TO TRUCK, BARTER AND EXCHANGE? On the nature of our digital economies

Following my previous post on the calculation of arbitrage opportunities and relative prices in the TF2 economy, I received many messages making more or less the same, terribly apt, point: Countless exchanges on Steam, Valve’s trading platform, did not meet the criteria of a market exchange. What does this mean? And why is it important for our research into the size and nature of the Steam economy? Finally, what does this debate have to do with Adam Smith’s relevance to digital economies?

One reader, for example, wrote to say that more often than not she used Steam in order to ‘give away’ to fellow players items that she had in her backpack (and which were surplus to requirements). Or to indulge in gift exchanges with them. The point of those communications was that, if people are prone to effect ‘trades’ that are, in reality, ‘give aways’ or some form of ‘gift exchanges’, the computation of relative prices utilising the observed exchange ratios from such ‘trades’ will be badly compromised (as these exchange ratios are, despite their social significance, economically meaningless and, therefore, perfectly capable of skewing our computed relative prices).

My immediate answer is to acknowledge their point and to answer that we are doing our best to leave out of our computations exchange ratios that are clearly economically senseless. Of course, this also means that we, as analysts, exercise a great deal of arbitrary discretionary power (in deciding which ‘trades’ to ignore) while formulating our estimates of relative prices. Nevertheless, my main rejoinder here is that, while I acknowledge the point, …such is (economic) life! Even in ‘real’, or analogue, economies, many agreed prices between buyers and sellers reflect factors beyond supply, demand and bargaining power. Why would, or should, digital economies be any different?

These thoughts led me to a more general, almost philosophical, consideration. Adam Smith, the ‘patron saint’ of economists, believed that the foundation of all economic activity was none other than the human “…propensity to truck, barter, and exchange one thing for another.” (See Book 1, Chapter 2 of his Wealth of Nations, 1776). Was he right?

It depends on what he meant by ‘exchange’. If he meant the sort of exchanges that occur in a stock exchange or in some anonymous marketplace (including your local supermarket), then Smith was profoundly wrong: Multi-person (as opposed to Robinson Crusoe-like) economic activity began life, both among animals and humans, on the basis of fully socialised reciprocity. As I argue in Section 1 below, the history of human economies is built on cooperative non-market interactions – even money was not ‘invented’ as a mere lubricant of market exchanges (despite what most economists think). Exchanges are never of the purely market sort envisioned by economic theorists and free market enthusiasts (see Section 2). Thus, the idea that our economic history can be understood in terms of some natural instinct for pure market trades is, at least for me, false.

Setting aside, for the moment, the pertinence of Smith’s hypothesis about the foundations of our ‘analogue’ social economies, could it be that Smith’s theory is spot on in the case of digital economies? Surely, when people trade on Steam, or within other video game communities, they are propelled by an instinct to “truck, barter and exchange” motivated purely by self-interest. In this blog post I present my tentative conclusions on this question. But first, let us look carefully at both sides of the argument about the nature of economic activity and the meaning of exchange.

1. If not barter and exchanges, what was the essence of human economic activity (before, that is, multiplayer computer games came along)?

Smith presupposed that humans, from a very early stage, based their economic activity on trade. He imagined that pre-historic hunters, gatherers and shepherds would gather together, or meet in pairs, and indulge in exchanges of arrowheads for tanned leather or of rabbits for chunks of stag meat. That, at some point, tired of being at the mercy of the double coincidence of want, they homed in on some numéraire–asset (e.g. salt, cowries or some precious metal) which they began to treat as a universally accepted currency for the purposes of greasing the wheels of trade.

While this story is nicely satisfying, it is most probably factually wrong. Anthropologists of note have demonstrated that the evidence points to another evolutionary process that yielded money. One that was based not on money’s utility as a lubricant of trade but, rather, as a unit of accounting for debt! Take for instance the first archaeological evidence of accounting books, dating to 3500BC and unearthed in Mesopotamia (contemporary Iraq). They come in the form of tablets on which ancient accountants had painstakingly carved a log of who owed what to whom, of how much grain each resident within some temple jurisdiction had stored at the communal warehouse, of how much barley was owed to those working in the temple. What is beguiling is that the unit of account often took the form of silver coins that, in fact, did not even circulate (or had not even been minted). Indeed, everyday use of coins as a means of exchange was not witnessed for several thousands of years after it was used to record debt obligations…

If I am right, the abundant archaeological and anthropological sources at our disposal suggest that pure exchange was not civilisation’s foundation, despite Adam Smith’s conviction. No, what set us on a course to modern societies were social conventions of production and distribution based on hierarchies, obligations, social norms that determined individual and caste entitlements and, of course, organised violence. Even though we developed money very early on, it did not buy much.

2. Regardless of the origins of money, have pure exchanges taken over all economic activities?

Even in today’s highly monetised society (where almost everything is for sale), it is not hard to see remnants of the older societies, in which money did not buy much. Consider the following scene: It is Thanksgiving. Mum has slaved over a hot stove all day to produce the turkey for everyone and, at the end of the day, it is expected that each ‘recipient’ of her ‘gift’ will help with the washing up. One family member declares that he will not be participating in the clean up and asks: “How much should I pay you to get out of my washing up duties?” The answer, most probably, is: “Get lost!” I use this example to illustrate that it is perfectly possible to be utterly familiar with money, to use it all the time, but to refrain basing decisions on what to produce and for whom exclusively on monetary transactions. Indeed, my argument above has been that civilisation was built on a division of labour that had little to do with pure exchanges. It is only in the past two or three centuries that impersonal pure exchanges have taken over. And even now, it has not done so fully.

3. Pure and impure (or contested) exchanges

What is a pure exchange? It is an impersonal market exchange where the relationship between the buyer and the seller is fully reducible to the objects exchanged (e.g. apples and oranges) and to the exchange ratio (e.g. two apples for one orange). Moreover, it is a relationship that expires the moment the transaction is completed.[i]

Impure exchanges, in contrast, involve social obligations and thoughts such as “I shall do X for Jill not because of what I expect Jill to do in return for me but of what she would do if she were in my place.” In impure exchanges there is no expectation of equivalence of that which is being exchanged (e.g. your mother does not need to feel that her labour, to produce the turkey, is somehow equivalent to that you will be putting into the washing up later). In pure exchanges, by contrast, traders worry about little else than being ‘short-changed’. The market rules supreme when ‘equivalence’ in values traded is the name of the game.

Of course part and parcel of impure exchanges is hierarchy. At Thanksgiving, mum pulls rank and everyone else does as they are told. Hierarchy involves exchanges between unequals which are more often than not supported by social norms and (unequal) gift exchanges reflecting the ‘givers’ social status. In many societies, for instance, a ruler would often bestow a precious gift upon one as a means of deepening their recipient’s subservience.

For centuries, if not millennia, the pure and the impure coincided in the exchanges that kept the world going. Even though money was often involved, fully monetised, pure, exchanges did not ‘arrive’ until fairly recently. When exactly? Around three centuries ago and, in particular, after land and labour were commodified (following, e.g., the conversion of bonded peasant labour into free labourers working for a wage).

Today, we have come as close as possible to creating a realm of pure exchanges. When you walk out of the supermarket checkout, your social relationship with the ‘seller’ (who is not, of course, the checkout employee) is non-existent. It is as close to a pure exchange as it can be. The financial markets are the ultimate realm of pure exchanges, especially when trading is done not even by humans but by algorithms (or algos) trading with one another on our behalf. And yet this is not to say that we have a social economy predicated entirely on pure exchange.

