Gold Farming Research – Part Two

You are reading the series of the research of Gold Farming.
• Preface
• An Overview of Gold Farming

An Overview of Gold Farming

The Origins of Gold Farming

Gold farming is a modern fraction of the relatively much older phenomenon of “real-money trading” (RMT); traced back to 1987 and the first cash payments between players for items or for improving characters within text- and basic graphics-based multi-user dungeons (Hunter 2006)¹. RMT trundled on during the 1980s and 1990s but seems to have remained a phenomenon of the global North.

This began to change at the end of the 1990s for reasons related to two different markets. First, MMORPGs really began to grow in industrialised countries with the launch of Ultima Online in 1997. Alongside that growth was a strong take-off in real-money trading; facilitated by the founding of eBay (Lewis 2006). Whenever a new online game was launched, items would be available for sale on eBay within a few weeks.

As discussed in more detail in Section C1, the sale of in-game items and currency was initially a cottage industry of players-turned-traders based in industrialised countries. In 2001/2002, though, this changed as mainly US-based traders perceived the opportunities for outsourcing to low-wage labour locations such as Mexico and East Asia. Some of these traders were North Americans of Asian origin, or Asian students studying in the US (Jin 2006).

This “outside-in” model met up with an “inside-out” model that lay behind the second reason for growth of developing country gold farming: serving the domestic/regional market. There was a particular growth in online games-playing in East Asia from the late 1990s.² The games played were initially foreign imports such as Blizzard’s Starcraft but in 1998 the South Korean firm NCSoft launched Lineage, the first significant non-industrialised-country MMORPG. The game soon spread to China and Taiwan and Japan. In-game traders began paying those around them to earn “adena” – the Lineage in-game currency (Chan 2006). As Lineage began being played outside East Asia (particularly when Lineage II was launched on American servers in 2003), and as East Asians increasingly played Western MMORPGs, the “adena farmer” model spread. It seems likely that it was at this point – from late 2003/early 2004 – that the number of gold farms rose from dozens to hundreds and then thousands (Jin 2006b, PJ 2007).

Summarising gold farming’s origins, it was likely in existence from at least 1998 but under (Western) radar. Earliest recorded activities include:
• “Inside-out” model: Korean cybercafés being converted into gold farms to serve the local market in 2001 (Huhh 2008); Chinese gold farmers working for Korean and Japanese players from 2001 (Jin 2006)³, and
• “Outside-in” model: a US firm outsourcing to Mexican gold farmers in 2002 (Dibbell 2003).

Earliest usage of the terminology that I have been able to locate dates to 2003 on TerraNova (Kiblinger 2003) but it must go further back than that.

Activities of Gold Farmers

What do gold farmers actually do? They sit at an Internet-connected PC and they play an online game for the purposes of making real money. Gold farmers make real money in three main ways, listed in what are likely to be (Carless 2007, Robischon 2007)⁴ decreasing order of activity and earnings:

• Selling in-game currency: akin to purchasing real-world foreign exchange. The currency is purchased via a web site at a clearly stated real-to-virtual currency exchange rate, with payment typically made via PayPal or credit card. Then the purchaser is messaged and told to “meet” one of the gold farming avatars in game, and the in-game currency is transferred. For example, in mid-2008, you could buy 5m Runescape gold for around US$20.
• Power-levelling: payment is again made via a web site but this time the gold-farming firm is provided with the purchaser’s game username and password. Their staff then “play” the purchaser’s character in the game, building up its levels (of combat or other skills). Once the character has reached the agreed level, it is handed back to the purchaser. For example, in mid-2008, it would cost around US$50 to have your World of Warcraft level-40 character power-levelled to level-60. Very much related though much less often offered is account-selling; where the gold-farming firm sells a high-level character that they have made: a level-75 Final Fantasy XI account would cost around US$400.⁵ Likewise, quest-completion services are offered: the gold-farming firm will take over your character and play it to complete a particular in-game activity such as a quest; completing the Runescape Lost City quest, for instance, costs US$15.
• Selling in-game items: the transaction occurs in very much the same way except that an item rather than currency is transferred. Individual items have sold for as much as US$20,000 (Davis 2006). A more typical example, though, would be that you could buy a Star Wars Galaxies AT-RT in mid-2008 for US$10.⁶

