Are female skins sold at a lower price? Evidence from the Fortnite game

ABSTRACT Much significant research has been done to shed light on discrimination of females in, for example, labour markets. Less is known, in contrast, about the amount of discrimination in the virtual world of online gaming. In an early study, Castronova (2004), using data for the EverQuest game, finds that female avatars receive about 10% lower prices in online auctions. In this research note, we re-examine the pricing of avatars using data for the Fortnite game and find small and statistically insignificant price difference between female and male avatars.


I. Introduction
Building on influential earlier work by Castronova (2004), we study data for Fortnite, an online survival game produced by Epic Games and People Can Fly, to re-examine the role of gender for the pricing of avatars in online gaming. Specifically, we test for a price difference between female and male avatars. An important difference between our work and the earlier research by Castronova (2004) is that he studies avatars of the EverQuest game, the prices of which were built in auctions where the player of the game acted as buyer and seller. In the Fortnite game, the prices are determined by the producer in the in-game shop.
In the virtual world of online gaming, gender discrimination may show up, for example, in a tendency of online gamers to opt for a male rather than a female virtual character (avatar). In fact, Castronova (2004) documents that male users are more prone to selecting their main character as male instead of female. Moreover, he reports that the auction price of female avatars is about 10% lower than the auction price of male ones, and concludes that female avatars are traded by a primarily male player base at a discount, where the difference between the prices of avatars of different gender is statistically significant.
In this research note, we re-examine the role of gender for the pricing of avatars in online gaming and test whether there exists a price difference between female and male avatars. To this end, we study data for the Fortnite game taken from the website of Skin-Tracker 2020. The data consist of the name, definition, and the prices of all listed outfits (that is, avatars). Moreover, with the help of photos provided for each outfit, we can control for physical appearances of avatars, such as gender. Additionally, we observe the prices labelled for each outfit in V-Bucks, the virtual currency allowing gamers to buy the equipment or costumes in the in-game shop.
In contrast to the results of Castronova (2004), we do not find a statistically significant price difference between male and female avatars. Based on the outfit data, we find the difference between the average prices for male and female avatars to be rather small. Taken together, our empirical results show that, while the prices for female avatars are lower than the prices of male avatars, this difference is not statistically significant.
The virtual currencies used in online-gaming platforms and the main determinants of prices in in-game shops, where avatars are bought or auctioned by online gamers, have become under scrutiny in recent empirical research (see, e.g. Cerulo 2011;Nagy and Bernadett 2014;Xu et al. 2017). For this reason, we also study the pricing of fantasy outfits (that is, nonhuman outfits) and observe a significant positive price difference relative to male outfits. Moreover, we show that the time of the outfit release has a significant effect on its price, which is in line with the findings of Castronova (2004).

II. Earlier literature
Much research has been done in recent years to study whether females are discriminated in the workplace, as witnessed by the large gender-gap literature. Against the background of this research, an interesting question is whether females are discriminated in the virtual world of online gaming. Discrimination may show up, for example, in a tendency of online gamers to opt for a male rather than a female virtual character (avatar). In fact, Castronova (2004) reports that male users are more prone to selecting their main character as male instead of female. Moreover, he reports that the auction price of female avatars is about 10% lower than the auction price of male ones, and concludes that female avatars are traded by a primarily male player base at a discount, and that the difference between the prices of avatars of different gender is statistically significant. Lucas and Sherry (2004) examine gender differences in video game play by using a questionnaire survey among gamers. They document differences between male and female gamers and report that female gamers report less frequent play, less motivation to play in social situations, and less orientation to game genres featuring competition and three-dimensional rotation. Dickerman, Christensen, and Kerl-McClain (2008) analyse gender and racial stereotyping in video games. They find that depictions of women in video games are scarce and highly stereotypical. Women are usually minor characters, are seen as victims rather than heroines, and are depicted in highly sexualized ways. In later research, Lynch et al. (2016) perform a content analysis of female characters in video games  and analyse more than 500 titles. Results indicate that sexualization has diminished since an observed height in the 1990s. Despite an increase in games featuring playable female characters, games still depict female characters more often in secondary roles and sexualized them more than primary characters. Jennings and Messer (2018) conduct and experiment with children (aged 7 to 14) and find that the gender of the player is one factor which is strongly associated with the choice of avatar. However, the willingness to pay to play with the same gender avatar is very limited. With respect to this willingness to pay, no gender-specific differences are detected. This finding questions the pricing strategy of game developers -documented by Messer (2015). Messer (2015) shows that the default avatars are mostly male and that, for female avatars, price premium are charged.