Take for instance the employment contract through which employees ‘sell’ their labour to employers. By definition, this is an impure transaction. Think about it: the employee cannot simply ‘leave’ after selling her ‘asset’ (i.e. her labour) to the employer. She must stay on the premises while her labour is ‘diffused’ through the company’s network. So, as long as people are hired by employers, our analogue economy can never be a realm of pure exchanges. Something other than “trucking, bartering and exchanging” must be going on.

4. Steam trading: A realm of pure exchanges?

Digital economies, like Steam’s exchange platform, come closer to Adam Smith’s concept of an economy that sprang from a penchant for pure exchanges. People meet up online, enter into mutually beneficial trades with minimal other social obligations to one another, bear no debts (monetary or social), and walk off with whatever item they managed to acquire (via bartering) without any need to maintain a ‘relation’ with the person they bartered with. An economy created as if in the image of Adam Smith?

Not quite. The exchanges that we observe on Steam are not exactly pure. As readers have perceptively remarked, many exchanges are highly impure. People simply use Steam in order to unload onto others items from their backpack that are surplus to requirements. Often, they will accept remarkably low ‘value’ in return. Value equivalence, a prerequisite for trading-proper, is simply absent. But then again, this is how trade began and continued to be practised for thousands of years: People sold goods that they produced, or gathered, over and above their ‘planned’ (i.e. needed) volumes. It is not that they produced these goods in order to exchange them (the very definition of a commodity being a good produced in order to be traded) but, rather, they sold the quantities that ended up being surplus to requirements. It is no exaggeration to suggest that, until the rise of industrialisation and market societies, less than 5% of goods produced were commodities. So, until a couple of centuries ago, most trade happened because of unexpected, or unplanned, surpluses: very much like those in the TF2 economy!

In short, Steam trades are not always pure exchanges happening in some moral-free zone where social obligations are perceived to be non-existent. An unspecified (and impossible to compute accurately) number of trades take place at exchange rates that do not reflect the relative bargaining of buyer and seller but, instead, are determined by other social and gaming factors. In technical terms, this means that, while our arbitrage data is not affected (since the volume of arbitrage opportunities is independent of the reasons for which some items are sold cheaply and resold expensively), our relative price estimates are. Ideally, we would like to have some ‘gift exchange’ radar that alerts us to all instances of Steam ‘trading’ where people are far from trying to get the best possible ‘bargain’ for themselves. If we possessed such a radar, we would use it to decide which trades to turn a blind eye to when computing relative prices. Of course, that ‘radar’ is missing. So far we are utilising crude methods of ‘visual’ inspection, leaving out of our calculations those relative prices that seem, economically, silly. Clearly, we need to work on coming up with such a radar. Any suggestions from you will be most welcome…

An interesting twist to this debate is added by the opportunity that TF2 players have to buy gifts for anonymous strangers by purchasing a Secret Saxton from Steam. (See here for information on gifting, for those unfamiliar with it. In brief, purchasing a Secret Saxton immediately sends an item to some random player within the digital vicinity of the ‘giver’, followed by a public statement of who was the ‘giver’). The popularity of these Secret Saxtons provides ample evidence that TF2 players are replete with motives that go well beyond the urge to “truck, barter and exchange”.

As the following diagram reveals, random gifts to fellow ‘strangers’ are frequent and peak heavily around a two-day period during which we have ‘learnt to believe’ that we ought to give presents to people with whom we have some form of social relationship with. Hardly evidence that those trading on Steam are only indulging in ‘pure’ exchanges, i.e. that they are guided by nothing more than the pursuit of value equivalence.

Recapping, gifts are an important part of the TF2 economy and this, on its own, casts doubt on the purity of all other exchanges and, by association, on our relative price estimates. Our technical difficulty regarding the latter is that, unlike the case of gifting via purchases of the Secret Saxton item (for which we have perfect data), gift exchanges on Steam (as confirmed by readers’ messages) must be prevalent and truly capable of spoiling our data on relative prices.

You may recall that in my previous post, I made a big deal out of our finding that no item emerges as a common currency, a numéraire. Our close study of the TF2 economy revealed that there are times when different items are traded most frequently in different periods. If one item had ‘evolved’ into currency status, that item would have been the most traded all the time. Confounding my expectations that keys would play that role in TF2, Steam trading data shows that there are five or six items which alternate as currencies. How come? Why does one item within the TF2 economy not evolve into a currency given that everyone’s (trading) experience would improve (in terms of ease of ‘closing’ a mutually beneficial trade)? One possible explanation that the preceding discussion is pointing to is this: Because Steam trading, at least of TF2 items, are instances of impure exchanges; that is, trades that are not fully guided by the principle of trading items of equivalent subjective values. Norms of the social valuation of one item relative to another, beliefs of what one is ‘entitled’ to expect to receive for some valuable item (as opposed to what one actually would be happy to receive in exchange for that hat), expectations of a continued social relationship between buyer and seller – all these ‘impurities’ would, if indeed present, prevent the evolution of one item into the role of common currency.

Last, but not least, the impossibility of using a potential item in order to account for debts on Steam trading (unlike in Mesopotamia), offers another clue as to why no item dominates as the game’s currency unit.

5. Conclusion

Many economists believe that philosophising over the nature of exchanges is a luxury they do not need in order to analyse and understand an economy. They are wrong. The nature of exchanges, whether they are pure (i.e. asocial) or impure (replete with social norms and part of intertemporal social relations), makes a difference when it comes to predicting economic activity. Thus, to understand the exchanges we observe on Steam, it is crucial that we grasp the network of social relations within which they are embedded. The prevalence of gifting and the fact that no specific item has emerged as a form of money in trades of TF2 items should alert us to the intriguing social conventions that are part and parcel of our community’s trading decisions. How will these conventions change or mutate when participants are given the capacity to buy and sell, among one another, using real dollars? Would it make a difference if any dollar profit made through such trades can be taken out of Steam (i.e. monetised)? I suspect the answer to these questions are in the affirmative. But we must wait and see.

[i] Another definition of a pure exchange is that of an exchange where the parties involved could, potentially, draft a complete contract in which all relevant aspects of the exchange may be included in a manner that leaves no room for legitimate contest of its terms ex post.

63 Responses to TO TRUCK, BARTER AND EXCHANGE? On the nature of our digital economies

  1. Brandon says:

    Thank you for another absolutely fascinating post. I feel like I am smarter for having read that.

    One thing interests me. Watching real life markets, we can have little idea what the relationship is between parties. However, online, we have several markers that might serve. Valve is able to know whether the parties on either side of a trade are on each other’s friend lists (either directly or are two or so hops from each other on a network graph), share communities, or even regularly play together on the same game servers. It would be interesting to compare the activity of likely “strangers” with the activity of likely “friends” to see whether one behaves significantly more like a pure exchange.

    • Dhruv says:

      Friends are often added purely to commence a trade hence in many cases the two parties classified as ‘friends’ undertaking a trade are actually not acquainted.

      • David C. says:

        I would imagine though that could be mitigated somewhat through other data patterns. Most simply, I’d be suspicious of trades occurring between friends whose “friendship” has lasted less than five minutes before the transaction started.

        • Matt says:

          Another feature you could look at would be network connectivity among friends, that is, generally friends of friends are also often friends. If I’m trading with someone who is at best 7 edges away in a friend-network, that person is not likely someone I am acquainted with.

        • Darren says:

          Yes, you could also look at the number of friends-in-common.

      • Amit says:

        Also, I believe Valve might be able also track friends amount of times two ‘friends’ chat (But this id doable only if it doesn’t conflict with Valve’s privacy policy) I personally chat only with my real friends, and barely with strangers…
        So I do believe Brandon is right, and it might help build that radar Valve is looking for. A trade done by two users who play often on the same servers, and chat a lot shouldn’t be considered in the math.