Beyond the basic “gold farmer” epithet are many different roles that gold farmers may play in-game (expanded from Davis 2007 and Zonk 2007):
• Hunter: these players kill non-player characters/”monsters”. They will then take any currency dropped from the kill, and sell valuable dropped items in-game for more currency.
• Fighter: helps the hunter to kill NPCs but does not take the drop.
• Gatherer: undertakes a non-fighting role in the game to gather resources that can be sold for in-game currency, such as mining or wood-cutting.
• Producer: undertakes a non-fighting role in the game to manufacture items that can be sold for in-game currency, such as blacksmithing or potion-making.
• Banker: stores assets; may also “mule” the assets from one area of the game to another (e.g. a bank or trading location).
• Dealer: delivers the currency or item to the purchaser in the game.
• Marketer: “barkers” who generally advertise gold farming services to other players; “peddlers” who contact individual players with advertising messages.
• Leveller: takes over a purchaser’s character and plays it in order to raise it levels (or plays a firm-owned character, raises its levels and then sells that account).
  In practice, a gold-farmer’s avatar may play more than one of these roles (e.g. both marketer and dealer: Bell 2006). Gold farmers may play these roles alone or work in groups (e.g. Jin 2006d, Dibbell 2007). Some of these roles (e.g. barker) seem more likely to be automated than others

Some field evidence from gold farms – such as Ge Jin’s (2006c) videos – suggests everyone within a gold farm plays online; though potentially in the different roles just described. However, that may be true only in those gold farms that sell through an intermediary. Other evidence suggests greater division of labour, at least within those firms that sell direct to players. One entrepreneur reports that of his 20 staff:

“Three people do research and find the most efficient way of gold farming and power leveling, three people who are good at English do marketing and the rest are divided into four groups, the best gold farmer as group leader, farming gold.” (Wang 2008)

We can summarise all this using the internal value chain model, as shown in Figure 1 (adapted from Porter 1985).

Gold Farming Firm Internal Value Chain

Taking a small gold-farming enterprise, then there are three support roles:
• Management: just one or two people (typically the owner-managers) would handle the administration, management and HR management tasks.
• Research: this could be a separate role, as described in the quote above (we should perhaps call these “gold horticulturalists”). But in other cases, firms rely on their gold farmers’ knowledge of the game for their “research” including on occasion the playing time their staff invest outside their working hours (Dibbell 2007).
• Technical: a technical staff member would cover technology management and procurement, or that would be outsourced to a cybercafé if the gold-farming was based there rather than at dedicated premises.
• Customer relations: responsible for all the non-operational primary activities that involve contact with potential and actual customers.

For those gold farmers that sell direct to customers (as opposed to those that go via brokers and exchanges), marketing is undertaken via the web (e.g. messages to gamer communities, or adverts on key web sites) or in-game. When a customer makes an approach, it is handled by customer reps through email, phone and/or online chat to understand what is required (see e.g. Admin 2007). A payment is then made. Gold farmer staff then produce the item, currency or service. In many cases, actually delivery happens separately. After-sales “service” would normally only consist of follow-up messages to past clients to see if they wished to purchase more.

Without getting inside a gold-farming enterprise it is difficult to provide more than this basic description. Field research would be needed to identify strengths and weaknesses of the ten value chain areas, and to look at balance of effort between the areas.

Mapping the Gold-Farming Sub-Sector

Estimates of the size of the global computer games market vary considerably but focusing on just the games (i.e. not consoles or other hardware), one may estimate that the global market for the five game types (arcade, console, PC-based, online and mobile) was around US$33bn in 2008 (The Hindu 2007, Jin & Chee 2008). There were perhaps some 50m online games players worldwide of whom around 20m were playing monthly subscription online games (White 2008, Woodcock 2008). Extrapolating up from average subscription fees of US$10⁷, the latter figure would suggest something in the region of US$2.5bn per year subscription income.

