The Effect of Non-Monetary Promotions on Product
Returns
By Lisanne Oosting
The Effect of Non-Monetary Promotions on Product
Returns
By Lisanne Oosting S2964937 l.oosting.2@student.rug.nl Master Thesis University of Groningen Faculty of Economics and BusinessMSc Marketing Management & Marketing Intelligence
14-06-2020
Management Summary
These days, a lot of companies struggle with high percentages of product returns. These product returns are very costly for companies (i.e. around 550 billion dollars in the United States alone already). Therefore, it is important for companies to gain insights in the drivers of product returns in order to hopefully reduce these costs. This research focuses on the effect of non-monetary promotions on product returns.
Research about the antecedents and consequences of product returns are on the rise, however, there are still some topics which remain largely unexplored. One of these topics is the effect of marketing promotions on product returns. Whereas a lot of research has been done about the effect of marketing promotions on sales, it is also interesting to see what the effect is on product returns. A popular marketing promotion a lot of companies nowadays use is giving away free products or samples when customers buy a product. This promotion has been used to increase sales and to increase the share of wallet of the customers (Liu et al, 2011). Numerous studies have researched the effect of this promotion on sales, however, few studies have looked at the effect on product returns. This research tries to fill this gap by investigating the effect of non-monetary promotions, in this case gifts, on the product return probability.
To investigate this effect, data from a European retailer that sells products online and, in their brick,-and-mortar stores in The Netherlands was used. The dataset contains information about the orders of the Dutch web shop and shows if a product was being returned or not. Moreover, the data shows information about whether the order contained a gift and if it does, it also shows which type of gift it is. This study distinguished between two types of gifts: gift based on a given product (e.g. product related gift) or whether the gift is based on the basket value (e.g. basket related gift). Besides, it is being investigated what the effect of the gift on product return probability is when the gift is related to the ordered product (e.g. contains the same product category). Moreover, other factors could also influence the product returns and have been controlled for (basket characteristics, product characteristics, situation-specific factors and payment method).
To investigate the probability of a product being returned, a binary logistic regression model was used. This research finds that when customers receive a free gift when they order something, they are less likely to return the focal product compared to a customer who did not receive a gift. When splitting up this gift into basket related and product related gifts, it can be noted that both types of gifts will also lead to a decrease in product return probability.
probability than when the gift is unrelated to the focal product. When splitting up the gifts into basket related gifts and product related gifts again, it again shows that when a gift (both basket related and product related gifts) is related to the focal product, it will lead to a lower product return probability.
Other interesting insights from this study are that the return probability increases when customers orders products over the weekend, have more items in their basket or pay with Afterpay. Moreover, the return probability decreases when customers order products during the holiday period (since it is likely to be a gift for someone), which are on sale or for which they have to pay shipping costs.
Preface
This thesis displays my final work for the Master Marketing Management & Marketing Intelligence. During this programme I have learned a lot for which I am grateful since it helped me develop my skills and knowledge which I can deploy in my future career. Writing this thesis has been challenging but also very enjoyable since it helped me to put my skills obtained during my study to practical use, obtaining insights from real-life data.
Table of Contents
1. Introduction ... 7
2. Theoretical Background ... 10
2.1 Product returns ... 10
2.2 Effect of gift with purchase on product returns ... 11
2.3 Moderating effect of basket value gifts and product-based gifts ... 12
2.4 Moderating effect of product relatedness ... 13
3. Methods ... 15
3.1 Data ... 15
3.2 Key variables ... 15
3.3 Control variables ... 17
3.4 Data inspection and cleaning ... 19
3.4.1 Outliers ... 20 3.4.2 Missing values ... 20 3.5. Multicollinearity ... 21 3.6 Descriptive statistics ... 22 3.7 Model Formulation ... 23 4. Results ... 25 4.1. Model Fit ... 25 4.2. Model Results ... 26
4.2.1 Results Main Variables ... 27
4.2.2 Results Control Variables ... 29
5. Discussion ... 34
5.1. Theoretical Implications ... 35
5.2. Managerial Implications ... 36
5.3. Limitations and Future Research ... 38
6. References ... 40
1. Introduction
Nowadays, a lot of companies are struggling with high percentages of product returns. The average return rate for e-commerce companies is around 20 percent (Orendorff, 2019). For apparel retailers the return rates are even higher, around 30 percent, whereas for brick-and-mortar stores, the return rate is around 5 to 10 percent (Kaplan, 2019). There is even a special day on which high product returns are expected, namely National Returns Day which was on the 2nd of January (Business Insider, 2020). On this day this year, it was expected that 1.9
million packages would be returned in the United States alone (Business Insider, 2020). These product returns are very costly to companies. By 2020, it is expected that product returns will cost around 550 billion dollars solely in the United States (Statista, 2019). Four years ago, this figure was four times less, approximating 137.5 billion dollars (Statista, 2019). Therefore, obtaining insights into this product return behaviour and in turn hopefully reducing these costs, is an important driver for companies.
Research about the antecedents and consequences of products returns are on the rise. Surprisingly, not a lot of products are being returned with reason of being defect. However, most dominant reasons for returning products are 1) it does not meet customer’s expectations and/or, 2) the customer found a better price or product (Powers & Jack, 2015). Whereas there already has been a lot of research done about product returns, there are also still a lot of topics unexplored. For instance, the effect of marketing instruments on product returns is largely unknown, whereas the impact of marketing on sales has been examined numerous times.
Nowadays, a lot of companies are giving away free products (also called “gift with purchase”) or samples when consumers buy something. This has been a popular non-monetary promotion to increase sales and to increase the share of wallet of customers (Liu et al 2011). While various studies have researched the effect of this marketing technique on sales, few studies have been performed with regard to the effect on product returns. Whereas this technique may actually help to reduce the amount of product returns or opposite, may increase returns. Previous research (Lee & Yi, 2017) has found that by offering a free product, customers are less likely to return products. However, research by Raghubir (2004) has shown that gifts can also have a negative side: namely, harming the brand.
This research sheds light on the effect of non-monetary promotions, in this case gifts, on the product return probability. Therefore, the main research question is:
In order to answer this question, multiple sub questions have been formulated:
1. What is the effect of gifts on the product return probability?
2. How does the effect of gifts on product return probability depend on whether the gift is based on basket value or based on a product?
3. How does the effect of gifts on product return probability depend on whether the gift and focal product are related to each other (e.g. same product category)?