        However it might get complicated as you can trade things from one game to the other and those two people might not like the other game, so they have nothing to do with it…
        Anyway, best of luck making that radar.

        • John L. says:

          Even that is not a perfect criteria. I for example only talk with my Steam friends on IRC, TeamSpeak, AOM and external communication services. They are still my gaming friends, I just like those services better.

          In response to David C. above, that also wouldn’t work. I often bring RL friends to Steam by gifting them a game through it. After “initiated” we would play together constantly online, so that criteria also wouldn’t work.

          The problem you face when dealing with people in such level is that every person has an unique personality and unique traits that makes them interact with the same object(Steam) in very different ways. One way to deal with that problem would be to generalize the algorithm, but that would leave a big margin for error on any calculation. The other way to solve this problem would be to go deep inside the psychology of the player and get the universal instincts that drive them to such decisions, but that will require a great deal of work and might as well be impossible.

          Or we may go around the issue and find a way to get relevant data without the need to discriminate pure and impure exchanges. That might work too.

  2. Zeus says:

    Interesting post once again professor.

    Regarding the “gift exchange radar” issue….do you think there might be a correlation between the time period that separates the acquisition of a certain item and the subsequent trade (or giveaway) that might shed some light upon the nature of the “transaction”.

    Perhaps gamers that actively pursue arbitrage and pure trading are more inclined to proceed swiftly with a pure sale or purchase. On the other hand the impure exchanges might signify a longer time period between acquiring and item and giving it away since one (I would think) is not inclined to give away a newly acquired item.

    Just a thought. Not too silly I hope.



    • Matt H says:

      My experience has been quite the opposite, at least among my group of friends. When I gift a game to a friend, I tend to have bought the game purely for the purpose of gifting it. Most items traded rather than gifted tend to have been held onto for a while before a good deal can be found. This isn’t always the case though, (I held onto a copy of portal for over a year before i could find a friend who didn’t own it), and I’m not sure how much my friends and I fit into the standard.

  3. Garrett Greenwood says:

    You may want to consider features outside of each trade when deciding whether it took place in a pure or impure context.
    Using machine learning(more specifically, logistic regression), you could train an algorithm to estimate the probability that a trade is pure/impure based on data outside of the trade. For example, the following features may be indicative of the purity of a trade:
    * Whether or not the two parties are friends on Steam
    * Whether or not the trade took place while the participants were in a game
    * Whether or not the trade took place in a server marked as a trading server
    * Whether or not any of the items involved were gift wrapped
    * The time spent by each party playing in that particular gaming session

    Additionally, you could potentially analyze the contents of the in-trade chat for more features to determine the purity of the trade. Because the chat potentially captures bartering, it could be useful as well.
    * The amount of numbers typed in the chat
    * The size of each message typed
    * The number of messages typed
    * The degree to which the chat was back and forth (a chat that looked like ABABAB would be valued high, while AAABBB would be low)
    * The amount of time spent in the trade window

    In order to train an algorithm, you would need to gather data of each of the features for a number of trades, then manually sort through them, deciding which is pure or impure. After using logistic regression to find a set of parameters that can be applied to the features, you can apply them to new data and automatically and efficiently classify a large number of trades.

    I wish you the best in luck in your endeavors. These write-ups not only provide insight into the economy of TF2, but are also an easily accessible resource for those like myself who know little about the inner-workings of economics. I’ve learned much from these posts.
    Feel free to contact me if you have any inquiries regarding the application of machine learning to this unique problem. Although I’m not an expert, I do enjoy the subject.

    • AZ says:

      I like your way of thinking. I wonder if economists are familiar with artificial intelligence and higher mathematics.

  4. Yaakov says:

    Much of this is well over my head, but do you account for multiple users ‘idling’, ‘crate storage’ accounts and so on, or are you basing data off the assumption (which I don’t see refuted anywhere) that one account == one user? How much of a gift is it really if the recipient is the sender?

  5. StephenMe says:

    It’s been my experience that there IS a currency in TF2, among those who engage in trading frequently. Metal serves as the currency for “common” items — most trades, even those not involving metal, are referenced against an item’s value in “refined” (where a Scrap Metal is valued at 0.111 Refined Metal, as per the crafting system that allows us to convert Scrap into Reclaimed Metal and Refined, and back again). Metal serves this purpose perfectly because it can be created out of any two common weapons that are unwanted, and is necessary in almost all crafting recipes to create other items.

    The flaw with metal is that it can only conveniently be used for lower-valued items. The trading window is easier to use with smaller numbers of items (and it used to only be able to handle 8 items). It would be unwieldy to hold upwards of dozens of Refined Metal items in a limited-space backpack — some purchases would even require hundreds of Refined, even though the vast majority of items are valued at less than 4-5 Refined.

    So keys serve as a second currency for “high value” items. The exchange rate shifts, and is often a matter of opinion, but it usually sits fairly steadily between 2.33 and 2.55 refined per key.

    And even keys are not enough for the most expensive items, like highly prized Unusual Hats — again, because of the limits of inventory space and trading window space. These usually are purchased using multiples of rare-but-accessible items with a finite supply — like Earbuds, and Bill’s Hats.

    The thing is, there is no such thing as a “twenty refined note”. To hold ten times as much of a TF2 currency, you need ten times as much space, and space is limited. Unless Valve were to introduce varieties of metal beyond Refined, or were to allow players to “stack” metal or keys in a single inventory space, a single currency is incapable of handling the full range of values of items.

    • AJP says:

      Now this is fascinating – I’ll bet that the valve folks have not considered this possibility. There are MANY instances of this kind of phenomenon that arise in nature that often get overlooked by researchers (often for long periods of time) because they are generally looking at these phenomena from the “outside,” so to speak. We often have to interpret the context in which the data exists – a very, very difficult thing to do many times.

      I hope Yanis see’s this post and incorporates this information. I’ll bet, if this is as prevalent as you say it is, a clear tired currency system will arise. Tier currencies (i.e., one in which there are multiple forms of the currency that each carry different values) arose to solve the exact problem you’re describing – it was much easier to, say, carry 100 gold coins than 10000 copper coins. You describe the same issue!

    • Starfire says:

      This is exactly what’s been seen in the MMOs Guild Wars and Runescape; something similar happened in Diablo 2 as well. GW allows stacking of crafting materials. One of (if not the) rarest crafting materials is ectoplasm. It has an intrinsic value in making some of the best armors in the game. Money has a ludicrously low upper limit. (100,000 carried 1000000 in your vault) So ectoplasm, which stacks and only consumes one inventory space, as the most valuable but still obtainable item in the game became currency. It’s still subject to some market forces because it is both constantly generated and consumed. (Since I stopped playing it it looks like two or three other items attained currency status but there are probably less than five all told.) GW is a relatively low item game.

      Runescape has Party Hats, an item which is completely worthless from a gameplay perspective, which are available in limited supply and of which there will never be any more (but could potentially be less if someone trashes one) so their value will only increase, but it’s entirely a value based on communal agreement simply because it is rare. (Runescape has a money cap so the phats serve as a way to get around that.) I don’t personally play runescape so I’m not sure about its item quantity.

      Diablo 2 had Stones of Jordan, which had in game value, were small but not stack-able (nothing of value was in D2), and (like ectoplasm) were constantly created and destroyed. Money in D2 was so worthless that the community required something more stable and evidently settled on SOJ because it was as small as you could get with the best benefits and most options. There were other items that could have sufficed as currency, but they weren’t quite common enough to be sacrificial. (Zod runes would look appealing as a currency because they were insanely rare or required obscene investments of time and energy to craft, but their benefit was probably too good to waste on being currency. Zod runes had an emotional investment you didn’t get with SOJ.) D2 used algorithms to generate most of the items in the game so there were a lot of items, but only a small number of them (perhaps less than GW) were consistent and therefore monetizable. (The whole point of a currency being that it is in some way consistent so you can use it as a standard.)

