Internal data from Second Life give us outsiders our first glimpse at the volume of trade inside virtual worlds. It's big. Really big.
Over a one-month period, the users of 2L had a stock of 8.35 million Lindens (L$) and used it to conduct L$19.2 million in trade. Economists call the ratio of trade to the money stock "velocity;" in this case, it's 2.30.
Economists also tend to assume that velocity is constant. It's the turnover rate of a single unit of currency and is determined by the efficiency of market institutions. Second Life doesn't seem enormously different than any other world in this regard, so let's assume the velocity is about the same in other worlds, such as Norrath.
Data from Norrath in Summer 2001 indicated that the typical user had about 7,700 platinum pieces (PP) in cash. At that time, EverQuest had 400,000 users and the exchange rate between PP and dollars was 100:1. This means that the total money stock was (7700)*(400000) = PP3.08 billion. Using a velocity of 2.3, total in-world trade would have been PP7.084 billion. At the then-current exchange rate, that comes to $71 million.
So, a game like EQ has $71 million monthly in in-world trade. Current data from eBay suggest that the combined export-import trade for all US games except EQ - games whose total populations nonetheless well exceed EQ's 400,000 - is only about $1.3 million monthly. Evidently, in-world trade is much greater than out-world trade.
The figures for EQ suggest that each user does about $18 of in-world player-to-player trade monthly. That number feels about right (seriously! I've actually played EQ quite a bit). If we think about the market as a whole, we have upwards of 10 million people in the various worlds right now. At $18 per person per month, that's an aggregate in-world trade volume of $2.2 billion annually. According to UN data, that volume of trade would make virtual worlds the 144th largest economy in terms of GNP, right up there with Malta and Albania.
On the other hand, this number is such an extreme telescoping of assumptions as to have little merit in an absolute quantitative sense. We only have the velocity number - the keystone of all this - because Philip Rosedale of Second Life was kind enough to let me see it. It's frustrating that we don't have better data, and that we're forced to put together this crazy-quilt patchwork of assumptions to get even the most basic feel for what's going on. (If you're the proprietor of a big game and can see how some of these data, if released, could benefit the entire industry, get in touch.)
However, these data do yield convincing evidence of a qualitative, if not quantitative, nature: Whatever we see on eBay is dwarfed by what's happening inside the games. A great deal of value is changing hands out there. And it's all denominated in gold pieces.