-level analysis, respectively.
3.3. Dependent Variable
The dependent variable in the study is sales in United States Dollar (USD), which has two components of total units sold (volume) measured as the count of individual UPCs sold multiplied by unit price of the same UPC. Sales, units sold, and unit price are unique across weeks and stores due to a possible price promotion applied in each time frame and store location. The time frame ranges from 1 to 106 where each number indicates the specific week. Weekly sales are the net sales excluding returns per UPC (Hypothesis 1) and per brand (Hypothesis 2 and 3). For Hypothesis 2 and 3, we eliminated the product categories with less than three in each brand for the sake of comparison. For example, if a brand has two product categories and one of them has eco-packaging appeal, the eco-packaging ratio of the brand becomes 50%, preventing comparison with a brand with many product categories.
3.4. Independent Variables
Eco-packaging. A dummy variable is created, indicating that either a certain product has a feature of eco-packaging directly and explicitly in its outer packaging (1 = eco-packaging, 0 = otherwise). To identify a product’s sustainability and eco-friendliness, we followed the guideline from the Sustainable Packaging Coalition [
45], which emphasizes the source of the materials that were used during the manufacturing process. Since neither the manufacturer nor the retailer maintains the eco-packaging as product information, one of the authors manually classified and identified the eco-packaging using a wireless UPC barcode scanner.
3.5. Control Variables
Featured advertisement. This variable indicates that a certain UPC is featured in a store-generated (yet often funded by the manufacturers) weekly flyer during the given week (1 = product item is featured in a flyer, 0 = otherwise). Independent of the eco-packaging or regular product, the effect of feature advertisement measures the financial efficacy of store-generated printed advertisements.
Price promotion and deletion. These variables refer to UPCs that have a temporary price deduction (price promotion) or a price deduction followed by removal from the shelf (deletion). Here, two price promotion variables are also treated as binary.
Holiday. A total of six holidays are included: Thanksgiving, Christmas, Independence Day, Labor Day, Columbus Day, and Martin Luther King Day. A typical model specification for the holiday effect is the indicator variable, “1” during the week of the holiday and “0” otherwise. Each indicator variable is repeated twice since the duration of data collection is two years.
3.6. Model Specification
We employ a fixed effect model to control for store and holiday effects. We control error correlation across weeks in each product to get unbiased estimates with clustered standard errors which often appear in large-scale scanner data [
30]. We cluster standard errors by UPC product for Hypothesis 1 and brand for Hypothesis 2 and 3. Equation (1) represents the model specification for the effects of eco-packaging on product sales (Hypothesis 1):
where Y = UPC sales in USD, Eco = eco-packaging (dummy),
i = the unique UPC,
j = store, and
t = week. The parameter
is the efficacy of eco-packaging on UPC-level sales. The positive and significant coefficient
indicates that eco-packaging indeed plays a role in consumer choice, supporting Hypothesis 1. Equation (2) aims to test brand spillover effects: the effects of eco-packaging intensity on brand-level sales. Notwithstanding the difficulty in data collection and analysis, many studies apply a multi-category analysis for brand spillover effects [
40] since numerous firms embrace both line extensions (within the same product category) and brand extensions (across different product categories). Since a single-product model cannot holistically reflect the current market phenomena, we apply the multi-product category model. With this model, brand spillover effect can be measured by the following model:
in this equation,
is the brand-level sales in USD in product category
i, brand
j, and week
t.
The component
represents the ratio of UPCs with eco-packaging in product category
i, brand
j, and week
t. The coefficient
is related to the brand spillover effect of eco-packaging from the own product category (line extension).
where
i = product category,
j = brand, and
t = week.
, on the other hand, is the average ratio of UPCs with eco-packaging in other product categories, except for
i, in brand
j, and week
t.
where
i = product category,
j = brand, and
t = week.
