Holdup, innovation, and platform governance with Ethan Bueno de Mesquita | a16z crypto research talk

Transcript

okay great uh I'm excited to uh introduce our speaker today for governance work uh governance week we have Ethan Bueno deos joining us uh from the Harris School of public policy at U Chicago where he's currently the acting Dean yet still finding time to do research and presenting to us today on uh a really interesting question that's very relevance to web 3 which is about holdup Innovation and platform governance thank you Ethan take it away thanks so much for having me it's really fun to be here it's great to hear what everybody's working on and appreciate the opportunity to present this work I'll say this is this is super early stage preliminary work um two months ago there were no results of any sort a month ago there were a bunch of results but it turns out every single one of them was wrong um so these ones I'm

convinced are right um but this is early work I'm really interested in your feedback and trying to think through what is the right set of questions to ask of this kind of a model um whether this is the right model exactly or not let me start with just some motivation platform economies of course or there the platform markets are common in lots of settings right so there's a whole extensive literature uh in organizational economics about platforms which are really just uh sort of markets characterized by sort of like a two-sided Market where there's kind of a a market maker in the middle who is putting one side of the market together with the other side of the market right setting prices for one or both sides of the market and and the service they're providing is connecting them so like a

canonical example of that from the not uh sort of Internet economy would be something like a credit card company which is putting sellers on one side and buyers on the other side together right and they're really just the intermediary um providing this matching Service uh between buyers and sellers um of course platforms have become particularly important in the world of um sort of the web and web Commerce you think about Amazon you think about Facebook you think about a lot of the sort of important sort of web 2 companies are their their whole thing is that what they are is running a platform that puts two sides of a market together whether that's content creators and content uh uh consumers like on a social media platform perhaps or if you think about

um say a a video game company that's putting video game creators and video video game players together on a two-sided Market um so I'm going to think about platforms platforms are characterized by lots of things this sort of ided Market is the essential nature of them but they are often characterized but we by what we think of as as lockin that is that once you make your way into say providing your product on or consuming a product on one platform it's hard or costly for a variety of reasons to move to a different platform and those lock in effects come from lots of things so they come from sort of network effects and external network externalities inside so you want to be it's more fun to be on a social media platform where all the other people are right and so you know

we're seeing this now with the decline of Twitter and everybody sort of vying for control um it's just more fun to be on a twitter-like platform when there's lots of other people producing content there too and so if one if one platform is already dominant it's kind of hard to move to another one because because of the the increasing returns to scale search costs are another possible source of of um source of lock in this like you know I buy my stuff on Amazon and if I wanted to go buy somewhere else it's not that there maybe aren't any other uh e-commerce companies where I could buy pens where I currently buy them on Amazon but I have to go find the name of that company I don't know it off the top of my head and it would take me so there might be lock and effects there might

also be sort of then the one I'm going to focus on today lack of interoperability so maybe I build a product to the specifications of your platform in some way and now if I want to move it to another platform its quality is going to be degraded because um because it wasn't built for that platform or I'd have to bear some costs to to revamp it for for a new platform you think so you think about I made a video game for Sony PlayStation and I want to move it over to I don't know any other um that was it uh we have a Sony PlayStation um I want to move it over to an to Xbox and maybe I got to rewrite some code right um or maybe it just doesn't work as well same thing I made an app for for for iPhones and maybe

moving it over to um to another to another uh operating system doesn't work as well so I'm going to focus on lack of interoperability but I think there's lots of forms of lockin that happen on platforms um and Creator lockin in particular uh create what we sort of call in in called in sort of industrial organization um a holdup problem it creates a Temptation for platforms right to increase prices that once you're locked in and it's hard for you to move somewhere else uh they want to extract some more rents for you they're like oh you're outside option has gotten worse uh let me take some of it and we see this a lot I think if you think about I'll talk about this a little more think about web 2 we see a lot of this kind of ratcheting up of the amount of the Surplus or the amount of

the rents that the platforms are taking over time time they get people in with kind of low prices and then as they get more and more locked in for some of these reasons um the platforms start ramping up prices on them and that creates some problems that creates some problems for markets in particular sort of the canonical holdup problem is if I'm thinking if I'm a Creator and I'm thinking about that and I know my rents are going to be eroded over time I start becoming more and more reluctant to invest in creating things of value because I know that I'm probably going to not get the Surplus that I generate that that I'm going to lose it because I get locked in okay and so I want to think about about that a little bit um so just to sort of give you a sense that

at least in in web 2 um platform holdup is a real thing you just I mean if you start paying attention to this you hear these stories just constantly of the App Store you know increased the fee for inapp purchases from 20% to 30% or Etsy increas the amount of fees it's going to charge to its sellers once you start thinking about it you're going to hear a story on in the newspaper or the radio every week about this happening okay there's lots of and lots of what looks to me like platform holdup going on okay so the approach I'm going to take to try and study that and what I'm interested in here is both that problem and potential solutions to that problem and I want to think about potential solutions that look like interoperability as well as potential solutions that look like governance um