      All of which leads me to suspect that there might be other factors in TF2’s lack of evolved high-level currency. For a secondary currency to evolve in a game there has to be a need for it. In GW and Runescape it evolved to circumvent a gold cap. In D2 it evolved because gold was completely worthless and the likelihood of finding a two way trade for two random items was effectively nil. TF2 doesn’t have a bad currency to replace or supplement. Items purchased in the store effectively use real money. The only reason for a currency to develop then, is in the trading for unpurchasables or for people who don’t want to spend real money. This isn’t a need born of total desperation or frustration with internal limitations that affects every single player in the game. There’s not as much pressure to settle on one.

      Items that are directly purchasable from a store, or that are given away, don’t become unofficial currency because it’s too easy to screw up the market. (The exception being, as in the Runescape case, that no one realizes they’re valuable until it’s too late and they’re exceedingly rare. TF2 players and MMO players in general are too savvy for this to happen much anymore, I suspect.) Conversely items that are too rare are useless as currency because you have to have some substantial quantity to make them monetary.

      And TF2 has a LOT of items that fall into this “rare but not too rare” category. Theoretically any item tagged as Vintage, Genuine, or Haunted could be currency, but they all have a Unique (read: common) version. The only thing special about them is a color change in the name. mildly impressive, but not really a hook. They feel more like “Oh, hey, I found a wheat penny!” than anything truly impressive.

      Sure they might be worth a little more but that’s not what you’re looking for with a currency item, especially one that’s not stackable. If your currency item’s going to consume precious inventory space then it should be reserved for things that are really worth it. The SOJ wasn’t used to trade for mundane items like common rares. You used a SOJ to buy things like high letter runes, legendaries, that last set item. A non-stackable currency is for stuff so rare that you could play for years and never see that one specific drop.

      So what does TF2 have that falls into this area? A fully maxed Strange weapon would probably be ideal except that the counter is reset whenever you trade them. You can get around this with giftwrap, but that’s clunky, not at all convenient, and seems risky. If you mess up then the investment is lost. It’s not stable enough.

      Tools, especially keys, would be ideal, but the basic key can be bought, which means that the market can be overloaded. The might be an acceptable medium-level currency, but I can’t see them being used for anything better than that. Two paint colors seem to have been uncrateable only, and are no longer available. Since they’re consumable if they never appear in the store for purchase I expect they’re currently fairly valuable and will only go up. But if they’re put up for sale they’d lose all value.

      The only category of items that really seems to have potential as possible sources of currency are the Unusual items. And the real problem there is the variety. 168 possible hats that can have an effect times 26 possible effects means 4368 potential items to use as currency. (This is an upper bound, it’s possible not all of the effects can be applied to all the hats.)

      Now you’re into the psychology of the group: are any of the hats “better” or more popular? Since these are functional items and part of the appeal might be to run around declaring “I have money” will an all class hat or a class specific hat come out on top? Which unusual particle effect is the rarest/coolest?

      I suspect that all of these hats collectively function as currency but the pool is too large for any one hat to really float to the top. Certain promotional hats, if they had a wide enough dispersal (but not too wide) might function well as currency.

      The other thing to consider is that TF2 players seem to like variety in their hats. While they might be willing to put up with three or four copies of the same hat simply for monetary purposes I suspect that they would be more likely to trade in a rare hat for one of roughly equal rarity and perceived worth just for the variety.

      I’m not familiar enough with the ins and outs of TF’s item hunt to know about specific coveted items, but since most items seem to be either entirely cosmetic or provide bonuses while being untradeable, I suspect that a currency will fail to develop.

      But if Valve wants to test the theory, they can make an extremely rare drop weapon with nifty stats that stacks and can be consumed-in-job-lots-to-craft-something-equally-cool-but-non-stacking and see how long it takes to become a currency. I don’t think it would take long.

  6. Allen says:

    Confounding my expectations that keys would play that role in TF2, Steam trading data shows that there are five or six items which alternate as currencies. How come? Why does one item within the TF2 economy not evolve into a currency given that everyone’s (trading) experience would improve

    The items I commonly think of as “currency” in TF2 are 1) metal 2) keys 3) buds. Note that in TF2, currency takes an item slot, so TF2 currency is “more bulky” than real world currency.

    I’d be really interested to see the value of the average trade involving only one of these items i.e. keys for hats, as opposed to keys for buds. The former is “keys acting as money”, while the other is “currency exchange”.

    My guess is that all trades that involve keys, buds or metal acting as currency, the values of the trades are very different, mainly because the three items’ values are so different, and because the items that people holding each of these currencies desire, are also different.

    My guess is a large subset of the population has never bought a key. My guess is also that a large percentage of the population has never possessed buds.

    From my own TF2 experience, I largely interact as a “consumer”. I have never tried to arbitrage. I have never bought a key. I own a pair of buds, my own. I have never tried to trade to acquire buds or keys. I do barter for items, and I buy items using metal.

    If you’d like to test the theory that one stable currency will arise, try introducing larger denominations of metal, such that a single item slot holds the same value as buds, and see what happens.

  7. Nick H says:

    This is extremely concise and enlightening. I thank you for your thoughts sir. My sole suggestion would be examining if at possible we can look at only ‘arms-length transactions’ in this new digital economy. Probably not, but I appreciate your analysis nonetheless.

  8. Jimmy says:

    Excellent read. I look forward to these posts!

  9. Zander says:

    I’ve noticed an interesting phenomenon with gift items. At LAN parties, I’ve seen people create servers and then only having two people log on. Each player would activate a Secret Saxton essentially generating a free item per person. I don’t know how prevalent this is, or what other forms it could occur in, but it indicates that some percentage of gift-trades actually have a profit motive. Not quite sure where I’m going with this, or how much it impacts existing models. I suppose it would be interesting, if Valve had a way of tracking these informal trades, to see their prevalence in relation to the update that gave out free Secret Saxtons.

  10. Andrew Plant says:

    I don’t know how much the numbers bear this out. but it occurs to me that from the inside the thing that most seems to resemble a currency is the metals. at least, when I see people soliciting trades in chat it usually involves metal. it’s easy enough to see why, the three types of metal, scrap; reclaimed; and refined, have a clearly defined relative value. it helps to improve the chances satisfying that dual coincidence of want too, because anybody with some surplus items can usually scrounge together some metal, and similarily metal is a major component in crafting just about anything a player might be after. I suppose it is not unlike salt becoming a trade staple because of it’s tremendous utility.

  11. R.C.B. says:

    You make a long and detailed point about the “purity” or lack thereof in the analog and TF2 economy, which you succinctly sum up with;

    “So, until a couple of centuries ago, most trade happened because of unexpected, or unplanned, surpluses: very much like those in the TF2 economy!”

    I believe, correct me if I am wrong, that this is in the context of showing how Adam Smith lacks application to certain markets, especially pre industrial ones.

    Many see his treatise as focused at industrial societies to begin with, so much of his premise seems to remain intact from my perspective. In any case, I think you are being rather hasty in your charactarization. Allow me to elaborate.

    Early formulations of General Equilibrium theory, as you know, attempted to give some mathematical rigor to the concepts of Smith. These formulations state, in little uncertain terms, that economic agents attempt to maximize their own unique utility functions that assign marginal utility to different goods. Of course, the endowments they start with are rarely already optimized relative to their utility functions. In other words, their endowments have surpluses and shortages of goods relative to their true “equilibrium” quantities; precisely the situation you describe!! Yet the toy-agents from the GE-model economy don’t have any problem doing as Smith said, and unloading their surpluses to finance their shortages; they tend to equilibrium, as you said in a prev. post.