There appear to be very few estimates framed in terms of the size of the gold farming sub-sector. Instead, most estimates are for the value of real-money trading (see Table 1). As noted above, these would be different if the RMT figures include “primary real-money trading” i.e. that conducted within or alongside games and sanctioned by the game company. A best guess is that the RMT figures cited here in fact exclude the primary market but only Lehtiniemi (2007) is explicit on this and better data-gathering and analysis are required.

Data from Yee (2005) and Castronova (2006) leads to an estimate of average spend on gold farming per Western player of something like US$10 per year⁸. Simple extrapolation from the 20m subscribed gamers, then, would suggest total market revenue for gold farming of US$200m per year. That is arguably an underestimate for a number of reasons.

First, if we stay with the same notion of 20m gamers, we can use the data produced by Everquest II’s Station Exchange to give two other estimates. Average annual spend per player (all players, not just RMT-active) was US$46. Multiplying up from global subscription numbers would produce a total global market of US$920m. Attacking from a different direction, the spend (US$1.87m) was 29% of that on subscriptions (40,700 players x US$13 average subscription per month x 12 months = US$6.35m). Applying that same percentage to the estimated annual total subscription income of US$2.5bn would suggest a global RMT market of US$725m.¹⁴

Second, we can step outside the 20m online game subscribers paradigm in three ways:

• Not all those subscribers are Western and we know little about the extent of gold-farming purchases among Asian players. However, the figures in Table 1 suggest that gold-farming trade within Asian nations may be higher, not lower than that in the West. Lehtiniemi’s (2007) estimate of US$1.6bn, for example, is well-considered and analyses the Western market to be only 18% of the global total. The 20m figure also excludes those (another 20m) playing free-to-play games; nothing seems to have been written about gold farming on such games but it is likely to exist and could easily exist at the level of subscription game purchases. Second Life and other non-game worlds make up the remaining 10m of online players. These allow and integrate real-money trading, so – again – we know little about gold farming for these virtual worlds.
• If we take the minimum estimates (see below) for employment in gold farming of 150,000; monthly salary of US$145; and minimum overhead of 100% on salary; that produces a total overall market of US$520m per year. Using the more likely estimate of 400,000 employees, we arrive at a total overall market of US$1.4bn.
• Unpublished research from two US game companies suggests the potential real-money trading market to be eight times greater than that generated by subscriptions (Eikenberry 2008). Those estimates do not tally with the actual experience of Station Exchange, but they do suggest a potential market of US$20bn per year.

In particular, putting the first and second bullet points together – incorporating the larger Asian market, the large potential in free-to-play games, and employment data – the best guess here will be that gold farming is a US$500m per year sub-sector. It seems unlikely to be less than US$200m per year, and it could well be more than US$1bn per year.¹⁵
No clear chronological trends are can be drawn out with confidence from Table 1 but there is little to suggest the gold-farming sub-sector is shrinking over time. If anything, the data leans slightly in the opposite direction though we do have a serious gap in post-2006 data.

Location
Location
Figure 2 (map from Johomaps 2008) shows the various global locations that have been mentioned as gold farming sites¹¹. Looking at the map, one might be struck by how much the locations resemble those for “nearshoring”: the outsourcing of work to locations close to major markets. There is some validity to this – the Mexican example was set up by a firm just over the border in Southern California (Dibbell 2005); Chinese gold farmers have particularly served the South Korean and Japanese markets (Jin 2006). In other ways, though, the trade appears globalised – all locations, for example, serve the US market – and location appears based on the presence of an online gaming community, broadband ICT infrastructure, and connections to major markets which can easily be independent of physical geography.