so the approach I'm going to take is I'm going to think about a Model A very simple model of platform competition okay where I'm going to think about a competition between two platforms and they are competing for two sides of a market okay for users and for creators uh I'm going to keep my model simple by for now only having one Creator okay lots of users but only one Creator um so that Creator is going to be very important the Creator can Exane make an investment in generating a better product which creates social surplus but they don't have to and it's costly to them to do so the platform would like them to in a certain way because create because investing in value um generates sort of more more it creates demand and it's going to create sort of you know more more profits to be to be shared

among these various actors and more Surplus there's going to be lock in due to limited interoperability um and I'm going to consider two kinds of responses to that lockin one is increasing interoperability and the other is devolving government governance such that uh the platform can in some way commit to the creators not to engage in this Behavior by giving them some governance rights okay yeah so like with the extreme form of what you mean by devolved governance just mean like the platform literally just is aifi like what like is so what I'm going to study here because I've only got the one Creator and I'm going to devolve power to the Creator is I'm going to give them a veto over changes to to fee so they're going to be allowed they can just say no uh and you might think about in a model

that was richer that had more than one Creator I might create some sort of a democracy where I let them vote on whether or not to I was say is this going to be meaningfully different from just instituting a commitment upfront so the platform could give the Creator veto power or the platform could just like you know use some form of immutable code to say this is our price forever is there is there a sense in which these are going to lead to different results not an equilibrium it's going to be the same got cool thanks um though I think you can we can talk at the end I think you can think about extensions here where those those wouldn't quite be the same that are I think interesting um but which I haven't gotten to uh funnily enough I think this is actually the

essentially the exact same slide Andy put up yesterday right which is just to convince you like like web 2 platforms and and Beyond are thinking about this kind of devolved govern governance um not always for holdup reasons or primarily for holdup reasons but they are interested in this kind of idea of devolving platform governance over important decisions and here we're going to think about them making devolving uh governance over actually a decision which is really business relevant right about about a Commerce decision these are the kind of questions I'm going to work try and think about when does lockin undermine Creator investment does interoperability solve the lockin problem can governance solve the lockin problem what is the platform's willingness to pay for each of these these two kinds of solutions look like how does the platform feel about about about interoperability

versus governance as solutions to this kind of problem there's going to be a bunch of results I think if I went through this you wouldn't have any idea what I was talking about and instead I'm going to do the model and then we'll talk about these results at the end instead okay um there's a bunch of sort of related ideas obviously I'm building on sort of enormous literatures on hold up which is a very old idea um on two-sided Market competition which is also uh a less old idea but still a very rich literature um democratization as a commitment device which is an important idea in the political economy literature um and then locking in platform governance and I know that this paper was presented I think here last year so just to give you a little bit of a sense of what's going to be different here

between these two papers this paper has a a lovely kind of fully Dynamic model of a platform economy where governance is a possible solution to a lockin problem but it's going to blackbox lot of the underlying economics so there's going to be no endogenous investment in quality and there's not going to be price setting in the like like the value of the of the of what's going on on the platform is going to be left as a black box to to further this sort of studying of the Dynamics of governance which I think is really nice and smart approach I'm going to do kind of the opposite so I'm going to do a two period model instead of a fully Dynamic model although I think you'll see uh it would be straightforward to extend what I'm going to do to kind of a a a markovian

uh uh fully Dynamic model but I'm going to do a two period model but I'm gonna I'm gonna open up that blackbox so I'm gonna I'm gonna actually model pricing and investment decisions um to try and get at this holdup idea uh and the so this investment by the Creator or investment by the Creator okay so you're sort of micr founding the response of the user exct exactly um and then I'm also at the end I'm going to ask about um about investment in in governance so so I'm going to give you results about willingness to pay for governance and then I'm going to do if I get to it at the very end I'll talk about suppose that that choosing to Implement governance or not was an endogenous decision inside this Market between these two platforms that are competing with each other what would happen all

right so here's the model I'm going to have two x identical platforms I'm just going to call A and B they're completely identical okay they're symmetric to each other I'm going to have one Creator I'm going to have a unit mass of users each with with per period unit demand okay so they want to buy one unit of whatever it is the creator is s selling uh each unit or they may they may want to depending on the price but they have demand for one unit um I'm going to name them I and they're going to have a type Lambda I which is going to be they're going to be heterogeneous and for now everything's going to be uniform although uh some of the close form Solutions depend on that but you could obviously nothing nothing nothing particularly important about the uniformity here's what happens in the

game when there's no governance okay um so at the beginning of the game platforms Set uh first first period fees these are going to be fees for now for now I'm not going to micr found this although we know from the ré and terroll uh results how to micr found this I'm going to have them only price one side of the market okay so they're not going to charge users um in fact if they could the platforms here might want to sometimes would want to subsidize users I'm going to have them only charge um only only only price one side of this market so they're going to charge fee transaction fees to the Creator for every transaction okay and we're just going to call those fees cleverly fee um so each platform is going to State a first period fee the Creator is going to