    Perhaps the subtext to your post is that Smith did not develop utility nor General Equilibrium calculus, and this was introduced by Marginalists/Neoclassical theorists, rendering Smith anachronistic. However I would be quick to note that I see many of the conclusions reached by these Marginalists to be inextricably tied to Smith’s ideas– especially the invisible hand/general equilibrium.

    However much I respect the attempt to infuse the topic with lessons from Anthropology, I maintain my belief that in your haste to distance the analog economy from the textbook economy (a legitimate difference which exists in reality) you got a little ahead of yourself and wound up with a couple oversteps and inaccuracies.

    I eagerly await more scholarly and rigorous insights from the Valve economics team, as I greatly enjoyed your Arbitrage/Equilibrium barter analysis post.

    A Fan From an Economics Dept.

  12. Bertil says:

    Hi, and thank you so much for your incredible blog posts… A year ago, I had no idea what an econometrician for a game company could do, and I had been one for three months; now, I have someone to look up to.

    Regarding your question to isolate pure exchanges, I’d probably start by recommending robust averages (averages on the median quantiles) on implied price rates—but we both can tell from experience that no well executed modeling plan resist the contact with actual data. Therefore, I’d rather recommend offering some of those data, for all well-intended parties to look further. I’d recommend Kaggle, if you are not sure how to do it; however, Valve’s fame is certainly good enough to justify a stand-alone appeal to candidates.

    Another thing (without trying to go into confidential details I got from clients): I was surprised at how many accounts feed others without anything in return. Anecdotal exploration tells me those tend to belong to family members, the elder generation giving to the younger. By lack of qualitative study, I’m not sure if this is a teenager using his parents’ accounts to farm, or parents having an account only to control stipends… but I’d consider symmetry of the exchange, both as a way to filter out the least pure exchanges and as a way to consider exchanges that are unpure at face value, but that add up to something fair.

    Finally, getting rid of extra stuff your inventory can’t carry probably signal that the rare ressource is inventory, at least at some point; our team has developed inflation models to encompass rarity between so many different candidate for currency. It lead to changing constraints models, and profile—leaving plenty of room for arbitrage opportunities, something that makes the game exciting, because most players feel smarter that most.

  13. Ferlin Sutton Léo says:


    Just though you might like to know the address of websites that deal only with tea mfortress 2 trades :

    and dota2 trades :

    I really liked your post.

  14. Flea says:

    “How will these conventions change or mutate when participants are given the capacity to buy and sell, among one another, using real dollars?”

    They already have a black market for this. A hat with some particle effects on it can go from $250 – $3,000 dollars. There’s also the bizarre world of buds, bills, keys as currency. You can buy a secondhand key for a dollar less than the store. Valve just isn’t getting a cut from this.

    Another factor to consider – alt accounts. People violate steam’s contract by simultaneously running DOZENS of tf2 accounts in text mode for the sole purpose of running tf2 in text mode to acquire drops, the hope being you can convert these drops into metal, keys. Drops are currently like a free lottery where you can get new tickets by putting on a series of fake mustaches. Even if your ticket doesn’t win big, you still get a little something.

    There’s a small handful of power players that try to manipulate the economy to their favor.
    For example, the tf2spreadsheet is a defacto way to ascertain value but it is only ran by one person.

    This is the only economy where 13 year olds think they can have the same amount of perceived wealth as a 30 year old utilizing real money and a server of alt account. Plenty of people try to prey on the ignorant. There are plenty of “I’ll give you a single weapon for your hat!” A craftable hat is worth 1 – 1.33 refined. A single, old weapon is usually worth half of a scrap metal.

    You should also look into scrapbanking. A single weapon normally sells for a scrap BUT you have to find a buyer. If you don’t want to go through the hassle, you can give two random weapons to a scrapbanker. Currently the TF2 crafting rules allows you to craft 2 weapons of a similar type together to form a single scrap. What the scrapbanker is essentially hoping to do is to give away 1 scrap but sell the two weapons he receives for a scrap each. Websites like the tf2 warehouse do something similar on a massive scale.

  15. Kevin Marks says:

    You might find this Ignite talk by Christopher Marks from Google IO called Everything I know about economics I learned by playing video games interesting, as it contrasts RuneScape trading (Which does have an official currency) with TF2 ( it starts at 25:27 if the YT link resets)
    Slides are up on slideshare

  16. leTjuhl says:

    A big question to me is how the game’s systems influence the economic activities (i.e. exchanges of goods, gifting of goods) of players. The amount of items one can place in one’s inventory in TF2 is limited. This is less of an issue for players that play quite irregularly and more of an issue for highly active players (as they will accumulate more items through gameplay). Players can free up inventory space by deconstructing duplicate items into crafting materials. That could also generate an “economy of duplicate generation”, i.e. players barter items with other players more willingly when, via the exchange of goods, they’d generate a duplicate item. Muling may be an issue, where a player creates multiple accounts and then gifts an item from his inventory to an account that is just there to store or hoard items for the “main account”.

    In general, the game’s systems put constraints on the users which they can and will try to circumvent via activities allowed by the game’s systems.

    Considerations like these would ideally also be covered by the above mentioned “radar”.

  17. hain says:

    In terms of identifying gifting; may I hypothesize that
    – gift trades usually have unusual exchange rates, as in 2 X for nothing, or 3 Y for a Z where 1 Y sells for 1 Zs normally.
    – gift trades occur between Steam friends (Secret Saxton is obviously an exception, I am assuming people do not give gifts to strangers if they had to choose who)

    • Wilson says:

      Actually people do give gifts to strangers willingly, there are plenty of ways to use a secret saxton to give an item to a friend but you will find a majority of players don’t do this.
      There is a website where players also give items away for free called tf2 raffle house, you can find it here:

  18. TK says:

    Good afternoon!
    Again an interesting post and quite insightful.

    You might try to use a formula using the friend-list of users and how often he is playing together with the guy he just traded with, as a first step to try to identify impure trade, because the propability for less ‘rational’ behaviour towards someone on your friends list or someone with whom you are playing more often should be higher than for exchanges between people who don’t identify themselves as ‘friends’ and you just only met.
    But well… as you know, there is most likely solution which isn’t influencing the statistics itself.

    I’m not sure if the lack of single currency is not necessarily telling us something about the underlying social norms.
    The distribution of the currency matters for a currency. If some people do not own the currency, they will start to use something else to trade. If it is true for a lot of different people you will allways have competing ‘currencies’.
    Also i think you have to look at the real differences between the trade goods / possible currency. As you can experience in other multiplayer games, where the programmers did not introduce some currency, something else became this currency – if you look at the “history” of Minecraft it became (as far as i am aware) an item (slime balls) that couldn’t be used for anything but was somewhat rare ( which reminds me of gold). And this item was universally useless.
    But this is somewhat different for TF2 items (at least most of which i’m aware of), because a lot of items have the same dropchances and (which might be hard to measure) have the same demand behind them.
    The second though is naturally somewhat undemrined by some real life examples (cigarets in prisons, if the preconception stays true on this example).

    Best Regards

    On a side note, is it just me or is it normal that i only see the comments on the last Post, but not the current one?

  19. Jesper says:

    Great post!

    One part of your analysis left me a bit confused though, where I think your argument was flawed.

    You used the example of a Thanksgiving dinner to explain how not everything can be bought or sold. While I agree that it is largely impossible to buy your way out of a social norm, I see it as perfectly normal to banter in these circumstances.

    “What if I take out the trash instead?”, one would might say. “Well, you don’t have to do the dishes if you take out the trash AND clean the temple”. And so the trading would commence.