These questions will be analysed by using data from a European retailer that sells products online and in their brick-and-mortar stores in The Netherlands. The data used in this study only contained data about the Dutch web shop and did not encompass any data about the Dutch brick-and-mortar stores, the market places the company is active in or about the sales of the foreign web shops. The timeframe from this data ranges from October 1st 2019 till March 15th
2020. The data encompasses 12 categories: jewellery, watches, bags, sunglasses, wallets, accessories, watch straps, headwear, belts, scarves gloves and others (i.e. keychains). Moreover, this retailer offers a lot of free gifts throughout the year and therefore this data is convenient and fitting to investigate the effect of gifts on product returns. To investigate the effect of gifts on the product return probability, a binary logistics regression model has been used.
The key insight of this study is that gifts decrease the return probability. This effect is supported for the overall gifts’ variable, gifts based on the basket value and gifts based on a product. Thus, when customers receive a gift, they are less likely to decrease the focal product. More specifically, if a customer receives a gift, the return probability decreases by 12.8% compared to when the customer receives no gift. Moreover, this research also found support for the moderation effect of product relatedness between the gift and the return probability. It is found that when the focal product is related to the gift (e.g. same product category), customers are less likely to return the product. This holds for all types of gift, the overall effect of gifts, basket related gifts and product related gifts.
topic. Moreover, previous research hasn’t looked at the effect of the nature of the free product (i.e. whether the gift and ordered product are related). Whereas this research also took this into account to investigate the effect on return probability. Besides, this study also extends previous research by investigating the effect of different types of gifts on product return probability: gifts based on the basket value or gifts based on a product. Previous research has not made this distinction yet but looked at gifts in total.
2. Theoretical Background
This section will elaborate on the theoretical background for this research. First there will be a section about previous research about product returns in general and what this study will examine by showing the conceptual framework. Then an elaboration, based on theory, will be given on the main variables for analysis: the effect of gifts on product returns, how this effect will be influenced by basket-related or product-related gifts and how product relatedness mediates the relationship between gifts and return probability.
2.1 Product returns
Over the last decade there has been a growing interest and growing body of literature with regard to product returns. Most research have focused on the effect of the product return leniency on product returns. Janakiraman et al. (2016) have identified 5 dimensions of product return leniency: time, money, effort, scope and exchange. They found that time and exchange leniency (i.e. offering longer return periods and cash back) will decrease product returns significantly.
It has been found by research that not a lot of products are being returned because they are faulty but most often products are being returned because it does not meet customer’s expectation (Powers & Jack, 2015). Research by Minnema et al. (2016) & Sahoo et al. (2018) have for instance shown that when products are having overly positive customer reviews, this may actually increase product returns. This is because by having all these positive reviews, customers are setting high expectations for this product and when they receive this product, it may then not meet these high expectations and therefore, it is being decided to return the product.
receiving a free product, this might trigger the effect of loss aversion when they would like to return the focal products. In that case, people are more likely to not return their products since losses weigh heavier than gains. This research will focus on this topic: the effect of non-monetary promotions on product returns. To give better insights in what this research will examine, the conceptual model is shown in Figure 1. The different parts of the conceptual model will be elaborated on in the following sections.
Figure 1: Conceptual Framework
2.2 Effect of gift with purchase on product returns
Nowadays, a lot of companies are giving away free products (also called “gift with purchase”) or samples when consumers buy something. This has been a popular marketing technique to increase sales and to increase the share of wallet of customers (Raghubir, 2004; Liu et al 2011). However, like mentioned before, little research has been done about the effect of this non-monetary promotion on the product returns. The endowment effect will be used to assess the relationship between gift with purchase and product returns. According to the endowment effect, when an individual owns a product, it will be more highly valued than those not owned (Thaler, 1980). This is because when a product is not in a person’s endowment
H1: Gift with Purchase
- H2a: Gifts based on basket value
- H2b: Gifts based on products Product Returns
H3a/b: Product Relatedness
_ _ +
+/-
Control variables:
Weekend Product Category
Discount Payment Method
Urbanity Items in Basket
be seen as a gain (Thaler, 1980). In accordance with the loss aversion theory, which is related to the endowment effect, losses are weighted more heavily than gains (Kahneman & Tsversky, 1979; Kahneman et al., 1991). An implication of this loss aversion theory is that people prefer to remain the status quo, also called the status quo basis (Samuelson & Zeckhauser, 1988). As a consequence, when people experience loss aversion, they prefer to remain the status quo since losses weigh heavier than gains (Samuelson & Zeckhauser, 1988; Kahneman & Tsversky, 1979; Kahneman et al., 1991). Thus, in the example of receiving a gift, people prefer to keep the product plus the gifts instead of losing them both.
Previous research about the effect of gift with purchase on product returns has found that by offering a free product, customers are less likely to return the products (Lee & Yi, 2017; Lee & Yi, 2018). Lee & Yi (2017) have found that offering a gift with purchase will lead to a decrease in product returns due to the effect of loss aversion and perceived ownership when customers get to choose their gift. Other research by Lee & Yi (2018) have focused on the effect of promotion framing on product returns. They found that companies who are offering a free gift as promotion are experiencing less returns than companies who are offering a bundle promotion (i.e. buy product A and product B for €X amount of money). This is also due to the effect that giving up a free product results in greater loss compared to giving up the bundle promotion. However, these previous studies have mentioned that more research needs to be done in order to increase generalisability and validity since they mostly used scenario-based experiments.
When using the loss aversion theory, it can be argued that when people have to give up their gift, they might prefer to remain the status quo. When a customer wants to return their focal product, they also lose the right to keep the gift and they have to return it. As a consequence, this customer might experience psychological pain due to the loss aversion effect (Lee & Yi, 2017). Therefore, it is being posited that:
H1: Offering a gift with purchase will lead to lower product returns
2.3 Effect of basket value gifts versus product-based gifts
theory (Rao & Monroe, 1989). As a consequence, when brands are offering price promotions, customers might perceive the product as less of quality due to the lower price (Raghubir & Corfman, 1999). This price-quality inference does not only hold for monetary promotions but can also be applied to non-monetary promotions. When customers receive a free gift with their purchase, they might make inferences about the quality of the ordered product or even the brand based on the gift (Raghubir, 2004; Raghubir & Corfman, 1989). The other theory which also supports this notion is the value-discounting theory. This theory was found to be supported in Raghubir (2004) its research. According to Raghubir (2004), when companies provide free gifts together with the purchase of products, consumers will imply that the purchased product was overpriced/had a high margin, that the gift had a low value or even both. So, this free gift can make customer create interferences about “the cost and margin structure of the promoted product or the free gift or both” (Raghubir, 2004). This is because customers know that these kinds of promotions will also erode the profit margins of the company. As a result, when customers believe that the value of the gift is low, they are also willing to pay less for the gift.