Figure 2: Reported Global Locations of Gold Farming

There is a generalised assumption that the great majority of gold farmers are based in China. In the absence of any better evidence, we will go along with this and guesstimate that China has around 80-85% of employment and output in this sub-sector.¹⁴ Figure 3 (map from AZGA n.d.) shows the locations in China that have been mentioned as gold farming sites¹⁵. Looking at the map, one might be struck by how much the locations resemble those in which foreign investment and econodevelopment in China has mainly taken place: the eastern and southern coastal provinces (Ogutcu & Taube 2002). This is probably explained by three factors listed above: relative wealth meaning more online gamers in these provinces; foreign, government and personal investment in these areas creating a stronger ICT infrastructure; and wealth/trade fostering good overseas connections. In addition, these are the areas of rural in-migration, creating a pool of potentially-employable labour.

What is unclear, on both a global and Chinese provincial basis is whether the blank areas on the map are blank because there is no gold farming, or simply because no-one has done any field research on gold farming in those areas.

Figure 3: Reported Locations of Gold Farming in China

Narrowing the locational lens one step further, there are reports of urban-based entrepreneurs being interested in moving operations to rural areas in order to cut costs (e.g. He 2005), and there are plenty of assumptions about gold farming being rural (e.g. Warner & Raiter 2005, Bearup 2006). In practice, there have been no reports from rural villages, and just a couple of reports from rural towns with populations of some tens of thousands (Thompson 2005, Johnson 2006). By and large, reports have come from cities as have (see below) characterisations of worker demographics. It thus seems reasonable – at least until ICT infrastructure diffuses more fully – to label gold farming an urban phenomenon (Jin 2006b).
Employment and Enterprise Numbers
Because no-one looks to have undertaken systematic measurement, employment and enterprise statistics vary and typically give no explanation for their basis. For example:

• Employment: in China, chronological estimates of employment are 100,000 (Barboza 2005), 200,000 (Jin 2006/Lewis 2006 from same individual source), 100,000 (Dibbell 2007), 500,000 (Neff 2007), 400,000 (Wang 2008). Alongside these, Kushner (2007) estimates 1 million gold farmers worldwide.
• Enterprises: in China, estimates of the number of gold-farming enterprises – known as “gaming workshops” – are: 30 in 2004 (Carless 2007), 200 in 2004 (Russell 2004 cited in Steinkuehler 2006), 2,000 in 2005 (Lewis 2006), 2,000 in 2007 (Carless 2007).

A simple average for the employment figures suggests just over 400,000 gold farmers employed worldwide if we use the guesstimate that China has around 80-85% of employment in this sub-sector.

The enterprise figures given appear to be an underestimate. Johnson (2006) reports dozens of gold-farming operations in a 120,000 population town in NE China. Jin (2006) found hundreds of gold-farming operations in a 2.5 million population city in eastern China. This suggests a minimum of about one gold-farming firm per 10,000 population¹⁶. If we extrapolated that up to the whole of China, it would suggest at least 130,000 firms throughout the country in 2006. A more appropriate extrapolation, though, might be to the total population of just those provinces in which gold-farming operations have been identified¹⁷. That would suggest around 50,000 firms in China.

We can then convert this into estimated employees using average enterprise size figures. Average IT sector enterprise size in industrialised countries is about five employees (NGRF 2006) and (see Section D2) field reports talk of most gold-farming enterprises with 2-10 PCs. However, a more conservative converter might be the average size of urban informal sector enterprises in Asia: 2.5 employees (Ranis & Stewart 1999). If valid, that would suggest about 125,000-325,000 people employed in gold farming in 2006 in China alone depending which of the enterprise extrapolations above we accept. There is little evidence for change over time so, a best guess for 2008 might be a minimum of 60,000 gold-farming firms worldwide (most of them micro-enterprises with just one or two staff) employing 150,000 people; but with 150,000 firms employing 400,000 people equally possible.

The latter employment figure is perhaps more likely. It is a better fit to the sourced figures given above, including Wang’s (2008) on-the-ground report that there were 20,000 gold farmers in Harbin alone, a city region with a population of 10 million. It also allows for the 60,000 enterprises figure to be correct with a less-conservative estimate of 6-7 employees per average firm.