    This is an impure exchange for sure, but does at least highlight that trades and exchanges can be employed in almost any situation today.

    • yanis says:

      Of course. But my point was not that exchanges cannot take place within a gift exchange setting. My point was that you cannot pay your way out of the obligations engendered by gift exchange. It was about the impotence of money under those circumstances, even in fully monetised economies such as ours.

  20. Kapsar says:

    I think that this system fits in rather well with an evolutionary economic perspective. The lack of currency could be explained by the limited size of the economy and the size of currency in regard to the equipment. It might simply be easier to barter with actual equipment than anything else. Additionally, if you have the opportunity to trade or give someone in your party better equipment while in game, it likely increases your ability to survive during the game. As one of the commenters above mentioned the timing of the trade or gift could dramatically change the value of an item for the person gifting/trading it.

    For example if I found a weapon that I couldn’t use but a team member clearly had lower quality equipment than what I found, it would greatly benefit myself and the rest of my team for me to simply give that weapon to my teammate. In this way there is something of an unwritten rule that winning for the team is the ultimate goal so contributing more than just your skills as a player would be considered very valuable. Additionally, I may receive a weapon from another person in the same game, or I could have been given one in another with different people. These interactions set a norm within the game that drive the gifting behavior.

    It might be interesting to look at individual users to see how their gifting behavior changes over time. They may initially look to do most of their bartering through selling, but if they had received a gift they may shift their gifting behavior. This would support your argument that impure exchanges are more common and help explain the lack of a currency.

  21. Engraver says:

    I don’t pretend to understand most of what you are talking about. At the start of the only econ class I ever took the prof said, “All the theories we will discuss are based on the assumption that the economy will grow forever”. Such a flawed assumption put me off of economics from then on, which says more about me than economics I suppose.

    When you brought up the subject of gifting in TF2 I was immediately reminded of the theory put forth by Eric S. Raymond on how the open source community works through a gift economy. You can find his paper, in all its incarnations here: The Cathedral and the Bazaar.

    The content may be more applicable to the people in the community who create the digital assets used in TF2, rather than those who use and trade those assets.

    I must admit I have no interest in anything in TF2 that does not enhance my ability to play the game better. This is especially true because user interface for “crafting” is so clearly designed to make it difficult to get the job done that I rarely use it.

    It’s the same reason I don’t play Angry Birds. After 15 minutes playing AB it was plain to see that the user interface was deliberately designed to make the game harder because the game designers couldn’t think of a better way to make the game itself more interesting and challenging.

    It’s similar to the typewriter manufactures that scrambled the letters on the keyboard to slow the typists so they wouldn’t jam the poorly designed mechanism. Deliberately making the interface hard to use is not an innovation in my book.



  22. Lisa Morton says:

    Another game that developed a de-facto currency (actually more than one, as the original item was eventually replaced as currency by another) was Diablo II – the in-game currency (“gold”) rapidly became useless due to rampant inflation and players exploiting bugs in the game to introduce counterfeit gold and items, which led to nearly all high-level trades in the game becoming denominated in “Stones of Jordan”, a rare ring that was useful both in itself and for the in-game ‘crafting’ of new items.

  23. Red says:

    I believe you could solve Europe’s debt crisis by introducing an economy based purely on key and hats.

  24. Kowalski says:

    Very interesting read, but like others have said, I think there are a few interesting ways you could find more information about each trade: friend status, shared community, how often one trades for less than ‘market value,’ even size or value of inventory compared to how often they trade. I’ve always been interested in how economies work and what affects what, the TF2 economy has always been more frustrating than anything for me because things seem to fluctuate in value every other minute and only a small handful of people own the vast majority of the rarer or special items.

  25. Zeus says:

    Another thought on the process of monetization.

    While the simplistic theory of monetization arriving on account of money (metals, wheat, cigarettes) acting as a lubricant, eliminating the double coincidence of want, does seem to be just part of the story you might need to elaborate on the debt accounting explanation.

    While the Mesopotamia records do offer some support to this, they by no means amount to proof, not even of concept.

    Personally I would choose the accumulation of wealth function of money to be its driving force as that (in my mind) would be the first thing the elites of the time would deem most important.

  26. Joe says:

    I’m not sure if you are familiar with the TF2 black market but it really drives the several different items being different prices. Off the top of my head, keys = $1.50ish. Bills Hat = $10-$11, Painted Bills hat (paint adds a little to the price depending on the color etc) $12-$13. Earbuds = $30. Maxheads = $65 , hats of undeniable wealth and respect = $300. Oh yeah and refined metal =~ $1. So really the stuff is anchored in the US dollar still.


    • Sonny says:

      This. This. This.

      A huge part of the TF2 economy is people selling items for cash. Which is not at all reflected any of the data that Valve has access to.

      I suspect that many “gifts” are actually trades where the other person is paying cash.

      Power sellers also try to disguse the fact that they are selling for cash by asking the buyer to throw in a low value item, so it looks like they are trading item for item.

      e.g. they will trade an unusual hat for a backburner. But the person who gives the backburner also sends $100 through paypal. The paypal part of it is invisible to Valve.

      This kind of thing will skew any of the data valve gets unless they implement some kind of cash trading system built into steam with comissions equal to or better than paypal.

    • SUNSPY says:

      The desire to sell items for real money drives the economy in directions that have been unexpected from where it started. Unusual hats are sold for MONEY (earbuds=money) and the only way to obtain those is buying keys, or buying them directly outside of Valve’s system. I’m sure this has been great for key sales, but I won’t be surprised if keys aren’t #1 selling item soon.

      Initially, people would freely trade unusuals for other unusuals, since every one had a stable earbud/cash price. Earbuds were cheaper relative to keys giving a reasonable discount to PP users over non PP users.

      The recent year has seen a shift to users only wanting promo items like earbuds which have stable demand and prices, or even rising demand and prices, which in turn drives even further demand and price increases for the said items. People don’t want risk, so public consensus or “sheepconomy” as I call it took over.

      IMHO, hurry up and take over the PayPal trading with Valve controlled money trading or Valve will suffer badly. There is talk of Valve acting against this black market ? that is insanity, unless Valve handles these trades(SALES) that people want to make. What of it Valve ? Steam Trade Beta discussions were very promising but Valve is being very.. SECRETIVE.

      That’s a shame, since it affects such a wide group of unusual traders, owners, investors and unboxers.

  27. George says:

    Yeah, you go Valve. Keep finding new ways to suck money out of the weak under-class. I hope you are putting that money to good use to build a new awesome game. And by awesome, I mean a game that any person can buy, install, and have a good time playing. Unlike the Left4Dead(1+2) community, which is the most vile and disgusting group of gamers ever. TF2 is full of little kids and people who have no idea how to play.

    Anything for a buck, right?

  28. Pax0 says:

    As has been noted, this is a superbly interesting read throughout.

    The observation I find most interesting, however, is that no one particular item has yet gained currency status- not even keys or crates!- and instead there are 5-6 items which are “pseudo-currencies” of alternating prominence.

    My first inclination would be to think that the various grades of metal used in the crafting system could easily form a multi-tiered currency system (Much like the British currency setup [Pounds sterling>Schillings>Pence] before its decimalization in 1971). However, in practice, this would actually be highly impractical, given that each unit of metal occupies its own backpack slot, greatly limiting its ability to be stockpiled and stored in large quantities, especially for players who have accrued a large collection of items.