The research by Raghubir (2004) showed that giving away gifts can harm the brand due to the value-discounting theory mentioned above. Customers believe that when they receive a free gift together with the purchase of a product, the value of the free gift is low, and this also leads to a negative spill-over to the brand. This research has been done when a gift was given away with a certain product, so product related gifts. With regard to gifts based on the basket value, this effect might be less clear. Then the free gift is not specifically related to a product or brand and therefore, consumers may not make these strong interferences about the cost and margin structure of the product or gift. Therefore, the negative spill-over effects might not be evident in the case of basket related gifts.
To test this, the following hypotheses are being posited:
H2a: Offering gifts based on basket value will lead to lower product returns H2b: Offering gifts based on products will lead to higher product returns
2.4 Moderating effect of product relatedness
However, it can also be the other way around since you already bought a piece jewellery and do not receive extra utility when getting one extra for free.
Consumers return products when they perceive the utility of returning the product greater than the utility of keeping the product (McFadden, 1974). The effect of the gift with purchase and the product relatedness can be two-sided. On the one hand it may be the case that when getting a gift with purchase which is unrelated to the focal product (e.g. low product relatedness), customers may perceive extra utility from the gift since it is a totally different product than what they ordered (Lee & Yi, 2017). If a customer thinks about returning the focal product, they might actually be persuaded by the unrelated gift as it adds extra utility. Therefore, the customer might decide to not return the focal product as this also means losing this unrelated gift which adds extra value and utility.
On the other hand, it may also be the case that higher product relatedness may actually create utility for the customer as can also be seen in loyalty programs. When extending this topic of product relatedness to the field of loyalty programs, it has been stated that rewards such as free products are often related to the products being sold as this is oftentimes more effective (i.e. Buy 10 coffee packs and get one for free) (Berman, 2006; Bijmolt, Dorotic, Verhoef, 2010). Rewards that show an overlap with the product and thus are related, are preferred because these types of rewards can increase the favourable brand association and attitudinal attachment (Roehm et al, 2002).
Since the effect of product relatedness can be two-sided, opposing hypotheses are being posited:
H3a: Receiving a gift which is related to the purchased product relative to receiving an unrelated gift will lead to lower product returns.
3. Methods
3.1 DataThe data for this research comes from a European retailer that sells products online and in their brick-and-mortar stores in The Netherlands. The data encompasses 12 categories: jewellery, watches, bags, sunglasses, wallets, accessories, watch straps, headwear, belts, scarves gloves and others (i.e. keychains). The timeframe from this data ranges from October 1st 2019 till March 15th 2020. Data before that period is also available, however, due to logistical
reasons (e.g. new systems in place) and time constraints, it was decided to start the data collection from October 1st. The collected data only contained information about the Dutch web
shop and did not encompass any data about the Dutch brick-and-mortar stores, the market places the company is active in or about the sales of the foreign web shops. The data was collected from different systems and platforms. First, the data about the orders and their returns are being extracted from Google Cloud Platform. This data was then enriched from data from another system of the retailer (information about the payment method, the ZIP code of the customer and information about the products itself). Moreover, data from Centraal Bureau voor Statistiek (CBS) is being used to retrieve how urban a certain ZIP code is. Furthermore, data from the Koninklijk Nederlands Meteorologisch Institut (KNMI) is gathered to retrieve data about the temperature and rainfall during a specific day. The raw data started with 455.003 observations. However, as will be discussed in one of the following parts, cleaning needed to be done in order to prepare the data for analysis.
3.2 Key variables
This company offers a lot of free gifts throughout the year. These gifts can be distinguished into two groups: Gifts based on basket value and gifts based on a given product. The gifts based on a given product can entail different categories (i.e. jewellery, wallets, water bottle, etc.) and is given away for free with certain products. Most often this gift is a product from the company its own brand, however, sometimes other brands are also offering gift which the company then also gives to their customers when they purchase certain products. An example of this type of gift is when you order Calvin Klein jewellery, you will receive a free Calvin Klein water bottle.
different types of products, such as a bracelet, sunglasses, laptop cases, discount vouchers, etc. An example of a goodie bag can be seen in Figure 2. This goodie bag was given away last Valentine’s Day, and it contained; a bracelet, a watchstrap, a chocolate bar and a €30 gift card. Another example of this type of gift is when the customer receives a free bracelet when they spent €50.
Insert Figure 2 here
Thus, the big difference between these two types of gifts is whether a gift is given away based on a specific product you purchased or based on the amount of €s you spent. These types of gifts have been operationalized in the dataset in the following way. Firstly, the gifts based on basket value have been operationalized. For this, it has been checked whether the order contained a basket related gift and based on this, a dummy variable has been created: Basket Related Gifts. If the order contained a basket related gift, this value would become 1 and otherwise it would be 0. Secondly, for gifts based on a specific product, another dummy variable has been created (e.g. Product Related Gifts). For this variable, it has been checked whether a product within the order contained a product related gift and if that was the case, the value would become 1 and otherwise it would be 0. Lastly, for the general effects of gifts, these two types of gifts are then combined to create the variable Total Gifts. Similar for this variable, dummy coding has been used where 1 means that there is a gift given away and 0 when there is no gift.
Another key variable is the Product Relatedness. This has been operationalized by creating a dummy variable and checking whether the product category of the gift is the same as the product category of the ordered product. If it was the same, the dummy variable would become 1 and otherwise it would become 0.
When looking at how often gifts are being given away, it can be stated that around 41% of all orders contain a gift which can be split up again in basket and product related gifts. From these gifts, 78% is a gift based on basket value and the other 22% is a gift based on a certain product which is being purchased. This difference is due to the fact that the gifts based on basket value receive more communication and are therefore more popular than the gifts related to a certain product.
return rate for this company is 17.17% which is higher than the average industry return rate of 9% (Kerr, 2013). However, this percentage is still lower than the clothing and shoes categories which show an average return percentage of 30-40% (Reagan, 2019). Customers could return their products free of charge within 30 days of the delivery date. That is why the gathered data until March 15th was enriched with data at April 14th to see whether any ordered products until
March 15th were returned in the 30-days the customers still had the chance to do so.
A summary of the operationalization of the key variables can be seen in Table 1.
3.3 Control variables
Several other variables will be used in this study to control for any other factors which might influence the product returns. Firstly, basket characteristics are being used as a control variable. A variable for the Basket Value in euros has been created and a variable for the number of Items in a Basket has been created. Moreover, customers had to pay €3.95 for shipping costs when their order was below €50. Above this amount the shipping costs were free of charge. This factor will also be taken into account by creating a dummy variable, Shipping Costs, with a value of 1 whenever the customer had to pay shipping costs and a value of 0 when the customer enjoyed free shipping.