In terms of the demographics of these employees, they appear to fall into three types (He 2005, Jin 2006b, Dibbell 2007, Wang 2008):

• Migrants from rural areas looking for work in urban areas.
• Young unemployed people based in urban areas.
• College students, who would work only part-time (and study part-time).

Many in the latter two categories would already be game players (Jin 2006, Jin 2006e). Female gold farmers do appear in accounts (e.g. Paul 2005, Yee 2006), but in practice the profession seems overwhelmingly male. Bell (2006) reports a 93%:7% male:female ratio, while Jin’s (2006) field survey in China reports only male gold farmers.¹⁸ In age terms, all reports (Bell 2006, Jin 2006, Yee 2006) state a range from late teens to early 20s, consistent with Barboza’s (2005) more specific 18-25 years range.

There is no direct data on investment levels but one can estimate a fixed investment of around US$800 for hardware and operating software per two workers (given many gold farms seem to work a two 12-hour shift system) (Nystedt 2007). Taking the 400,000 employment figure above, that would suggest a total investment of US$160m excluding premises. Actual per-capita investments would be higher, a) because not all PCs are shared by two workers; b) because firms also make fixed investments in telecommunications and games software; and c) because some (probably only the larger) firms make fixed investments in their web sites and marketing. However, such additional investments would be unlikely to more than double the original figure, so we can reckon that gold farms require a fixed investment per job of less than US$800. This compares favourably with, say, investment per job costs in manufacturing in China, which are quoted as at least 3-4 times higher for township enterprises (World Bank 1998) and dozens of times higher for urban processing plants (ADB 2003). In terms of employment creation, then, gold farming may be an efficient use of scarce capital in developing countries.

Pay and Work for Individual Gold Farmers
Gold farmers simultaneously play and work; hence the some-time title “playbourers” (Chan 2006). We can divide playbourers into amateur and professional. Amateurs are players who make some additional income from gold farming, but for whom it is not their main source of income or (for students) main activity. They will typically gold farm to cover the costs of their game play, or for pocket-money, or on an occasional basis (e.g. when they find a rare item or amass more in-game currency than they require or get tired of a game and sell their characters) (see, for instance, Jin & Chee 2008). Professional playbourers are those who make their main income from gold farming.
For the latter group, we can characterise their work via four main descriptive categories (Heeks 2008):
i. Pay. Gold-farming makes headline news in part when individual farmers make huge incomes. One Chinese student was arrested in Japan in 2006, having made US$1.3m over a two-year period, US$850,000 of which he had remitted back to China (Davis 2006b). The reality for most gold farmers, however, is a lot more mundane. They are employed by an enterprise, and they receive a pay packet.

Table 2 indicates the reported pay for gold farmers in China from various sources which are listed either as “real” where the evidence comes from face-to-face questioning, or “virtual” where interviews were by phone or email. Where known, the specific location is indicated, plus any salary comparators. Taking an average, we can say that pay appears to be about 1,000 yuan/US$145 per month. There is no clear trend within the data over time: other than the single-sourcing for 2004, yearly source averages are: US$137 (2005); US$139 (2006); US$130 (2007).

A few motes of other data have emerged, as shown in Table 3. The Russia figure is either out-of-kilter or indicates why this might not be a competitive location once Chinese gold farming took off. The remaining figures suggest other locations might still be competitive with China, and that – not surprisingly – managers earn more than workers in gold farms.

There appear to be four approaches to payment:
• Regular wages: these are merely an implicit assumption within some reports; in practice, it is not clear how widespread – if at all – such payment is.

Piecework: workers are paid by results; for example, Dibbell (2007) reports workers in China paid 10 yuan/US$1.25 for every 100 gold coins they gather in-game in World of Warcraft.