For this reason, parameter explains the brand spillover effect of eco-packaging from other product categories (brand extension). The interaction term of and estimates the comprehensive brand spillover effect of eco-packaging: simultaneous introductions in line extension and brand extension. If the coefficients , and are positive and significant, Hypothesis 2 is supported: positive brand spillover effect of eco-packaging on brand-level sales. However, one should note that each parameter estimates a different type of brand spillover effect, that is, for brand spillover from line extension, for brand spillover from brand extension, for brand spillover from line and brand extension.
Besides, to test the curvilinearity of brand spillover effect (specifically an inverted U shape) postulated in Hypothesis 3, we form the following model with the linear and the quadratic variable of eco-packaging as follows:
where
= the brand-level sales in USD in product category
i, brand
j, and week
t and the variable
= the ratio of UPCs with eco-packaging in product category
i, brand
j, and week
t. If the coefficient
is negative and significant, Hypothesis 3 will be supported: eco-packaging positively effects the overall sales of a brand with a peak point, but then decreases after.
4. Empirical Results
This section will reveal the statistical results of eco-packaging impacts on UPC-level sales and brand-level sales along with the curvilinearity test. The fixed models at UPC- and brand-level are estimated with a least square regression with robust standard errors for the sake of unbiased estimates of variance and covariance parameter estimation.
Table 1 shows the effects of eco-packaging with respect to UPC level. The
R2 of the model is 0.58 and the
F-value is 6441, which is statistically significant at the 0.001 level.
The dependent variable is UPC-level USD sales. Eco-packaging is an indicator variable in terms of eco-packaging. Holiday is a dummy variable for holidays including Thanksgiving and post-Thanksgiving, Christmas, the Super Bowl, Independence Day, Labor Day, Columbus Day, and Martin Luther King Day. Manager denotes manager special in a store. Deletion is typically carried out with a 6-week price reduction prior to removing the item from inventory. Sales adv. includes the weekly flyers and in-store newspapers. Store is a fixed effect of 10 stores.
The eco-packaging appeal has a statistically significant and positive effect on UPC-level sales ( = 4.387, p = 0.003), supporting Hypothesis 1. This result can be interpreted as: when a product is featured with an eco-packaging appeal, the sales of the product increases by USD 4.3 on average. Indeed, the analysis justifies continuous development and research on sustainability marketing in retail settings, especially in the context of FMCG.
With regard to the control variables, the holiday week dummy variable does not show any significant effect on sales; this implies that eco-packaging is not specific to a season or a time of the year since our holiday control variable is spread across seasons (Independence Day in summer, Labor Day in early fall, Columbus Day, Thanksgiving, and post-Thanksgiving in fall, Christmas and Martin Luther King Day in winter, the Super Bowl in late winter). Unlike other products (e.g., hot chocolate during winter) or product features (e.g., pumpkin flavor during fall), eco-packaging does not have a seasonality and can be featured in a product all year round. From promotion activities, we find a positive and significant effect of sales advertisement such as weekly flyers and in-store newspaper advertisement, which has the highest effect of 68.313. The coefficient of manager specials, however, is negatively and significantly associated with UPC-level sales. One possible explanation might be reduced unit price; the price promotion increases number of units sold but decreases the overall revenue since unit price is significantly reduced. Unless the change in units sold exceeds the change in unit price, the overall effect remains negative.
With regard to Hypothesis 1 and the main variable (eco-packaging own effect on UPC-level sales), we observe evidence of a positive correlation between eco-packaging and product item-level financial performance. This will contribute to the sustainability appeal as a viable marketing tool, which previously has had mixed results (e.g., [
31,
36]).
In Hypothesis 2, we posit that eco-packaging will positively influence overall brand-level sales due to a brand spillover effect. To measure this rather complex phenomenon, we decompose the model into three parts: brand spillover effect from line extension, brand extension, and combined introduction. We illustrate the regression results in
Table 2.
Ep_ratio is the ratio of eco-packaging in brands by category (Equation (3)) and
Ep_other is the average ratio of eco-packaging in all other product categories in the same brand (Equation (4)). The interaction term of
Ep_ratio and
Ep_other measures the spillover effect of eco-packaging appeal, where the line extension and brand extension strategy simultaneously exist in one brand.
