Learning from virtual economies: The (im)possibility of an autistic approach

Ever dreamed about being "Helicopter Ben" or even a benevolent social planner? The advent of massively multiplayer online games (MMOGs) could not only make your megalomania come true but open a new field for empirical research.

Despite the ever growing sophistication of mathematical modelling and quantitative analyses, economics still remains a wannabe natural science. The main problem, perhaps, is the difficulty of conducting controlled experiments. While controlled lab experiments in microeconomics (e.g. game theory) help gauging the explanatory power of models, they still remain highly abstract and thus raise doubts on the external validity of the results. On a macroeconomic scale, it is hardly possible to even conduct controlled experiments. Although the reliance on natural (quasi)experiments and use of instrumental variables has helped alleviating this disability, the use statistical tools remains rather vague and a Hessian "glass bead game".

The emergence of MMOGs and the virtual world might provide another battleground for econometricians, ultimately offer new insights in economic theory. As an example, consider the case of "Mudflation" in the popular game World of Warcraft (as I am no gamer, I apologize for any inaccuracy in my following narrative).

Mudflation: A WoW narrative

Like in a real economy, money (gold) is used in WoW as an intertemporal storage of purchasing power and a means to conduct transactions. The player obtains money for killing monsters and completing missions (quests). The money obtained can be used to purchase items such as weapons and potions, among other goods and services (e.g. repairing amory). While this simple consumer incentive drives the game play (better weapons, better equipment, better missions), it unintentionally gave rise to the emergence of an industry of "gold miners"; these are agents who perform nothing but killing monsters and accumulating money (hence "grinding"), ultimately selling the virtual money over eBay for real world profit. As more and more money was injected through this flawed mechanism, price level started to rise, leading to inflation in the virtual economy and harming the low-skilled, low income players.

Why am I telling this story? In response to the price surge, Blizzard devised more and more clever mechanisms that withdrew money from circulation (e.g. transaction costs, taxes), basically implementing an institution similar to a Central Bank which kept money growth in check. While mudflation in WoW has now stabilized, the phenomenon itself was no exception: Similar events occured in other games such as Second Life, Everquest or Final Fantasy. Obviously, virtual agents behave no differently than in real world. If so, it would enable scientists to conduct controlled experiments on a large scale.

Omniscience and omnipotence

The game designer is not only omniscient, but also omnipotent. Omniscience eradicates measuring errors and allows to control for all existing variables. Obtaining the data from these virtual economies is easy and cheap, therefore opening opportunities for data mining and EDA. Omnipotence allows the game designer to set up experiments where economic theories could be tested. These experiments can even be fun, as game designers can simply invent a suitable story (yes, the helicopter is possible!) and of higher validity than natural or randomized field experiments. Finally, as the number of gamers reach hundred of thousands, inference is not restricted to a small sample (like in game theory labs).

This potential has been recognized, and more and more research makes use of this sheer endless stream of data, with a main hub being the Virtual Economy Research Network (VERN). More and more questions are raised and answered: How is real money spent on virtual goods (SecondLife even has an estate market)? Does pressure play into SecondLife purchasing? Does the virtual economy in Everquest follow economic laws?

Concluding remarks

While there are certainly doubts on the utility of such studies for real world issues, I believe that studying virtual economies can ultimately deliver relevant insights for reality. And since economic models don't seem to fit reality, why not retreat into virtual reality where each assumption and shock can be programmed and its effects fully captured?

Even though this approach might seem very intriguing, it is awfully difficult to operationalize such experiments. A next step would be to devise clever experiments (embedded in the game play) which enable tests of economic theorems (I'm thinking for example about a cross-server Balassa-Samuelson Effect? Or the role of institutions, a still very hot topic?). Anyway, if you know more about it or have some new ideas on operationalization, please do comment on this little draft.

Interesting fact: The Gini coefficient in WoW exceeds those of the worst real world countries, with a figure as high as 0.9 (as of 2008). If you have a newer figure, please do let me know.

GX.

Sources:

[1] http://virtual-economy.org

[2] http://www.metanomics.net/

Data on game prices:

[1] http://www.gamepricewatcher.com/

[2] http://gamerates.com

[3] http://www.mmogchart.com/

[4] http://www.wowecon.com