Secondly, product characteristics are being taken into account. A dummy variable was created to see in which Product Category the product belongs. The reference category for this variable is jewellery since jewellery is being sold the most. Moreover, whether a product is on Sale or not will be included in the model. To determine whether a product is on sale, the price which will be looked at in this study is the Manufacturer Suggested Retail Price (MSRP). The amount the customer has paid for the product will be compared to this MSRP and when the customer has paid less than the MSRP, the product is on sale. Previous research by Petersen & Kumar (2009) have shown that products being purchased on sale are returned less often than products purchased at the regular price. Therefore, this effect will be controlled for in this study. Besides, a dummy variable has been created to investigate the effect of male, female and unisex items, called Gender Item. The gender of the customers was not readily available in the systems of the company since most customers do not fill out their gender when ordering as this is not mandatory. Therefore, it was being decided to look at the items and see whether it is a male, female or unisex (a key holder or a cleaning cloth for jewellery for instance).
and gift cards. This variable will also have a dummy coding where iDeal is used as reference level since this payment is one of the most popular payment methods.
Lastly, situation-specific factors may also the effect on product returns. A dummy variable will be created to control for the effect of Holiday periods (i.e. whether the day on which the customer made an order is in November/December). During the holiday periods, customers are more likely to return their products than during the rest of the year (Petersen & Kumar, 2009) and therefore, this needs to be controlled for. Moreover, the weather might also play a role in product returns and therefore data from the KNMI is gathered about the Temperature and Rainfall on a specific day. Another situation-specific variable which might play a role is the Urbanity of a certain place. This information is gathered from the CBS and comprises information about the number of inhabitants per postal code. Furthermore, the day of the week might also have an effect. This is included in the model as a dummy, called Weekend, where 1 means that the order was placed during the weekend and 0 when the order was placed during the week.
All these control variables can also be seen in the Conceptual Framework (Figure 1) and are summarized in Table 1.
Table 1: Variable operationalization
Variable Coding Description
Product Returns Dummy 0/1 1 when a product is returned
Total Gifts Dummy 0/1 1 when the customer received a gift
Basket Related Gifts Dummy 0/1 1 when the customer received a gift based on basket value
Product Related Gifts Dummy 0/1 1 when the customer received a gift based on a product
Product Relatedness Dummy 0/1 1 when the received gift entails the same product category as the ordered product
Total Gifts * Product Relatedness
Dummy 0/1 1 when the received gift entails the same product category as the ordered product
Basket Related Gifts * Product Relatedness
Dummy 0/1 1 when the received gift based on basket value entails the same product category as the ordered product
Product Related Gifts * product Relatedness
Weekend Dummy 0/1 1 when the order is placed during the weekend
Discount Dummy 0/1 1 when the product was on sale
Urbanity Number of inhabitants Number of inhabitants per postal code
Shipping costs Dummy 0/1 1 when the customer had to pay
shipping costs
Temperature In °C Average temperature in The Bilt on
the day of purchase
Rainfall In mm Amount of rainfall on the day of
purchase
Product Category Dummy 0/1 1 when the product is a certain
category (reference category is jewellery)
Payment Method Dummy 0/1 1 when a certain payment method is
used (reference category is iDeal)
Items in Basket In numbers # of products within the basket
Basket Value In €s Order value of the basket in €s
Gender Item Dummy 0/1 1 when an item is for a specific gender
(reference level is female)
3.4 Data inspection and cleaning
As mentioned before, three types of data sources have been used in order to investigate the research questions: Google Cloud Platform, back-end systems of the retailer, data from the CBS and data from the KNMI. When exporting from Google Cloud Platform, the data was inefficiently ordered. Thus, to make it efficient for analysis it had to be transformed to some extent. This raw data consisted of rows for every purchase and every return. So, when a customer purchases something and later on returns it, this return will show up as an extra row in the dataset. This is not efficient and had to be transformed in the way that the returns will be added to the row of the purchased product. This was done by splitting the dataset into two: a dataset with only sales and a dataset with only returns. Then these two datasets were being merged based on the transaction ID, product ID, product description and price. This resulted in a correct dataset where the product returns were placed as a column in the same row as the purchased products. With regard to the other data sources which are being used for this thesis, the format of the data was appropriate and therefore could be combined with all the other data without transformation.
which our payment methods exist of. Moreover, the raw data also contained rows about the delivery, in this case PostNL, which could be deleted from this dataset since it contained no new information about the ordered products. Besides, certain orders were not correctly shipped since it could be the case that the ordered product was not in stock and therefore never shipped to the customer. These orders have also been excluded from the analysis. Besides, it could also be noted that there were some fraudulent customers who for instance, received free products and did not pay anything. These fraudulent customers are being excluded from the data.
3.4.1 Outliers
In order to check for outliers in the dataset, multiple boxplots have been created. One variable contained outliers for which correction was needed. When looking at the outliers of Items in Basket it could be noted that there were 10 orders which contained more items in the basket than there were in real after checking the retailer’s systems (i.e. having 23 items in a basket meanwhile it should be 8). This can be due to a measurement error in the raw data and these orders are removed from the dataset in order to not influence the reliability.
Another variable which contained outliers was the variable Basket Value. For instance, some orders contained a value of €3500 whereas the average basket value is €109. However, this is completely explainable since some products in the assortment are more expensive (i.e. a 14k golden bracelet or an expensive watch brand). After inspecting these high basket values in the dataset, all these outliers are due to the fact that these customers indeed ordered more expensive products. Therefore, these outliers do not have to be corrected for and do not influence reliability.
3.4.2 Missing values
values, mean imputation has been done. This is a fast way to deal with missing values and can be done when there are only a handful of missing values (Vehkalahti & Everitt, 2018).
3.5. Multicollinearity
To investigate whether the independent variables are highly correlated with each other, a correlation matrix has been created. This correlation matrix can be seen in Appendix 1. Moreover, the VIF scores have also been calculated and these can be seen in Appendix 2.
There are a few marginal to strong correlations which need some clarification.
Firstly, there is a moderate correlation between Totalgifts and Productrelatedness (r = 0.62). However, this seems logical since Productrelatedness can only become 1 if the customer received a present and thus, Total gifts is 1. However, this should not influence the results of the model since Productrelatedness will not be included as a main effect in the model and only as an interaction effect. This is because as said before, Productrelatedness is dependent on Totalgifts becoming 1 and therefore should not be included in the model as a main effect but only as an interaction effect.