    Furthermore, it seems to me that in order for an item or commodity to attain currency status (either as a lubricant of exchange or demarkation of debts), the ratio between its supply and the volume of its practical use must be above a certain threshold, and the net supply (gross supply minus the amount consumed in said practical use) must neither be too high, nor too low.* For example, this would be why precious metals came to be used as “currency” in ancient societies: they were relatively rare (but not overly so) and had little practical use outside of jewelrymaking. This then, would explain why neither chests nor keys have managed to attain currency status, as they are consumed at very high rates, on top of the aforementioned storage issues.

    *Interestingly, by this principle, a return to any form of gold standard, as advocated for by certain fringe groups, would be next to impossible due to the importance of gold in the manufacturing of advanced electronics and various other industrial uses for it which have been discovered since the global abandoment of gold-backed currency in the mid 20th century, but that’s another discussion for another day.

  29. Guy says:

    Thanks for a very insightful and interesting post!

    This Makes me moist

  30. PurpleXVI says:

    Thoughts on the “Gift Radar.” I’d say that there are some things you could rapidly filter out in an automated fashion.

    Firstly: Trades where only one party gives something. These are obviously gifts, unless some sort of off-Steam service is being traded, but in that case it cannot be measured and should probably be filtered out anyway.

    Secondly: Trades made with people on one’s friend lists. Usually the people we have on there would be people that we appreciate for some sort of social reason(we like playing with them, talking with them or something else), rather than as regular trading partners, so there’d almost certainly be some sort of “impure” factors employed there(people are probably a bit more scrupulous in negotiating with people that they don’t want to alienate), than someone with whom they have no prior relationship. It could perhaps be regulated by some sort of little counter, if people communicate with this person more than X per month(would require no monitoring of conversations, merely marking that they exist), it’s assumed they have a social relationship that matters, if not, it’s assumed they’re socially distant enough for them to go for maximizing value in trades.

    The rest of what you posted was fascinating and I have to say, you have a great ability to make economics, and the history of such, interesting to the layman.

    • Seth Bauman says:

      There is a flaw in filtering out trades with people on your friends list, because making a trade using almost any trade website, you have to add the other person in order to complete the transaction. And, usually these trades are as close to the market value as possible since there is much more supply in these concentrated areas than in isloted trade servers where there is much more room for arbitrage or ability to make ill informed trades.

  31. Jonathan Wright says:

    Hi Yanis,

    The points made here remind me a lot of David Graeber’s book “Debt: The First 5,000 Years” (, right down to the examples chosen and the way the concept was introduced. I was surprised to not find a footnote or the conclusion referencing the book. Are theses ideas now well accepted within certain circles and Graeber is collecting and popularizing the ideas?

    Jonathan Wright.

    • yanis says:

      More or less. For instance, I have been arguing that the origins of money is to be found is debt and taxes since my first book was published in 1989. Even then, such arguments were not original, in that they reflected a long tradition of monetary history. Graeber himself explains that he has compiled a number of these points.

  32. There’s one additional contradiction worth thinking about here, and this is something which is just beginning to seriously impact digital culture more generally. This is the uneasy coexistence between various forms of player creativity (fan labor, fan exchange, fan culture) on the one hand, and digital commercial entertainment on the other. Valve is run by some very smart and honorable folks who have long represented the best aspect of gaming, but they’re constrained, like all commercial studios, by the need to earn enough money to keep their business going. Fan labor is mostly non-commodified in nature, and is spreading and scaling up in complexity very, very quickly. My best guess is, the entire field of videogame in-game economics is going to generate a whole range of complex hybrid or mixed economies – part patronage, part advertising, part commons, where player communities function as the primary legitimating bodies, rather than state regulators (though states are of course still there).

  33. Santiago B. says:

    Whilst this study was very well done, I believe the surveyors did miss something when they say that periodically only 5-6 items eventually become currency. The thing about those items, like keys for instance, require actual currency in order to get into the market, keys are not items that drop randomly on players. The main reason someone would buy a key would be to open crates, which once the key is used on the crate it is gone forever. That’s why keys can’t really represent currency, because unlike currency keys hold an actual practical value.

    What really did represent currency though were the pieces of coal during the last winter sale Valve did over steam. Everyday any player could complete a steam community based achievement (essentially a freebie) and 6 other achievements for video games and the most common prize was a lump of coal (much like someone would work at any service job and be paid their wage). With the coal one was able to cash a number of them in for a video game or coupon, or hold onto them so that the player would be entered to win a prize of every single video game on steam. Soon enough trading over steam turned from the bartering mentioned in the article to using coal as a lubricant in the market, and for many players the coal had no true value. It was very interesting to observe, the players even placed a value for coal, some players were even using paypal to purchase coal from others. If I can recall correctly, on the first day when the max coal a person could have earned from achievements was about 7 lumps, the value of coal was estimated at about 0.50 USD (or two pieces of refined metal as those of the TF2 community would tell you), once a week or two passed and millions and millions of lumps of coal were introduced into the economy, the value dropped to a mere 0.13 USD each. Very similar to currencies in real life where their values fluctuate based on supply and/or demand.

    The entire winter sale would have made quite an excellent subject to write about. Already there are two possible topics which stick out: 1) Whether the introduction of a free but limited resource would manifest into a temporary market fueled by the conflict Marx wrote of; and 2) When gamers purchase titles at greatly reduced prices, are they as content playing it as they would have been if they bought the title at full price, or do they just feel less compelled to play the title because they have a lower threshold to when they have fulfilled their “money’s worth”.

  34. Skyl3lazer says:

    Nice entry, but any chance that in the future you could make your images more readable? I’d love to inspect that graph in closer data but it’s actual size :\

  35. Tony Hua says:

    While impure exchanges definitely affect how the market functions, I believe a bigger hindrance to the development of the TF2 economy to a more pure exchange can be explained by the opportunity cost of trading. A partial explanation for why arbitrage continues to exist may simply be that it’s costly to trade. Trading in TF2 is very cumbersome by comparison to more refined markets such as the local grocery store. I could spent an hour going across trade servers trying to “buy” low and “sell” high, but I would be forgoing an entire hour of playing the game. In addition, the TF2 economy tends to be segmented into local “bubble” economies rather than a large unified system. As a trader, I may want to seek arbitrage in order to increase my stock of items, but finding arbitrage would involve jumping from server to server.

    So if there is a real and definite cost to trading, then anything that makes trading easier would reduce the amount of arbitrage in-game. Something like a Steam marketplace that allows players to easily and reliably put their items on a well designed Steam UI would push towards a market equilibrium faster. And if contemporary economic theory is true, then the Steam community would be better off as a whole.

    A common property for a currency is that it has to be widely accepted. While Keys would make for a seemingly ideal item to evolve into the role of a currency, it also isn’t widely accepted. Part of the reason is that Keys are very high valued items, and that makes them cumbersome to use in trade because a Key may be worth 5 Medic Guns, but only 2/3 of a particular Hat. Most players do not carry large enough volumes of items to make it feasible to trade for Keys. Since traders cannot split a Key up, they will have to accept a nonequilibrium price or forfeit the trade. It would be the equivalent of what would happen to the US dollar if every unit of currency below a hundred dollars was removed. If Keys could be split into one hundred small pieces (and they would need not take up too much of a player’s inventory space), then I can definitely see Keys becoming a currency in TF2.

  36. Gkm says:

    I disagree vociferously with the approach taken to estimate what is the money in this “economy” and its value.

    “when barter ceases and money has become the common instrument of commerce every particular commodity is more frequently exchanged for money than for any other commodity” Adam Smith

    Therefore simply look at the good that is exchanged the most and you’ll find the money.

    Next, what is its value. If Valve has the time spent online of all players since “goods” creation began, then that is the labor involved to create them. Take the number of trades of each item and divide by the labor hours and you have the valuation of the commodities or goods.

