Some simple economics of ICOs
I have a new paper out (ungated here) with Christian Catalini (or ‘Dan’ as we call him) that examines initial coin offerings (or ICOs). These are the big new thing in fund raising for ventures whereby startups issue tokens that are redeemable for services that they may provide or facilitate in the future.
ICOs have raised billions for startups over the past few years often in record time. Not surprisingly, there is good reason to believe that many of these are not on the up and up. After all, if people will give you millions ahead of time for digital tokens, what precisely is your incentive to develop a new service just to effectively have those tokens returned to you?
What we wondered, however, was whether ICOs might do something if such outright fraud or opportunism with guile was not possible? That is, what if you actually had to incur development costs and provide the services promised?
This turned out to be a non-obvious question. For any would be token buyer, it is not obvious what the token’s value will be in the future. If you potentially value a service, what you care about ultimately is the token-denominated price. If you are a speculator, you will care about the dollar-denominated price which is the exchange rate multiplied by the token-denominated price. So you need to have some expectations that the token-denominated price will not be so low or so high that there is little demand for tokens. All that is in the hands of the startup.
What we find is that the startup can, by setting the token-denominated price in a certain way, induce the revenue maximising (or monopoly) price for its services. (We assume those services are unique so a monopoly price is what we would expect in the absence of tokens). The start-up takes the tokens on issue and divides it by the number of customers (or units sold) it wants to target in a period. Then, there is just enough scarcity of tokens so the exchange rate adjusts so that the dollar-denominated price is the monopoly price. If the start-up wants to do the same in future periods, there is a rationale for setting expectations this way.
So far so good. Basically, although the timing of payments is a little subtle, a startup issuing tokens get use them to set prices the way they might ordinarily do so. Why, then, use the tokens and not just do that? If the startup had to raise equity finance, they might not want to cede that equity. If the startup had their own funds, then that is a reasonable question. Either way, selling tokens initially, allows information the startup does not have about the demand for their own services be revealled earlier. Buyers might purchase them (in the same way they purchase crowded funded goods through, say, Kickstarter) and this might reveal demand. Speculators may gather information and do the same. What the startup gains is the ability to set a minimum level of funds needed to issue the tokens and so insure themselves against demand being too low. In other words, ICOs are a way of aggregating information before important costs are sunk.
Things get more interesting when demand grows overtime and there are incentives to induce people to hold tokens for a period of time rather than spend them when the opportunity arises. Monetary policy then comes into play. The paper discusses this in detail but the bottom line is that there may be value to committing to various rules in monetary policy if you need to raise lots of money up front. The reason is that a token sale can really only cover revenue from ‘one period’ (whatever that is) and you may want that period to be the period with the most revenue. And that may be farther in the future. Saving is what makes monetary policy and, indeed, money interesting.
Looking at the ICO market, it is hard to imagine that participants (on both sides) really understand all of this. And our approach is really just a first pass. We wanted to put the best case forward for ICOs. The good news is that we found that case. The bad news is that it is not clear that case is driving current token demand.