The dependent variable is the brand–level sales in USD. Ep_ratio is the ratio of eco-packaging in a category of a brand. Ep_other is the average ratio of eco-packaging in other categories in the same brand. Holiday is a dummy variable for holidays including Thanksgiving and post-Thanksgiving, Christmas, the Super Bowl, Independence Day, Labor Day, Columbus Day, and Martin Luther King Day. Manager denotes manager special in a store. Deletion is typically carried out with a 6-week price reduction prior to removing the item from inventory. Sales adv. is anything in the weekly flyer and/or in-store newspaper. Store is a fixed effect of 10 stores.
Table 2 displays the spillover effect from three different sources. The coefficient
Ep_ratio is positive and significant (
= v8.245,
p = 0.032) and
Ep_other is negative and significant (
= −10.538,
p < 0.000).
Ep_ratio refers to the spillover effect from line extension within the same product category (e.g., the impact of Sun Chips biodegradable plant-based packaging on total Sun Chips sales). The coefficient of Ep_ratio is positive and significant at 8.245, suggesting that if eco-packaging intensity in a product category increases by one percent, the brand-level sales within the same product category increase by USD 8.245. Ep_other captures the spillover effect from other product categories (e.g., the impacts of Dannon yogurt packaging with recycled materials on the sales of Dannon protein shakes). We find a negative spillover effect for eco-packaging with regard to the brand extension; the coefficient of Ep_other is negative and significant at −10.538. The negative parameter of Ep_other shows the substitution effects [
40]. That is, the introduction of eco-packaging in one product category reduces the sales of other product category that carries the same umbrella brand name. Although the introduction of eco-packaging increases the own product category sales with the same brand name (
= 8.245, p = 0.032), the results reveal that consumers are not likely to extend that positive association (sustainability) to other product categories with the same brand.
The interaction term of
Ep_ratio and
Ep_other captures combined spillover effects in our model, that is, brand spillover from both line and brand extension. The interaction term
Ep_ratio × Ep_other is negative and significant (
= −20.419,
p < 0.006), indicating that as eco-packaging appeal is extended to its own and other product categories simultaneously with the same brand (concurrent line and brand extension), the brand-level sales of the own product category indeed decrease, showing a negative spillover effect (
Figure 1).
Four figures in
Figure 1 are drawn with the estimates in
Table 2 by setting the average sales at USD 217.957. For example, when both
Ep_other and
Ep_ratio are zero (no eco-packing in own category and other category), then the brand sales of own category are USD 238.264 (USD 217.957 + USD 20.307). When
Ep_other is one and
Ep_ratio is zero (the own category has no eco-packaging and 100% of products have eco-packaging in the other category with the same brand name), then the brand-level sale of own category becomes USD 227.726 (USD 217.957 + USD 20.307 -USD 10.538). Thus, when
Ep_ratio is zero, the discrepancy of sales between
Ep_other = 0 and
Ep_other = 1 is USD 10.538. On the other hand, when
Ep_other is zero and
Ep_ratio is one (all products in own category have eco-packaging while no eco-packaging exists in other categories under the same umbrella brand), then the brand-level sales of own category becomes USD 246.509 (USD 217.957 + USD 20.307 + USD 8.245). Lastly, when both
Ep_other and
Ep_ratio are one, then the brand-level sales of own category becomes USD 215.552.
Figure 1 indicates that as
Ep_other is close to one, the negative spillover effect is escalated. In conclusion, Hypothesis 2 is partially supported. While the line extension of eco-packaging has a positive brand spillover effect, the brand extension shows a negative brand spillover effect.
Next, we investigate the non-linear effect of eco-packaging appeal and the regression results are shown in
Table 3. For the multicollinearity issue of an eco-packaging variable, we centered the eco-packaging value in our model [
46]. The coefficient of the centered eco-packaging is positive and statistically significant (
= 178.897,
p < 0.0001). Besides, the coefficient of the squared term of centered eco-packaging is negative and statistically significant (
= 199.073,
p < 0.0001). Consistent with our expectation in Hypothesis 3, we find an inverted U-shaped curvilinear relationship between eco-packaging intensity and brand-level sales. The brand level sales have a peak at the point of 0.447 of centered eco-packaging by derivatives. This result indicates that as the eco-packaging ratio increases, the brand-level sales increase and reach the peak of sales, and then decrease thereafter.