    Not knowing if the goods are fairly generic in all respects, then some grouping may be required. For example you’re not going to be able to differentiate the value of a Mona Lisa hat from a Picasso hat- they’re just hats as long as they serve the same function or no function at all.

  37. AZ says:

    One observation about the comments on this blog post. The community of gamers is by far more serious and intelligent than the community of journalists, politicians and (some I hope) economists !

  38. Obey says:

    One thing that complicates the establishment of a currency is who can “mint” said currency. Metal is continually being made, but like crops in agricultural barter economies, metal is also consumed. Buds are not being produced (unless I’m mistaken); there is a certain number of them out there and that number very rarely changes. Further, one currency can permanently be converted (or traded to Valve, as it were) when items are crafted.

    That said, many traders discuss trades in terms of metal: I frequently see in-game discussions where someone offers 1.33 refined for a strange “Something”. People are trying to establish a primary currency; it may simply not have happened yet. While there are millions of users making millions of trades, the entire economy has only existed about 4 years. And this will all be in flux each time a new game is introduced into Steam Trading (like Dota 2, perhaps).

  39. Kaseluris-Nikos-1959 says:

    My explanation why no TF2-item is becoming “money” is: because there is NO NEED for this!!!

    All these items are virtual-goods and at the same time “real-world goods” because through the steam-platform gamers use real-world-currency to own them. THIS real-world-currency is the currency for the TF2-items.

  40. Kaseluris-Nikos-1959 says:

    My explanation why an TF2-item is not becoming money (currency is a unit-of-measurement of money) is that there is NO NEED for this!

    TF2-items are “virtual-goods” [1] and at the same time “real-world-goods” because gamers use “real-world-money” in the steam-platform to buy them. THIS real-world-money/currency is the currency in the TF2-economy.


  41. Random Person says:

    Very pleased you now allow comments.

    I’m a little unclear on the pure vs unpure exchange and how most trades would be classified. Certainly if I run into a person I’ve never met before on a trade server, trade with them, leave, and immediately forget their name and have them and everyone else there forget mine, that would be a pure trade in the sense that it’s disconnected from social obligations. But most trades don’t occur in that fashion. If you really want to move an item without having to join a ton of trade servers and just *hope* you run into someone who wants that item, you use one of various web sites. Someone has mentioned the TF2 trading Reddit, and there are various other sites including TF2 Outpost and TF2 Trading Post. Here people propose trades and other users can post to accept the trade or negotiate different terms. If you are happy with a deal, you can leave a positive review that enhances the trade partner’s reputation.

    At a low level of trade this does not mean anything, but anyone trading more valuable items or anything for real money (you speak as if this is a future possibility, but it happens all the time already) will look for a trader with a good reputation (see From reading the reviews left for traders people tend to evaluate a trade based upon three criteria:

    1. How good a deal they believe they got.
    2. How rapidly the trade was completed (ideally a proposed trade is accepted immediately on posting on the website, a friend request sent and accepted, and the trade completed within minutes from initiation).
    3. How *polite* the seller or buyer was.

    People may want to get a good deal for their trades, but they also want to trade with someone who is likable. For instance, if a trader calls your unusual a “crappy low-tier hat” with a “cancer” effect, you’re more likely to reject their offer even if it is reasonable or even generous. Various reputation trackers at different sites may pressure traders to follow certain social niceties (and definitely pressures them to not rip people off!). Would this make the trade less pure?

  42. Jeremy Kun says:

    Just so you know, the marketplace for Diablo 3 (known as the Auction House) alleviates a number of these issues to create more transactional purity. First, gift exchanges between friends are done in-game by dropping items on the ground, and cannot be realistically performed in the marketplace. Hence they are not tracked, and hence they don’t pollute economic data. Second, the exchanges in the Auction House are entirely anonymous, so there is literally no relationship between the two parties in the exchange.

    Of course, there are some differences: ‘gold’ is a forced in-game currency and, while there is a bartering system, it does not occur in the Auction House. Moreover, there are many separate auction houses: one for real-money exchanges, one for in-game gold exchanges, and one for ‘Hardcore’ players (where when your character dies, you permanently lose all of his gear and progress in the game). I’d love to see a comparison of the development of these three isolated, but inherently similar, economies. It’s unfortunate that wonderful economists like yourself don’t have direct access to Blizzard’s data, but it’s something to think about.

    Keep up the great work! I’m already a hooked reader.

  43. Dan Mawson says:

    Speaking of information you can gather from Steam Data, this suggestion which cropped up on Ars Technica is quite interesting:

    Tundro Walker | Ars Tribunus Militum
    jump to post I looked at the first comment in one of the links, and the person said “I haven’t picked up Dead Rising 2, b/c it hasn’t hit my threshhold yet.”

    This seems like an amazing idea they should build into Steam. You can go find all the games you want to buy, but you just tag them with a price you’re willing to pay based on what you think you’re going to get out of it. Then, Steam will keep that list logged, and when the games go on sale and go below your threshhold, you’ll get an email or pop-up or some other notice letting you know. Or, maybe you could set it to auto-purchase.

    This could benefit 2 fold.

    1) it would let Steam track stats of what folks think a game is worth

    2) it could help folks buy games w/o having to actively price-monitor all the time (in the long run, a user will probably buy way more games if the sytem baby-sits the price-monitoring for them.)

    3) Steam could set threshhold limits of their own, like “if X people want to buy this game at $10, then we know we should have a sale on it for $10 pretty soon before they all lose interest”

    I haven’t dug into Steam that much, so maybe this already exists. If not, it seems like a really good idea.

    The applications go more broadly than this however, for example if someone implies they would buy a game for $10, but when it comes on offer for $12 they buy it anyway does this tell us something interesting about consumer behaviour in terms of time inconsistent preferences etc.

  44. Synagonism says:

    My explanation why any TF2-item is not becoming money (currency is a unit-of-measuring quantities of money) is because there is NO urgent NEED for this!

    TF2-items are “virtual-goods” [1] and at the same time “semi real-world goods” because gamers use “real-world-money” to buy them. I call them “semi” because game-companies are clever enough and do not allow to shell back for real-money, only to other gamers. My opinion is that this real-world money used to buy the virtual-goods is the semi-money in this virtual-economy.


  45. John says:

    Dear Yanis,

    Thanks for another interesting post. It sounds as though you may have read David Graber’s latest book, Debt the first 5000 years. Given that the gaming community is made up of many very smart and talented people, I hope your blog will lead to greater interest in economics, and its relationship with the political, social and cultural spheres more broadly.

    Kind Regards,


  46. Don Quigleone says:

    Are not impure trades pure in a different sense?

    For instance, I could give a person a gift in order to gain goodwill with a person. I am trading the monetary value of the gift, for the benefits of a better relationship with the person. Likewise, it is a sign of an unequal relationship when one member gives more then the other (and puts up with it). That person feels they get more from giving gifts to that person then the the other party would gain from giving gifts to him.

    The Chinese have a concept of “Guanxi” which can be summed up as reciprocal relationships. A chinese person will do favours for another in order to gain guanxi with that person. That person will, in the future, be obligated to repay that favour, and in fact, it can quickly become a never ending chain of favours between the two people.

    For instance, you might contact the official you helped pass exams to smooth the way for your son to get a visa to go abroad. Later, that same official might also call on you again. Etc etc. But if either party feels they are being exploited, they will break it off.

    In the thanksgiving example, you have a similiar case. In a family, people freely exchange their labour with one another. But how willing will people be to do the tasks if they feel another member isn’t pulling their weight? I might choose to do the dishes so that my mother will have more good will towards me, and not nag me so much (escape from nagging is a good too! Albeit a kind of racket…)

  47. Adam says:

    “at least for me” is intellectually lazy. Something is either true or false for everyone (except in the case of opinion, which this is not)