• Quotas: workers are set a quota that they must achieve during their shift. For example, Sage (2004) reports Chinese workers having to gather 1 million gil during their shift on Final Fantasy XI (see also Yee (2006), Bell (2006) and Steinkuehler (2006)). Quotas may be linked to a bonus payment (PJ 2007).
• Targets: workers are set a target to achieve during their shift but these are less directly linked to positive or negative consequences (Paul 2005).

One would predict downward pressure on wages from in-sector factors (Heeks 2001). Competition between enterprises is strong as there are few entry barriers into the sector. Production and overhead costs cannot be squeezed; and overhead costs may rise as game companies take more aggressive action against gold farmers. Gold-farming firms operate in a grey area that lacks regulation or organisation of labour. As noted below, training requirements for new worker entry are small, thus creating a competitive labour market through the large pool of reserve labour competing for jobs.

ii. “Fringe benefits”
Meals and dormitory accommodation are reported as included from field research (Barboza 2005, ARD-TV 2006, Jin 2006, Dibbell 2007). There is a report of poor hygiene and poor food quality in some gold farms (PJ 2007).

iii. Working hours
An assumption of two-shift working, 12 hours per day, seven days per week has arisen in relation to gold farming. Data sourcing in Table 2 indicates those listing shorter hours all conducted their interviews at a distance, whereas those who visited the gold farms describe 12/7 working. Breaks are allowed within the shifts (Jin 2006), and at least some gold farmers report getting sick leave or time off when required (though that is doubtless unpaid) (Paul 2005).

iv. Skills and training. Training requirements for entry into gold farming appear limited: He (2005) quotes two days. However, in the overall scheme of developing country employment it would be inappropriate to describe these jobs as “unskilled”; better to call them at least “semi-skilled” since they require a set of IT skills that most developing country workers lack. Those playing high-level characters or installing automation game code would approach or enter the “skilled” category.

Other aspects of the working lives of gold farmers are discussed in Section E1 below.

The Consumer Market
Discussion and comment on gold farming has centred largely on the producers. The consumers are rarely mentioned and could usefully be the topic of more research¹⁹. We can make a rough estimate of their numbers using two methods²⁰:
i. Player surveys. Yee’s (2005) survey showed 22% of MMOG-player respondents had bought in-game currency²¹. From Woodcock (2008) and White (2008), we can estimate 20m active subscribers in mid-2008. That would suggest 4.4m consumers buying gold farming services. This is likely a lower boundary since it is not clear to what extent the 20m free-play, item-pay gamers make use of gold farmers. As noted above, this could easily double the figure to 9m consumers.
ii. Financials. Reusing the calculation above of 150,000 minimum employed earning US$145 per month with salaries at most half of total revenues, that produced a market size of US$522m. Taking figures (Yee 2005, Castronova 2006) which suggest those who do buy spend an average US$45 per year, that would suggest 11.6m consumers.
It is assumed that customer location follows the demographics of game play. One corroboration of this comes from traffic analysis (via http://www.alexa.com) of gold-farming web sites. For example, 50-70% of traffic on four well-known English-language sites – www.ige.com, www.brogame.com, www.thsale.com and www.wowmine.com – comes from the US. The next three locations (each with a few percentage points share) are the UK, China and Canada. As would be expected,Chinese-language sites (e.g. www.5173.com) derive the vast majority of their traffic from China.