Another marginal correlation can be seen between Basket Value and Items in Basket (r = 0.54). When taking into account the VIF scores, it can be stated that this correlation should not influence the results. The Basket Value shows a VIF score of 1.69 and Items in Basket shows a VIF score of 1.60. The values are both below the threshold value of 4 and therefore multicollinearity should not be a problem in the model.
The variables Basket Related Gifts and Total Gifts show a strong positive correlation (r = 0.86). This seems logical since most of the gifts are gifts based on the basket value (78%). However, these two variables will never be put in the same model since Basket Related Gifts is a part of the Totalgifts. Therefore, this correlation should not be a problem for the results.
3.6 Descriptive statistics
After having cleaned and prepared the dataset for analysis, the dataset contained 126.504 observations. These 126.504 observations (products) can be split into 93.484 unique transactions across 82.239 different customers. As mentioned before, 41% of all orders contain a gift. Of this 41%, 78% is a gift based on basket value and the other 22% is a gift based on a certain product which is being purchased. Of these gifts, around half (51.70%) of the gifts is related to the ordered product as in it contains the same product category.
The dataset contains 126.504 sold products across 13 categories: jewellery, watches, bags, sunglasses, wallets, accessories, watch straps, headwear, belts, scarves, gloves and others (i.e. keychains). The product category which generates the most sales is jewellery with 56.58% of all transactions being jewellery. More descriptive statistics about the sales and returns of product categories can be seen in Table 2.
Most products are female products (64.70%). Male and unisex items account for 31.70% and 3.60% of all items. Moreover, customers spend on average €108.54 and order on average 1.87 items per transaction.
Table 2: Descriptive statistics of sales and returns per product category Product
Category
Sales % Sales Returns % Returns
3.7 Model Formulation
To investigate the probability of a product being returned, a binary logistic regression model will be used. This model is being chosen since the dependent variable, product returns, is binary and thus can have two values: 0 for the product not being returned and 1 for the product being returned. All the variables mentioned above can be formulized in the following binary logistic regression model, which represents the probability of product i being returned given order j.
![#$ = 1] = Λ *α + -!"# ./ = exp (4 + -!"# .) 1 + exp (4 + -!"# .)
The -!"# . can be specified in the following binary logistic regression formulas where the variables can differ between each product i and each order j. Two models have been formulated since Total Gifts will never be combined in one model together with Basket Related Gifts & Product Related Gifts.
Model with Total Gifts:
4. Results
This section will discuss the results of the binary logistic regression models. Four different models have been run in order to test for different effects. The first model will only include the effect of Total Gifts plus all the control variables. The second model will only include the effect of Product Related Gifts and Basket Related Gifts plus all the control variables. The third and fourth model will also include the interaction terms. Hence, the third model will look at the effect of Total Gifts plus the interaction terms of Total Gifts and Product Relatedness plus all the control variables. The fourth model will then look at the effect of Basket Related Gifts, Product Related Gifts plus the interaction terms for both types of gifts and all the control variables.
4.1. Model Fit
The fit of all the four models will first be discussed and these results can be seen in Table 3. As can be seen in Table 3, all 4 models are performing significantly better than the null model/intercept only model. This is because when comparing the Likelihood Ratio test, all 4 different models show significance at the 1% level. When comparing the Akaike Information Criterion (AIC) across all models, Model 4 shows the lowest AIC (AIC = 105325) and thus results in a better fit. When interpreting the McFadden Pseudo R2, the higher the Pseudo R2 the better the model fit is. Model 4 then again shows a better fit than the other models since it has the highest Pseudo R2 (0.0929).
Thus, it can be stated that Model 4 shows the best fit. However, all models will still be analysed since all four models calculate something different.
Table 3: Model Fit
Measurements Null Model Model 1 Model 2 Model 3 Model 4
4.2. Model Results
To investigate the effects of all four models, the odds ratio results will be presented. These results can be seen in Table 4 and will be elaborated on in the sections below.
Table 4: Results binary logistic regression models (in odds ratios)
Variables Model 1 Model 2 Model 3 Model 4
Intercept 0.117*** 0.118*** 0.120*** 0.119***
Totalgifts 0.879*** 0.931**
Basket Related Gifts 0.872*** 0.905***
Product Related Gifts 0.892*** 1.073
Afterpay 2.508*** 2.507*** 2.507*** 2.507*** Mastercard 1.339*** 1.340*** 1.340*** 1.340*** Amex 0.913 0.917 0.919 0.924 Bancontact 0.436*** 0.436*** 0.436*** 0.436*** Giftcards 0.896* 0.894* 0.896* 0.893* Paypal 0.995 0.994 0.997 0.996 Visa 1.179** 1.178** 1.177** 1.177** Items in basket 1.195*** 1.196*** 1.196*** 1.197*** Basket value 1.002*** 1.002*** 1.002*** 1.002*** Male 1.359*** 1.357*** 1.356*** 1.347*** Unisex 1.461*** 1.456*** 1.467*** 1.467*** Signif. Codes: *** p < 0.01, ** p < 0.05, * p < 0.1
4.2.1 Results Main Variables
In all models, most main variables are significant. As can be seen in Model 1, Total Gifts shows an odds ratio of 0.879 and is significant at the 1% level. This means that when a customer receives a gift when purchasing an item, the relative odds for returning a product decreases by a factor of 0.879 (12.1%) compared to when the customer does not receive a gift. This is in line with H1 which stated that offering a gift will lead to lower product returns.
When splitting the Total Gifts into Basket Related Gifts and Product Related gifts, as is shown in Model 2, it is demonstrated that both types of gifts are significant (p< 0.01) and lead to lower product returns as the odds ratio is below 1. With regard to when customers receive a gift based on the basket value, the relative odds for returning a product decreases by a factor of 0.872 (12.8%) compared to when they do not receive a gift based on basket value. This result support H2a where it is being posited that offering a gift based on basket value will lead to lower product returns. With regard to when a customer receives a gift based on a specific product, the probability of returning the bought product decreased by a factor of 0.892 (10.8%). This result is contradictory to H2b where it is being stated that offering gifts based on product will actually lead to higher product returns. A possible explanation for this might be that the expected negative spill-over effect is not that strong. It might be the case that the loss aversion effect is stronger in the sense that customers prefer to keep the gift and thus not return the ordered product.
