Tuesday, January 03, 2006

The Social Scientist’s Dream—As Close as Your Keyboard

Critics of social science have long pointed out–and social scientists regretted–the difficulty of doing proper controlled experiments in their field. An economist can offer arguments to show that price control has bad consequences. He can provide extensive anecdotal evidence to support those arguments. But he cannot take a hundred identical societies, impose price control of fifty of them, and see how average measures of individual welfare compare ten years later.

That problem has now been solved—not, perhaps, for price control, but for a considerable range of propositions in the social sciences. The solution is provided by a relatively new phenomenon in high tech entertainment—Massively Multiplayer Online Roleplaying Games.

Consider Blizzard’s World of Warcraft, the MMORG I know best. It has several million customers on several hundred servers. Each server (I simplify slightly) is an identical virtual world, with a population of perhaps ten thousand players, and, since each player may run multiple characters, several tens of thousands of virtual inhabitants. Behind each of those inhabitants—not counting the non-player characters provided by the game—is a human being.

On each of those worlds, characters are making (virtual) stuff, using stuff, buying and selling. Trade occurs at fixed prices between players and non-player characters, at market prices between players, either directly or though an auction house—on which prices and transactions can be freely observed.

The economic experiments you can arrange are by market participants, not legislators or regulators, unless you have an inside line to Blizzard. Working at the market level with quite a modest research budget, you could easily arrange to be a substantial fraction of the supply or demand for a single item out of the many being bought and sold. Ten graduate students running ten characters each on a single server would be sufficient to experiment on elasticity of supply and demand, workability or otherwise of strategies for manipulating the market, and the like. To test various forms of the efficient market hypothesis, try to develop simple rules for buying and selling on the auction house that yield a positive return. The same graduate students can be running other characters on other servers, doing statistical work involving multiple identical worlds.

I am an economist, so my brief examples are economic ones, but there should be opportunities in other fields as well. One respect in which the worlds represented by different servers are not quite identical is in their populations. Servers whose internal clocks are on Pacific Standard Time are populated mainly by people from the west cost of the U.S.—with an occasional Spaniard or Korean. A server on Korean time—I am told the game is very popular in Korea—will have a rather different population. That should make it possible to do extensive studies of differences and similarities in social norms across a wide range of societies—without spending a penny on airline tickets or hotels.

So far as the cost of the game is concerned, a little over a thousand dollars a year—a small fraction of any serious research budget—will buy you a hundred characters each on every server. Most of the cost of such a project would be the time of the researchers—and grad students are not very expensive. If you select them properly you may get a good deal of their time for free, since from their standpoint you are paying the cost of their recreation.

A whole new world of research.

[Any curious WoW players who read this blog can find me and my family on the Feathermoon server.]

13 comments:

Anonymous said...

http://www.npr.org/templates/story/story.php?storyId=4946772

The above is a link to an NPR article - I assume you've seen it - where WOW was used (serendipitously, as it were) as a model for how a population might react in the event of a pandemic.

It seems you are not alone in seeing the potential MMORPGs have for certain kinds of social experimentation.

Anonymous said...

I've been involved in MMORPGs for many years and you're exactly right. I've seen first-hand, for example, the rapid inflationary effects of a quickly increasing money supply in a MMORPG. That's a trivial example, of course, but it was very interesting nonetheless.

Anonymous said...

I wonder how serious a distorting effect the presence of gold farmers (individuals who dedicate their entire gameplay to the generation of in-game money, so that they can trade this in-game money for real-world money) would have on such research.

They seem to simultaneously raise the price of big-ticket items, which they attempt to monopolize, and depress the price of more common commodities, because their dedicated pursuit of valuables substantially increases the supply of these goods over what would be available from those who don't play for financial gain.

Essentially these people break the isolation of the economy, bringing factors of an outside economy to bear on the economy you wish to use for testing. The world of Azeroth has international trade. US dollars will buy you weapons and armor. You might even be able to speculate in Azeroth's currency market.

Anonymous said...

Hey, David, long time since the HPO days...

Anyway, forgive me if this is a silly question--I'm unfamiliar with these sorts of games--but isn't this rather like playing penny-ante poker, or playing it with funny money?

How can you get socially relevant data when people are not facing real-world cosequences for their actions and risk-taking?

Peripatético said...

RPGs are good labs indeed, but , that’s my point, are the findings applicable to organic world? When you are playing you’ve a different perception of the risks you are facing. There are, of course, other possible gains or losses. In addition, the gamers profile (age, education, sex, social and economics background) may be not representative and it exists other moral and cultural issues that may be irrelevant in a dragon’s lare but they are outside.