The dependent variable is the brand-level sales in USD. Eco-packaging center is the centered ratio of eco-packaging in a category of a brand to address the problem of multicollinearity. Holiday is a dummy variable for holidays including Thanksgiving and post-Thanksgiving, Christmas, the Super Bowl, Independence Day, Labor Day, Columbus Day, and Martin Luther King Day. Manager denotes manager special in a store. Deletion is typically carried out with a 6-week price reduction prior to removing the item from inventory. Sales adv. is anything in the weekly flyer and/or in-store newspaper. Store is a fixed effect of 10 stores.
5. Discussion
The objective of our study is to discover the relationship between eco-packaging appeal and financial performance on two levels of product hierarchy (UPC and brand) in an FMCG context. We also test it in two different product hierarchies, UPC and brand. To contribute to the body of knowledge on the brand spillover effect of eco-packaging, we further decompose brand spillover effect: those from the own product category (line extension) and those from other product categories (brand extension) with the same umbrella brand. Nonlinearity is also evaluated to address the rather complex mechanism of subject matter, which might otherwise misguide practitioners. What emerged from our empirical study is more complicated than the current literature has demonstrated. Like the prior literature, our results show that eco-packaging does contribute to its own product sales. This tells practitioners that sustainability is indeed preferred by general consumers in relatively low-involvement, frequently purchased product categories. This result appears to validate the findings and expectations of previous studies [
8,
34].
The test for brand spillover effect reveals rather interesting results: brand spillover effect is present when eco-packaging intensity is high in the own product category, but not when the intensity is high in other product categories even if they both carry the same umbrella brand. In other words, eco-packaging brand spillover effect comes from the line extension strategy, but not from the brand extension strategy. Our finding confirms finding from previous literature on reciprocal brand spillover effects [
40,
47] but extends knowledge to a diverse area of research: the sustainability in packaging strategy. While previous studies focus on the traditional traits of the core products in brand spillover effect, we devoted the study to packaging which was considered as rather peripheral and not a meaningful factor of purchase decisions [
19]. An inverted U-shaped relationship between eco-packaging intensity within a brand and brand-level sales implies that adding eco-packaging appeal to products would actually increase the whole brand image immediately. However, consumers react to this sustainability appeal in a rather non-product-related way up to a certain point. After a peak point, the positive association transfer is weakened and consumers look for other traits in products. To the best of our knowledge, our study is the first to fully explore the phenomenon (nonlinearity of eco-packaging return on brand-level sales), measuring the actual behavioral pattern of consumers. Without this full understanding, scholars and practitioners might be misguided due to either underestimated or overestimated financial efficacy of eco-packaging appeal. Our results support the previous studies in the field of the theory of reasoned action [
29] and its causal link of positive attitude → purchase intention → actual purchasing. While the first correlation has been extensively studied (attitude and intention), the last linkage has been largely ignored. The current study fills the gap in the literature. For practitioners, our study also provides some meaningful and useful normative guidelines on eco-packaging in own products, across products, and product development strategy.
Companies should acknowledge the full impact of excessive packaging on the environment, local and global communities, and animal wellbeing where the businesses are embedded. Firms can be proactive with regard to sustainability since it does financially pay and justify the expenses. We assume there is no conflict between product and brand managers since adding eco-packaging is financially rewarding at both levels, although disproportionately. Investors should be patient with a firm’s pro-ecological commitment since at a brand level, it requires more resources and research to realize increasing rewards. Practitioners should not interpret our results in an incorrect way in that simply adding a large assortment of eco-packaging appeal guarantees brand-level success. Rather, brand managers should focus on continuous innovation and manufacturing techniques to reduce price and improve quality of the core product to appeal to larger segments [
36]. Furthermore, brand managers should conduct ongoing research to monitor sales so as not to extend past the peak where brand-level returns seem to decrease after a certain point.