1: Maxson (2006) claims the practice dates back to the early 1980s with item and character trading on the game Avatar, which was written in 1979.
2: This growth was enabled generally by growing ICT (especially broadband) infrastructure, IT skills and incomes. It was enabled specifically by spread of cybercafés that provided group locations for young people, and micro-payment methods (Martinsons 2005). In turn, a number of these can be traced to the Asian currency crisis of 1997. Governments sought to spend their way out of this by investing heavily in ICT infrastructure. Some of the unemployed set up new businesses – either the PC kiosks in which games could be played, or actual games companies. Others among the unemployed turned to games-playing to fill the empty hours. Other push factors include leisure time for urban youth and spreading awareness of games. Pull factors include availability of attractive games.
3: PJ (2007) claims Koreans travelled to China in 2003 to train Chinese workers to gold farm.
4: Figures from Robischon (2007) for Everquest II suggest 59% of RMT expenditure on currency, 38% on characters, 3% on in-game items. See also Tae-jong (2006) who cites estimates that 60% of “profits” (most likely a miswriting for revenues) comes from selling in-game currency.
5: Some firms also offer account renting: use of a high-level character for a limited period; e.g. 7 days use of a level-70 World of Warcraft character for US$49.
6: “Escort services” have also been tried (Dibbell 2007). The virtual equivalent to guided ascents of Everest, these involve a group of gold farmers accompanying the customer on difficult or dangerous activities, such as dungeon-running in World of Warcraft. Non-gold-farming items that a number of gold-farming firms offer include sale of “keys” (codes used to activate a game account) and game cards (pre-paid amounts allowing playing time of say, 30 days, on an account).
7: World of Warcraft is the dominant game; Blizzard had first quarter revenues for 2008 of US$300m on a subscriber base of 10m, making US$10 per subscriber per month (Androvich 2008). That average also fits with the actual US$15 monthly fee in Western markets but less than that in Asian markets. It also fits with fees for two of the other most-played online games: Runescape (c.US$7 per month) and Final Fantasy XI (US$13 per month). US$2.5bn per year also makes sense given World of Warcraft has half the market and makes roughly half that figure in annual revenue. However, this figure could be an underestimate since Barboza (2005) claimed US$3.6bn per year in online game subscriptions in 2005.
8: Yee’s survey work found 22% of players had taken part in real-money trading and spent an average US$135. Castronova estimates this was, on average, spent over a three-year period, giving a spend of US$45 per year per RMT-active player, and hence a spend of roughly US$10 per player averaging over all players. Statistics from Station Exchange – the RMT system introduced by Sony Online Entertainment on two of its Everquest II servers – are, supporting Yee, that exactly 22% of players registered to take part in trading (Robischon 2007). However, the spend was much higher than in Yee’s survey – averaging around US$240 per RMT-active player per year, and around US$46 per year averaged over all players.
9: There are two potential downsides to the Station Exchange data. The legitimation and trustworthiness of Station Exchange RMT would probably increase player’s likelihood to spend compared to the normal “grey-to-black market” status of most gold-farming trades. If so, the SE figures overestimate the size of the overall market. (Put another way, comparing the US$46 and US$10 per-player annual spend data from our two sources might be taken as evidence that gold farming’s crepuscular status was artificially suppressing consumer expenditure.) Second, the total Station Exchange trade figure could include arbitrage/re-selling, e.g. a player buying currency for a lower price and selling for a higher. The SE trade figure would therefore exaggerate the extent of final-consumer purchasing; again causing the per-player figure and any extrapolations therefrom to be an overestimate. Robischon (2007) claims this is not so by noting the top three buyers were not in the list of top 20 sellers. This falls short of being a comprehensive refutation of the arbitrage possibility.
10: One reason for not jumping feet-first in with the >US$1bn estimates is that almost all the data foundations here come from 2005/2006. As discussed later in the paper, since that time currency exchange rates have fallen severely and it seems likely that profits have done likewise, thus greatly reducing market revenue. There also appears to have been more action taken against gold farming. The only counter-trend is the very high annual growth rates in numbers of online gamers. These trends might or might not roughly cancel each other out.