The data
We collected the data set from the website Skin-Tracker in June 2020. This website listed a total of 717 avatars, where newly released avatars were depicted at the top of the older ones. All avatars had a name and were accompanied by a description attached to the photos also shown on the website. Figure 1 presents some of these Fortnite avatars.
Although most avatars appeared as 'Item Shop' (they have a price denominated in terms of V-Bucks), some appeared as 'Leaked Skins' (new and unreleased cosmetics, see Hamstead 2020; Pro Game Guides 2020) or 'Promotional" (cosmetic items that are awarded as a result of a promotion, see Gamepedia 2020). Cosmetic items did not have a price labelled with them and, hence, were not sold in the in-game shop. Therefore, we excluded 'Leaked Skins' and 'Promotional' from our data and focused on the avatars that have a price determined in V-Bucks.
Although there are various types of avatars in the in-game shop, their V-Bucks prices did not differ much. There are mainly six different prices in our data: 800, 1200, 1500, 1600, 1800, and 2000 V-Bucks. Because avatars with a price of 1600 and 1800 V-Bucks were sold only in one season, we did not include them in our empirical analysis. The final version of our data set comprised 548 avatars along with their names, descriptions, and prices.
We categorized avatars according to their gender based on the photos depicted on the web page: We defined three categories: male, female, and fantasy. We sorted avatars that we could not subsume under the male or female categories into the fantasy category. In total, our data features 145 males, 185 females, and 218 fantasy characters (Table 1).
The average prices for males, females, and fantasy avatars in our data are approximately 1199, 1181, and 1472 V-Bucks. While the average price of fantasy avatars is higher than the average prices of female and male avatars, there is no substantial difference between the average prices of male and female avatars. Hence, the descriptive statistics do not lend support to the view that online gamers discriminate between female and male avatars.

Regression models
In order to study our data in more detail, we estimated the following regression equation by the ordinary-least-squares technique: The dependent variable is the price of an avatar, i, denominated in V-Bucks. The dummy variable 'female' is one for females, and zero otherwise. Similarly, the dummy variable 'fantasy' is one for fantasy avatars, and zero otherwise. The estimation results (Table 2) show that the price of female It is not always easy to classify avatars unambiguously. For example, 'Boxer' could be classified as fantasy. But since, 'Boxy' can be classified as female, we classified 'Boxer' as a male. Source: https://skin-tracker.com/ avatars is 18 V-Bucks lower than the price of male avatars. This price difference, however, is not statistically significant. In contrast, the price of fantasy avatars is 273 V-Bucks higher compared to the price of male avatars, and this difference is statistically significant.
Next, we tested whether newly released avatars are sold at a discount (as found by Castronova 2004). To this end, we extended our regression equation to include a variable that captures the release time of an avatar: where the variable 'time' assigns consecutive numerical values to every observation according to the time an avatar was released by Fortnite. We sorted the observations such that the older releases appear before the newer ones and the assigned 'time' value increase for every newly released avatar. The estimation results show that older avatars are sold at a higher price in the in-game shop, that is, the price of avatars has decreased over time. While female avatars are sold at a lower price of roughly 7.6 V-Bucks compared to male ones, this outcome again is not statistically significant. Fantasy avatars are priced that are approximately 287 V-Bucks higher than the prices of male avatars.
When interpreting the results for the variable 'time', it is important to note that Fortnite (unlike the online platform analysed by Castronova 2004) does not use auctions to selloff avatars. Rather, it sells all avatars at a predetermined prices, that is, it lists all prices in V-Bucks and players buy an avatar at these prices in the in-game shops. The decline of prices of avatars over time, thus, can be interpreted as evidence that Fortnite applies a quasiskimming strategy. A skimming strategy implies that a product is sold at a high initial price, and that its price declines over time due to customer satisfaction and intensifying competition. The pricing of avatars by Fortnite can be viewed as a quasi-skimming strategy because, while the price of a single avatar stays constant, the prices of newly released avatars decrease over time, that is, new outfits are sorted into lower price categories.
It also is worth noticing that, in the Fortnite game, players do not have to choose an avatar once and for all. In EverQuest, the game that Castronova (2004) has analysed, choices of characters, in contrast, are permanent. The potentially temporary choice of an avatar in Fortnite may imply that players, even in case the player base is mainly male experiment with avatars and switch back-and-forth between genders, so that there eventually the data we analyse do not show signs of a significant gender-price gap.
Like Castronova (2004), we next re-estimated our regression model with the dependent variable being measured in terms of logs. The estimation results corroborate that there is no statistically significant price discrimination between female and male avatars, and that new avatars are sold at a discount relative to older avatars. Furthermore, fantasy avatars receive higher in-game shop prices relative to male avatars. To be more specific, in the model where we take into account the release time of an avatar, female avatars are sold at a discount of only À 0:5% relative to male avatars, the price of newly released outfits is seen to be lower relative to the price of older outfits.