Great blog, keep reading.

David Friedman said...

Richard asks:

"but isn't this rather like playing penny-ante poker, or playing it with funny money?"

My wife, reading over my shoulder, answers "No--because they care about it."

In the real world, a thousand dollars has a certain value to someone, which affects how hard he will try to earn it. In a virtual world, a thousand pieces of gold have a value to someone, which we can observe in the time and trouble people take to earn that.

Peripatetico comments that the population distribution isn't the same as in the real world. But economics isn't about "the economy of the U.S. in 2006." Any interesting economic proposition ought to apply to a wide range of societies.

Anonymous said...

You wrote:
"In the real world, a thousand dollars has a certain value to someone, which affects how hard he will try to earn it. In a virtual world, a thousand pieces of gold have a value to someone, which we can observe in the time and trouble people take to earn that. "

But of course the value we assign to monetary units is also reflected in how we spend it, and the conseqenses and benefits of having more or less of it. While we can get some idea of the value of gold by looking at how much trouble people are willing to undergo to earn it, it's also important to note that in the game economy people are going to be fairly free about spending it.

Additionally the very nature of the world is that it is designed to be enjoyable. Therefore the time spent earning money is intended to be a rewarding experience apart from the monetary reward itself.

In the game world, there are very few consequenses for being poor, and once you get to a certain level of money, diminishing rewards for being rich. This devalues gold a certain amount.

In the real world, a person has severe consequences for having no money, and there are generally fewer diminishing returns for being filthy stinking rich. :)

The WoW economy is an interesting beast, and I'm interested in seeing what observations you'll be making about it in the future.

Personman said...

I just got an awesome book on this subject for Christmas - it's called Synthetic Worlds, and it's by Castronova. He has a website and blog devoted to the economic study of MMORPGs over at http://terranova.blogs.com/

Good post, and definitely check out his site if you haven't before.

Anonymous said...

If you're looking for interested social and economic models in MMORPGs, I suggest Puzzle Pirates at http://www.puzzlepirates.com, which has very dedicated players and a fairly complex economy, which I could explain more if you'd like.

At least one person has played some interesting experiments on the players. The experiments are listed below. In one event, the player held essentially a sealed bid auction for a purely luxory item (a familiar), which is known to sell for 2 Million Poes (the currency of the game). Minimum bid was 10000 Poe, which can be made in 3 hours by a lot of the players. The auctioner would keep all the bids, and all bids and bidders would remain secret. Despite predictions on the forums of a huge turnout, nobody entered.
In a similar event which ran at the same time on a separate server, there was no minimum bid but people could bid more then once, the familiar was given to the lowest unique bid. Again, predictions of a huge turnout, but the winning bid was 4, and that bidder was the only one to bid above 3. Both interesting statements on that cultures willingness to risk and reaction to predictions.

~Pete, who also went to Mudd
Rumsfeld on the Cobalt server

The experiments:
http://forums.puzzlepirates.com/community/mvnforum/viewthread?p=468090#468090
http://forums.puzzlepirates.com/community/mvnforum/viewthread?p=478636#478636

mark said...

Richard Posner wrote on this topic as well when he filled in for Lessig

http://www.lessig.org/blog/archives/002124.shtml

Jack Balkin paper on virtual worlds

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=555683

Great blog BTW

Anonymous said...

I'm not an economist, but is empirical testing really a sensible way to test the truth of economic theories such as the effect of price controls? Aren't they better tested by evaluating (criticising) them at a theoretical level, than by trying to run lab experiments? After all, what if the result of a MMORPG experiment contradicted the theory? Where would that get us? Would it really lead you to doubt the truth of the theory?

David Friedman said...

"After all, what if the result of a MMORPG experiment contradicted the theory? Where would that get us? Would it really lead you to doubt the truth of the theory?"

It might give you hints on how to improve the theory.

Suppose you start with a simple version of the efficient market hypothesis. You then discover that there are buy/sell rules that let you make money on the WoW auction house. That pushes you towards a more sophisticated version of the theory, perhaps one that takes better account of transaction and information costs.

If no such theory works, you might have a clue to patterns of behavior that don't fit the rational model but are still predictable.

I have an article, published and a draft on my web site, on economics and evolutionary psychology. Part of the motivation for it was observing (not in MMORGs) patterns of behavior that don't seem to fit the economic model, and trying to make sense of them.

Anonymous said...

Here are some articles about EVE Online and its economics. The first is about a player organization that went public with an IPO. The second is about someone scaming others with a fake business deal.

Trust Me, the IPO http://www.escapistmagazine.com/issue/25/3

The Great Scam, http://static.circa1984.com/the-big-scam.html