11: There are many secondary reports of these locations but primary reports of gold-farming exports from other than China include: Mexico (Dibbell 2003), Romania (Lee 2005, Thompson 2005); Russia (Weir 2004, Rickey 2005); and Indonesia (Lee 2005, Paul 2005). All except Thompson appear to be based on interview by phone or email rather than field visit, and one may also note the lack of post-2005 sources. Primary reports of gold-farming for the local market include: India (Sengupta 2007); Malaysia (Singh & Choo 2006); South Korea (Huhh 2008, Jin & Chee 2008); and the Philippines (Magno 2006).
12: By 2008, this had risen to 163m broadband users, 78% of China’s 210m Internet users (CNNIC 2008).
13: An element of this that re-appears enough to be noted is the claim that Indians don’t play computer games because they do not have enough time (Kandaswamy 2007, GameGuru 2008).
14: A figure justified only on the balance of reportage in academic, informed comment and mainstream media sources. This is a very weak basis for statistics but, in the absence of field data, the least worst option. Other than reported locations, a couple of other snippets suggest China’s domination. Jagex found tens of thousands of automated money-making accounts (bots) on Runescape were registered from China (Ward 2008). Real-to-virtual currency rates on a wide range of MMORPGs (Final Fantasy XI, Guild Wars, EVE Online, Runescape, World of Warcraft) all spike in late January/early February (Davis 2008b). This is claimed to be because of the Chinese New Year holiday, which reduces supply (for some corroboration see, e.g., WoWMine 2008).
15: Primary-reported locations – city/town and [province] – for Chinese gold farming include: Chongqing [Sichuan] (Barboza 2005, Chan 2006); Fuzhou [Fujian] (Barboza 2005); Hangzhou [Zhejiang] (Jin 2006e); Harbin [Heilongjiang] (Wang 2008); Jaixin [Hebei] (Jin 2006e); Jinhua [Zhejiang] (Jin 2006c, Dibbell 2007); Liaozhong [Liaoning] (Johnson 2006); Lishui [Zhejiang] (Jin 2006f); Nanjing [Jiangsu] (Dibbell 2007); Shanghai (Honge 2006, Lim 2007); Shenyang [Liaoning] (Wang 2008); Shenzhen [Guangdong] (He 2005); Suzhou [Jiangsu] (Jin 2006e); and Zhousan [Zhejiang] (Jin 2006e).
16: One per 10,000 is a low estimate from the figures given – two or three per 10,000 could be more likely. Figures in this ballpark are supported by the reports of the one-time main gold broker, IGE, which stated it had “tens of thousands” of suppliers in China (Carless 2006).
17: Fujian, Guangdong, Hebei, Heilongjiang, Jiangsu, Liaoning, Shanghai, Sichuan, and Zhejiang with a total population of just under 500m.
18: As a background note, Yee (2005) reports that Western MMOG players in his survey had an 85:15 male:female split.
19: A rare item of data from a gold-farming firm on the games from which customers order shows – not surprisingly – that orders mainly come from games with high subscription levels: World of Warcraft, Runescape, and Final Fantasy XI (Carless 2007).
20: Both estimates ride heavily on Yee’s (2005) survey work and, in the latter case, Castronova’s (2006) estimate that the US$135 average spend reported in Yee’s survey should be spread over three years of game play. This data was US-specific and is about four years old at the time of writing thus, for example, pre-dating the main anti-gold farming activities including delisting from eBay. That suggests the figure could be an overestimate. On the other hand, it excluded expenditure on power-levelling (Yee 2008), and was some way less than that reported in the Station Exchange’s first-year experience (Robischon 2007); suggesting it may be an underestimate. Finally, average spend may be asymmetrically distributed, with a larger proportion of sub-average purchases.
21: There were no differences across gender but some mild correlation with increasing age (quite possibly because the older you get the more money and less time you have); something borne out by the Station Exchange data showing the Everquest II average buyer age was 34.2 years compared to an all-player average age of 25 years (Robischon 2007). For some insights into the real-life experiences of a gold purchaser, see: http://www.wowgoldfacts.com/. Steinkuehler (2006) cites a smaller (n=793 as opposed to Yee’s n=1923), earlier (2004) and single-game survey which suggested 31% of Lineage II players bought in-game currency via eBay. Lim (2007) reports 20% of players buying currency/services but this is likely a rounded repeat of Yee’s figures.
Back to Top

Related posts:

1. Gold Farming Research – Part One
2. Gold Farming Research – Part Three
3. Gold Farming Research Part Four
4. Gold Farming Research Part Five
5. Gold Farming Research Part Six – the last one