Fortnite scandal
With Fortnite being a popular online-gaming platform, an incidence concerning the design of avatars has received much public attention. Specifically, after the release of Season 6, some players noticed a change: 'Epic added boob jiggle physics' (Wilhelm 2018) to some of their female characters. 1 Epic reacted fast and patched the problem. However, Wilhelm (2018) regards Epics' communication policy as strange, because Epic apologized by arguing that the addition of the bonus physics was careless and accidental. He argues that it took time and effort to create the jiggle and that this surely did not happen accidentally.
When such an incidence takes place, a company might set up a task force and crosscheck several of its procedures. This might affect subsequent pricing decisions in an attempt to avoid further incidences and scandals. Hence, in order to inspect whether the Fortnite scandal had a significant impact on price differentiation between female and male avatars, we separated the avatars released before Season 6 from the ones released after Season 6. Studying the pricing before and after the scandal separately accounts for the possibility that the scandal resulted in a structural break in the pricing. Such a structural break, in turn, easily could result in biased estimation results. In order to show that our results are not subject to such a bias, we report in Table 3 the estimation results for regression models that we estimated separately using data before and after the scandal.
The estimation results show that, before the scandal occurred, the price difference between male and female avatars was larger than in the full sample ( þ 34:8 > þ 18:1 V-Bucks). After the scandal, male avatars were sold at a lower price compared to female avatars and the price difference took on a value of À 16:2. Moreover, before the scandal occurred, the total numbers of females, males, and fantasy avatars were 73, 70 and 78. After the scandal had gathered steam, the number of female avatars increased to 112 whereas the number of male avatars increased only slightly to 75. In sum, Fortnite reduced the price of male avatars more strongly than the price of female avatars, so that the price difference has become negative. The subsequent increase in the number of female avatars may even further helped to even out the price difference that existed before the scandal broke out.
In order to analyse more formally whether the impact of being a female avatar on the in-game shop prices relative to the prices of male avatars changed after the scandal in Season 6, we created new dummy variables for the two genders that capture the average price differences among every gender before and after Season 6. The estimation results (Table 4) confirm that the price of male avatars were lowered to a larger extent than that the female avatars. The price  difference between genders is not statistically different. Hence, we do not find evidence of price discrimination between female and male avatars. When we specified our regression model in terms of logs of the dependent variable, the estimation results were consistent with the results for the anti-log specification.

IV. Concluding remarks
This research note contributes to the literature that studies discrimination in the virtual world of online gaming. In earlier research, Castronova (2004) finds that the gender of avatars matters for the auction pricing of avatars, and that female avatars are sold for significantly lower prices than male avatars. Upon analysing data for the online game Fortnite, we also have found that female avatars are sold at lower prices relative to male avatars, but this price difference is small and statistically not significant. In contrast, the release timing of avatars and whether avatars are released as 'fantasy' outfits have a significant impact on the pricing of avatars.

Disclosure statement
No potential conflict of interest was reported by the author(s).