Figure 3: Return Probability for the Different Types of Gifts
Model 3 and 4 show the effect of the interaction term between gifts and product relatedness. In model 3, the variable Total Gifts has been included in the model together with the interaction term. The variable Product Relatedness itself has not been included as a main effect in the model since this variable is dependent on Total Gift being 1. Including this as a main variable will then be almost equal to the interaction term and therefore should not be included. Model 3 shows that Total Gifts is again significant (p<0.01) and shows that when customers receive a gift, the probability of returning a product decrease by 6.9% (Exp(β) = 0.931). The interaction term between Total Gifts and Product Relatedness shows that when the product is related to the bought product (e.g. having the same product category), this effect becomes even stronger. This means that when customers receive a gift which is related to the ordered product, the probability of returning a product decreased by a factor of 0.897 (10.3%). This supports H3a where it is being posited that when receiving a gift which is related to the focal product, it will lead to lower product returns.
receiving a Basket Related Gift which is related to the ordered product, it will lead to lower product returns.
With regard to the main effect of Product Related Gifts in Model 4, it is interesting to see that it does not show any significance anymore as was the case in Model 2. This is the effect when Product Relatedness is equal to 0. Hence, a gift based on a product does not show a significant effect when the focal product and the gift are not related. In contrast to this finding, the interaction term between Product Related Gifts and Product Relatedness is significant at the 1% level. Thus, when a Product Related Gift is related to the focal product (e.g. Product Relatedness is equal to 1), it will lead to a significantly lower product return probability. More specifically, if a customer receives a Product Related Gift which is related to the focal product, the probability of returning a product decreases by a factor of 0.757 (24.3%). This also supports H3a where it is being stated that when a gift is related to the focal product, it will lead to significantly lower product returns.
A visualization of Model 3 and 4, which includes the interaction term of Product Relatedness, can be seen in Figure 4.
Figure 4: Return Probability for the Different Types of Gifts including Interaction Terms 4.2.2 Results Control Variables
Across all four models, different control variables show interesting significant effects. Moreover, the results between the four models barely change which makes it reliable to interpret. For this section, the odds ratio plus the significance of the control variables will be interpreted from the results in Model 4 since this model showed the highest fit and since all models show similar results.
1.051 (5.1%) This might be explained by the fact that customers often go shopping, in this case online shopping, in the weekend for hedonic reasons and for purpose of entertainment (Babin & Harris, 2015). During the week, people often shop for products they really need since they might have less time because of work. As a result, when customers order something in the weekend, they can also show a higher probability to return something due to the hedonic shopping motives in the weekend.
For the Discount variable, when a customer orders a product which is on sale, the probability of returning the discounted product will decrease by a factor of 0.835 (16.5%). This is in line with what Petersen & Kumar (2009) have found. Customers may not be that critical when a product is on sale and it lacks fit than when it is not on sale and it lacks fit (Petersen & Kumar, 2009).
Shipping cost also has a significant effect on the return probability. When a customer has to pay shipping costs, the return probability decreases by 21.4% (Exp(β) = 0.786). An explanation for this might be that when customers have to pay a shipping fee, they have to incur extra costs for the desired product. In that case, they might be really certain about the product they want and are willing to incur these extra costs.
The Temperature also plays a significant role on the product return probability. When the temperature rises, the return probability decreases by 0.8% (Exp(β) = 0.992). This effect might also be related to the hedonic shopping motives. When the temperature is high, people might shop for things they really need. But when the temperature is low and people cannot go outside, they might shop online for hedonic reasons and to have something to do (Babin & Harris, 2015).
The Holiday control variable also shows a significant effect on the return probability. When customers order products during the holiday period (November & December), the return probability decreases by 13.5% (Exp(β) = 0.865). This is contradictory to the result found by Petersen & Kumar (2009). They found that when people order products during the holiday period, they are also more likely to return these products. An explanation for this contradictory result might be that the different categories of the retailers are also perfect for gift-giving. Thus, most products bought in November and December are meant as a present for someone else and thus are being returned less often. This is supported by research by Petersen & Kumar (2009) where it is being stated that “products received as a gift are less likely to be returned than products not received as gifts”.
probability of returning an item increases by 0.2% (Exp(β) = 1.002). Moreover, if there are more items in one basket, it will also lead to an increase in product return probability. More specifically, when the items in one basket increases with 1 unit, the return probability then increases by 19.7% (Exp(β) = 1.197). Both effects can be explained by the substitutability of items. When customers order more products which are substitutable, they are more likely to keep one of them and return the other product.
With regard to the Gender of the items, it can be noted that both male and unisex products are being returned more often than female products. More specifically, male items experience an increase of 34.7% (Exp(β) = 1.347) in return probability compared to female items. Moreover, unisex items experience an increase of 46.7% (Exp(β) = 1.467) in return probability compared to female items. Both are quite high but can be attributed to the case that most items this retailer sells are female items (64.7%). Therefore, to make this effect more generalisable, this relation should also be tested in other situations as well.
The effect of the different Product Categories on return probability can be seen in Figure 5.
Figure 5: Return probability per category compared to Jewellery
The categories which experience the strongest negative return probabilities, and thus have the lowest product return probabilities compared to jewellery, are others & accessories. This can be explained by the fact that the items categorized as others and accessories are oftentimes products which are sold when customers want to check-out. When the customer goes to the basket, he or she will be persuaded one last time to buy more products and will see other products which he or she can buy. Examples of these type of products are a cleaning cloth for jewellery, hand cream, a chocolate bar or a birthday card. In turn, these items are often not being returned since these items can be complementary to the product which they might need (the cleaning cloth for instance). Another reason might be that these items oftentimes are not very expensive and therefore are often being added to the cart to get to the minimum amount for free shipping. Since these items can have a low price and can be handy sometimes (the hand cream or chocolate bar for instance), these products are less likely to be returned by customers. Figure 6 shows the return probabilities of all Payment Methods compared to the reference level iDeal. As can be seen, the biggest effect on a high return probability is the payment method Afterpay. When customers use this payment method, the relative odds for returning a product would increase by 150.7% (Exp(β) = 2.507) compared to when the payment method would be iDeal. An explanation for this can be that when using Afterpay, you don’t already have to pay and therefore there can be a low commitment when paying with this payment method. Since customers don’t have this commitment while checking out, they might order more than when they actually already have to pay. Besides, they might also order products which they are not sure of or don’t necessarily need. Thus, when customers pay with Afterpay, they might already know in the back of their head they might return the products. Regarding the two most popular credit card types in The Netherlands, Visa and Mastercard, it is demonstrated that they both show an increase in product return probability compared to when iDeal is used. The explanation given by Afterpay might also be extended to these payment methods: customers might already know in the back of their head they might return the product when paying with Visa or Mastercard.
