Earlier this fall, an article came out on Geekwire’s website featuring an interview with Gabe Newell, the man in charge of one of the most successful PC game developers and distributors – Valve Corporation. This interview highlights a lot of strange, interesting, and puzzling phenomena found in the online PC market that puts a lot of economic principles to test.
Valve is in a unique position to collect data and analyze economic theories, being the largest distributor of online PC games through their Steam distribution service. Through Steam, Valve can observe consumer behavior, purchases, and game activity in real time and base policies on data trends.
I want to highlight three interesting insights that emerge from the interview, and provide some brief commentary on what makes them so strange, and how we might go about exploring their consequences. Some of the language of the interview is a bit unclear, so below is my attempt to glean the basic results of it.
Piracy and Service
On the subject of internet piracy where it is so easy for players to simply download proprietary software without paying, Newell surprisingly describes:
One thing that we have learned is that piracy is not a pricing issue. It’s a service issue. The easiest way to stop piracy is not by putting antipiracy technology to work. It’s by giving those people a service that’s better than what they’re receiving from the pirates.
On a personal note, I find it extremely refreshing to hear comments like this when many developers and trade associations would rather spend millions of dollars on rent-seeking for government-backed intellectual monopolies and litigation against violators. That’s money that could have been spent innovating and improving games as opposed to simply enforcing the status quo of already-produced games.
Newell and Valve would rather focus on the serving the customer. And Newell’s statement is true, to boot – that people will buy games if they expect a better quality experience from the product than if they were to illegally download them and risk a lower quality experience. Downloading software increases the risk that one will be exposed to a virus, a poorly-modded version of the game, an incorrect keygen, or other unwanted discomforts. Pirate software also may void the possibility of getting custom support or add-ons to games. By focusing on customer service and feedback, Valve can sidestep much of the issue by out-competing pirate software which has crap service.
How Do Consumers React to Price Changes?
Now we did something where we decided to look at price elasticity. Without making announcements, we varied the price of one of our products. We have Steam so we can watch user behavior in real time. That gives us a useful tool for making experiments which you can’t really do through a lot of other distribution mechanisms. What we saw was that pricing was perfectly elastic. In other words, our gross revenue would remain constant. We thought, hooray, we understand this really well. There’s no way to use price to increase or decrease the size of your business.
Assuming Newell’s data is correct, this is a very intriguing conclusion to make.
For noneconomists, price elasticity (of demand) is the responsiveness of consumers to change how much of a good they demand when the price of that good changes. Technically, it’s the ratio of percentage change in quantity demanded to the percentage change in price. It’s usually a proxy to measure how “flexible” consumers are, and businesses judge how much revenue they can raise by altering the price. A business’ revenue is simply the multiplication of the price (per unit) times the amount of units sold.
Perfect elasticity means that the (percentage) change in quantity demanded is infinitely larger than the (percentage) change in price. Intuitively, this means that any attempt to raise (or lower) the price will result in demand dropping to zero, and nobody will buy the product, period. This is usually the case (in a less extreme form) where there are many substitutes for a good (e.g. cola-drinkers will switch from Coke to Pepsi when Coke gets too expensive).
If demand for Valve games truly is perfectly elastic, that means that players have a certain price in mind that they will buy games if, and only if, the games are sold at that price.
Valve continued testing the elasticity of their customers’ demand to find falsify this theory:
But then we did this different experiment where we did a sale. The sale is a highly promoted event that has ancillary media like comic books and movies associated with it. We do a 75 percent price reduction, our Counter-Strike experience tells us that our gross revenue would remain constant. Instead what we saw was our gross revenue increased by a factor of 40. Not 40 percent, but a factor of 40. Which is completely not predicted by our previous experience with silent price variation.
This would imply a very elastic demand, but not a perfectly (or infinitely) one. That implies that the (percentage) change in quantity demanded still overpowers the (percentage) change in price, but customers are still willing to purchase the good at different prices. This means that total revenue (remember, it’s price times quantity sold) skyrocketed since a slightly lower price is overpowered by a much greater quantity.
As for accounting for the variance among these experiments, several explanations might be possible. Different games might face different demands by consumers, so an experiment in one game might not be perfectly replicable on another. Newell’s team accounts for the variance in the data with a different explanation:
Then we decided that all we were really doing was time-shifting revenue. We were moving sales forward from the future.
This means that more people who were probably going to purchase the game in the future (when they had saved enough money perhaps) purchased it immediately upon noticing the drastic reduction in price. This also risks something called “cannibalization” – a phenomenon where two products from the same company (in this case, present and future purchases of the same game) compete with each other and the success of one product may destroy the other (buying the game now prevents one from buying it in the future). But again, Valve’s research falsified this hypothesis:
Then when we analyzed that we saw two things that were very surprising. Promotions on the digital channel increased sales at retail at the same time, and increased sales after the sale was finished, which falsified the temporal shifting and channel cannibalization arguments. Essentially, your audience, the people who bought the game, were more effective than traditional promotional tools. So we tried a third-party product to see if we had some artificial home-field advantage. We saw the same pricing phenomenon. Twenty-five percent, 50 percent and 75 percent very reliably generate different increases in gross revenue.
It’s difficult for met to interpret what the final results were, but it seems to be that the price is fairly elastic, and this effect is amplified by word of mouth: a few people realize that a game (maybe their favorite game they already own, or one they had always wanted to buy but it was previously too expensive) is now vastly cheaper, and tell their friends to buy the game. The amplification may also be due to the fact that many of these games are multiplayer, so someone who already has the game benefits from their friends buying the game and joining the fun. This, in turn, creates a greater incentive for the game-owner to tell their friends about the sale.
“Free to Play” is More Attractive than Simply “Free”
Valve also experimented by advertising some of its games as not only “free” in the ordinary sense of the term (a price signal), but “free to play.” They discovered a surprising surge in demand when something was “free to play” but not necessarily when it was simply “free.”
Why is free and free to play so different? Well then you have to start thinking about how value creation actually occurs, and what it is that people are valuing, and what the statement that something is free to play implies about the future value of the experience that they’re going to have.
I think this difference highlights two interesting facts:
First, language is extraordinarily meaningful and we construe all of our actions within the web of language. A single word, and all of its connotations, can dramatically alter our actions.
Second, Newell seems to imply here that players are making an intertemporal choice (comparing the value of something now compared to the same object’s value in the future) – expecting that something that is “free” means that the developer is no longer interested in maintaining the quality of the product, whereas something that is “free to play” implies that the developer will continue to maintain and marginally improve the product in response to customer feedback. Something that is “free” simply means instant gratification with high diminishing marginal returns, but something that is “free to play” may imply a longer stream of benefits extending into the future, thus making its present value much higher. Again, this boils down to the players interpreting the semantical difference of “free” versus “free to play.”
I can’t say it any better than Newell’s own summary of the odd economics of modern gaming:
“We don’t understand what’s going on. All we know is we’re going to keep running these experiments to try and understand better what it is that our customers are telling us. And there are clearly things that we don’t understand because a simple analysis of these statistics implies very contradictory yet reproducible results. So clearly there are things that we don’t understand, and we’re trying to develop theories for them.”