Figure 6: Return probability payment methods compared to iDeal
5. Discussion
The aim of this research is to investigate the effect of non-monetary promotions, more specifically gifts, on product returns. To investigate this, data from a European retailer has been used. The data used for this research only contained information about the Dutch web shop and did not encompass any data about the Dutch brick-and-mortar stores, the market places the company is active in or about the sales of the foreign web shops. Moreover, the timeframe from this data ranges from October 1st 2019 till March 15th 2020. The retailer has a lot of
non-monetary promotions throughout the year where they are giving away free gifts. Therefore, the data from this retailer is convenient and fitting to investigate the effect of gifts on product returns. Different sub questions have been posited:
1. What is the effect of gifts on the product return probability?
2. How does the effect of gifts on product return probability depend on whether the gift is based on basket value or based on a product?
3. How does the effect of gifts on product return probability depend on whether the gift and focal product are related to each other (e.g. same product category)?
These sub questions were being investigated by means of multiple binary logistic regression models and hypotheses. The results with regard to the hypotheses are summarized in Table 5.
Table 5: Summary results hypotheses
Hypothesis Supported?
H1: Offering a gift with purchase will lead to lower product returns. Yes
H2a: Offering gifts based on basket value will lead to lower product returns Yes
H2b: Offering gifts based on products will lead to higher product returns No, the opposite
H3a: Receiving a gift which is related to the purchased product relative to receiving an unrelated gift will lead to lower product returns.
Yes
H3b: Receiving a gift which is related to the purchased product relative to receiving an unrelated gift will lead to higher product returns.
No
To investigate the effect of different types of gifts, hypothesis 2 has been investigated. There were two types of gifts investigated: 1) when customers receive a gift based on their basket value, and 2) when customers receive a gift based on another product. Regarding the gifts based on basket value, it has been supported that the relative odds for returning a product decreases compared to when customers do not receive a gift based on basket value. This is in line with H2a. With regard to the gifts based on a product, it has been shown that the return probability also decreases compared to when customers do not receive a gift based on a product. This result is contradictory to H2b where it was expected that the return probability would actually increase. A possible explanation for this might be that the expected negative spill-over effect might not be that strong in this case. It might be that the loss aversion effect is stronger in the sense that customers prefer to keep the gift and thus not return the ordered product.
The last two hypotheses, H3a and H3b, were opposing hypotheses since the effect of product relatedness between the gift and the ordered product on product return probability could go both ways. One the one hand it could be the case that when the gift is related to the ordered product (e.g. same product category), customers might appreciate the gift more just as in the case of loyalty programs. On the other hand, however, it could be the case that unrelated gifts might create extra utility as it is not similar to the product the customer ordered. These hypotheses have been tested with regard to product relatedness on the three variables Total Gifts, Basket Related Gifts, and Product Related Gifts. With regard to all these types of gifts, it has been shown that when the gift is related to the product, it will lead to a decrease in product return probability. Thus, this supports H3a where it is being posited that when a gift is related to the focal product, it will lead to a decrease in product returns.
Some interesting findings have also been found for certain control variables. For instance, the return probability increases when customers order products: 1) during the weekend, 2) when they have more items in their basket and have a higher basket value, 3) which are categorized as sunglasses or gloves and 4) pay with Afterpay. Contrarily, the product return probability decreases when customers order products: 1) during the holiday period, 2) which are on sale, 3) for which they have to pay shipping costs, 4) when the temperature increases, 5) which are categorized as Others and Accessories, and 6) pay with gift cards or Bancontact.
5.1. Theoretical Implications
limits the validity construct of customers response of returning a scenario product versus a real purchased product by for example the opportunity costs to return a product physically. This research used real-life data from a European retailer and therefore increases the validity and generalisability on this topic.
Secondly, previous research hasn’t looked at the effect of the nature of the free product. Whereas this research also took into account the relation of product relatedness between the gift and the ordered product on product return probability. Previous research by (Lee & Yi, 2017) has only looked at the effect of gifts on product return and has mentioned the effect of product relatedness for future research. This research has partially found support that product relatedness moderates the relationship between gifts and product return probability.
Lastly, this study extends previous research by also investigating the effect of different types of gifts: gifts based on the basket value or gifts based on a product. This study found evidence that gifts based on the basket value and based on a product both reduce the product return probability significantly. Previous research has not looked at this distinction between gifts but looked at gifts in total (Lee & Yi, 2017, (Lee & Yi, 2018, Raghubir, 2004). With regard to the gifts in both studies by Lee & Yi (2017) and Lee & Yi (2018), customers had to retrieve an experience from their mind with regard to when they did receive a free gift when they bought something. Hence, it was not clear what the type of gift was in these studies. Moreover, in another study in their research, the researchers showed the participants a hypothetical case in which when they bought a focal product, they received a free gift. When extending this scenario to this research, it can be deemed similar to the Product Related Gifts mentioned in this research. Lee & Yi (2017; 2018) have found in both studies with regard to gifts given away with a product that customers were less likely to return the focal product than when they did not receive a gift. This is also confirmed in this research in the sense that Product Related Gifts will lead to a decrease in product return probability.
5.2. Managerial Implications
new brand. By doing this, they will create awareness for this new brand, and it will not harm the product return probability.
Secondly, it is also interesting to see when product relatedness matters for giving away free gifts. For all types of gifts, this research has found that when a product is related to the gift, it leads to a lower return probability. Hence, when retailers are offering free gifts, it is recommended to have a gift which can be related to the product. For instance, if the company only sells jewellery, it is recommended to also give a piece of jewellery as a gift rather than something totally unrelated, such as a beanie. This is especially recommended for gifts based on a certain product. Since this research has found that when the gift is given away by a certain product (e.g. Product Related Gift) and is unrelated to the focal product, it shows no effect on the product return probability. However, if it is related, it leads to a significant decrease in product return probability. Therefore, it is especially important for Product Related Gifts to have a gift which is related to the focal product in order to decrease the product return probability.
Thirdly, this study has found support that during the holiday period, products are being returned less often than when there is no holiday period. This is contradictory to the result found by Petersen & Kumar (2009), where it is being found that during the holiday period, products are being returned more often. This is due to the effect of gift-giving. Products being bought as a present for someone, are being returned less often, as also found by Petersen & Kumar (2009). Therefore, if a company is in a similar industry where a lot of products can also be given away as gifts, it might be an opportunity to focus more on these gift-giving events. For instance, communicate throughout the year also about gifts for birthdays, Mother’s Day, Father’s Day, etc.
Afterpay but paid with the gift card he or she still has, the customer then shows a return probability of 11.64%. These different scenarios show that the models used in this study can be easily used for managers to interpret the return probabilities for different scenarios.
5.3. Limitations and Future Research
This research also shows some limitations and recommendations for future research. The results of this research might be different for other retailers and industries. The product categories of this retailer are quite specific, and it might be the case that the effect of gifts might differ for clothing retailers for instance. As said before, the return rates differ per industry and therefore it might be interesting to see whether these results also hold for other industries. This is something future research can investigate to increase the generalisability across different industries.
Another issue may be that this dataset consisted mostly of female products. Therefore, in order to increase generalisability and validity about which type of products are being returned most (male, female or unisex items), more research needs to be done.
Moreover, some other control variables of product characteristics might also be taken into account for future research. This retailer also has photos of products together with how it looks on a person. However, these photos are not available for all products. Information about this type of data was unfortunately not available. But it might influence the return probability as people can make better informed decisions when they see how the product fits on a model. Therefore, future research should also take this into account in order to control for even more factors.
Another interesting research opportunity might be to see what people will do when they have a basket value below the amount of receiving free shipping. As shown in this research, customers are less likely to return the product when they have to pay shipping costs. However, as experienced by the retailer, sometimes customers order something extra just to get to the free shipping amount. Future research might investigate this in order to see how this will influence the product return probability.
having non-monetary promotions, it is likely that there will be a decrease in product return probability. However, if the sales skyrocket because of this promotion, it is interesting to see whether this decrease in product returns probability is still as strong when the absolute product returns might increase. Hence, whether it financially still leads to a strong decrease in costs for product returns. Moreover, another factor to take into account for this is the costs of these free gifts. Whether the costs of giving away these free gifts can be offset by the decrease in product return probability or not.
6. References
- Anderson, E.T., Karsten, H., DI, S., and Lei, K. W. K. (2009). “How price affects returns: The perceived value and incremental customer effects.” Unpublished Manuscript, Northwestern University, Evanston, IL.
- Babin, B.J., Harris, E. (2015). “CB7”. Cengage Learning, 7.
- Berman, B. (2006). “Developing an Effective Customer Loyalty Program”. California Management Review, 49(1), 123-148.
- Bijmolt, T.H.A., Dorotic, M., Verhoef, P.C. (2010) “Loyalty Programs:
Generalizations on Their Adoption, Effectiveness and Design”. Foundations and Trends® in Marketing, 5(4), 197-258.
- Business Insider (2020). “UPS expects a record-breaking 2 million package returns today, and it reveals a dark truth about holiday shopping,” (accessed April 1, 2020), [available at https://www.businessinsider.nl/ups-predicts-massive-returns-on-national-returns-day-2019-12?international=true&r=US ]
- Janakiraman, N., Syrdal, H.A., Freling, R. (2016) “The Effect of Return Policy Leniency on Consumer Purchase and Return Decisions: A Meta-analytic Review” Journal of Retailing, 92(2), 226-235.
- Kahneman, D., Knetsch, J.L., Thaler, R.H. (1991). “The Endowment Effect, Loss Aversion, and Status Quo Bias”. The Journal of Economic Perspectives, 5(1), 193-206.
- Kahneman, D., Tversky, A. (1979). “Prospect Theory: An Analysis of Decision under Risk”. Econometrica, 47(2), 263-292.
- Kaplan, M. (2019). “The Growing Problem of Customer Returns,” (accessed March 13, 2020), [available at https://www.practicalecommerce.com/the-growing-problem-of-customer-returns ]
- Kerr, J. C. (2013) “Serial returners, beware: Retailers are tracking you,” (accessed April 8, 2020), [available at https://www.today.com/money/serial-returners-beware-retailers-are-tracking-you-6C10900265 ]
- Lee, S., Yi, Y. (2017). ““Seize the Deal, or Return It Losing Your Free Gift”: The Effect of a Gift-With-Purchase Promotion on Product Return Intention”. Psychology of Marketing, 34(3), 249-263.
- Lee, S., Yi, Y. (2018) “Retail is detail! Give consumers a gift rather than a bundle”: Promotion framing and consumer product returns”. Psychology & Marketing, 36(1), 15-27.
- Minnema, A., Bijmolt, T.H.A., Gensler, S., Wiesel, T. (2016) “To Keep or Not to Keep: Effects of Online Customer Reviews on Product Returns”. Journal of Retailing, 92(3), 253–267.
- McFadden, D. (1974) “The measurement of urban travel demand”. Journal of Public Economics 3(4), 303-328.
- Orendorff, A. (2019) “The Plague of Ecommerce Return Rates and How to Maintain Profitability,” (accessed March 13, 2020), [available at
https://www.shopify.com/enterprise/ecommerce-returns]
- Petersen, J.A., Kumar, V. (2009) “Are Product Returns a Necessary Evil? Antecedents and Consequences”. Journal of Marketing, 73(3), 35-51.
- Powers, T.L., Jack, E.P. (2015) “Understanding the causes of retail product returns”. International Journal of Retail & Distribution Management, 43(12), 1182-1202. - Raghubir, P. (2004) “Free Gift with Purchase: Promoting or Discounting the Brand?”
Journal of Consumer Psychology, 14(1&2), 1-186.
- Raghubir, P., Corfman, K. (1999) “When Do Price Promotions Affect Pretrial Brand Evaluations?” Journal of Marketing Research, 36(2), 211-222.
- Rao, A.R., Monroe, K.B. (1989) “The Effect of Price, Brand Name, and Store Name on Buyers' Perceptions of Product Quality: An Integrative Review”. Journal of Marketing Research, 26(3), 351-357.
- Reagan, C. (2019) “That sweater you don’t like is a trillion-dollar problem for retailers. These companies want to fix it,” (accessed April 9, 2020), [available at https://www.cnbc.com/2019/01/10/growing-online-sales-means-more-returns-and-trash-for-landfills.html ]
- Roehm, M.L, Pullins, E.B., Roehm JR, H.A. (2002) “Designing Loyalty-Building Programs for Packaged Goods Brands”. Journal of Marketing Research, 39(2), 202-213.
- Samuelson, W., Zeckhauser, R. (1988) “Status Quo Bias in Decision Making”. Journal of Risk and Uncertainty, 1(1), 7-59.
- Statista (2019) “Costs of return deliveries in the United States from 2016 to 2020,” (accessed March 11, 2020), [available at
https://www.statista.com/statistics/871365/reverse-logistics-cost-united-states/] - Thaler, R. (1980) “Towards a positive theory of consumer choice”. Journal of
Economic Behavior and Organization l, 39-60.
7. Appendices
Appendix 2: VIF-scores
Variables Model 1 Model 2 Model 3 Model 4
Totalgifts 1.122 1.975
Basket related Gifts 1.187 2.096
Product Related Gifts 1.114 2.811