choose which platform to join they're not going to I'm not going to allow multihoming for now so they're going to join only one platform um they're going to choose whether the product the quality of their product okay and that's going to be a binary Choice it can either be sort of the Baseline quality the default quality or it can be high quality okay uh do getting high quality is going to be costly and um they're going to choose the price that they charge for the product on on the platform that they join and they're not going to be able to price discriminate across users so they're going to set a uniform price for the product users then observe all this they decide whether to whether to uh purchase the product on the platform or not okay so the users are obviously only going to join because I only have one Creator in

this model they're only going to join one platform they're going to join the one where if they're going to join it all they're going to join the platform where the product is available um if we had more creators obviously we could get a richer model of of users moving across platforms Ethan this might be kind of a dumb question I'm just trying to understand where this is starting from like so before period one neither platform has any users is that yeah I mean you can think of it that way or you can think of their like you can think of this is the marginal Creator and and users making decisions but what is important what is going on here um that's relevant is that users care like this Creator is not small relative to this economy right users care in their platforming choices um about whether or

not this user this Creator's product is available so in a world in which um in a world in which these platforms were already enormous users were already places um and this product on the margin right like there was a you know Infinity of creators and this product didn't really matter decisions would be different because there would be no change in demand in response so what's important here like one Creator is not important what's important is that the creators are not small relative to the economy they're doing something that is consequential to users users are willing to go to a platform to get their product yeah I'm just it's interesting I'm just trying to think through like there's a it seems like there's a big

difference between the discussion of like glomming governance onto a platform that's already huge like expost because people have like learned over time about this holdup problem versus like a new platform like a new let's say web 3 platform that doesn't have any users yet and it seems like there's a big yeah that's interesting I mean there's not it's hard to capture that Difference game theoretically because game theoretic agents are so smart um oh I me think but I see what you're saying because it's like I feel like when I talk to people starting platforms now they're off even if it's web 3 and even if governance is built in they often still have this thesis that like the governance itself won't pay off for

a long time like no one's going to shop platforms based on governance which me like a behavioral claim yeah so no one's going to shop governance based on some one's going to shop platforms based well okay so yeah interesting users are not going to shop platforms based on governance here creators users are simply going to go to the platform with the best product at the best price right but which platform gets the best product at the best price will be a a function of equilibrium Decisions by creators and creators are thinking about tomorrow like that's what's important is that creators here are anticipating the holdup problem and they know that if they build a really good product today and it's and that that product doesn't work as well on the

competitor platform tomorrow they're going to lose some of their rents um and therefore their decisions about where to go and how good a product to offer do depend on governance um yeah so that's what's key period two is exactly the same in the sense that platforms Set uh second period fees creators can choose which platform to go to in which price users can decide whether to purchase the only difference is that the quality decision only happens once okay so they it's not so if you're going to invest in quality has to be at the beginning like this is like the end uh if we had a fully Dynamic model we could we could keep allowing quality Investments that would be fine but when you're doing a two period model

you want a sense of the beginning and a sense of the beginning or or early and late and so late it's too late to create the quality okay so you can create quality at the beginning um so the Creator can switch from period one to period the Creator can switch platforms from period one to period two and as I'll show you in a moment what lack of interoperability is going to mean is if you invested in quality you lose some of of that investment okay so what does inter operability mean quality investment is platform specific if you just offer the Baseline product we're going to assume those are interoperable across the two platforms so this is like in the sense of like in the past we've already had this idea or whatever uh if I buy quality um on platform a and I switch to platform B it gets degraded by

an amount Alpha okay so Alpha is just a parameter of this model which measures the amount of interoperability between these two platforms right so if Alpha equal 1 over Theta then when I moved I would go from quality Theta all the way down to Quality 1 that would be complete lack of interoperability right it would be like I was offering the Baseline product if if Alpha equal one that would be perfect interoperability and in between those two we have varying degrees of interoperability that make sense I understand this is a parameter but in principle that like the uh a competing platform might want to build in operability yeah and so well so I okay so here's what I would say I've thought about this a fair bit for this model I think that it's right to think in most settings that interoperability is a two-way street um and that you

can't unilaterally create right I'm going to show you what level of interoperability the platforms would choose in equilibrium um and it's not complete interoperability uh and I'm going to show you how it changes with governance okay payoffs are I think completely straightforward so users value quality they pay prices um and then this Lambda that's the user's type they're just heterogeneous in how much they value this product okay so the higher Lambda is the more I value the product whether it's right so Q can be one or Theta whether it's one or Theta I still value the product more if I'm a higher type but I value it more more um right there's compliments here if I don't

purchas I just get a baseline payoff of zero what do creators care about um creators care about how much money they make so n is the measure of users who purchase okay p is the price they charge per transaction Fe is the the fee they charge for transaction right so this is the this is the net revenue per transaction this is the total revenue and then if they invested in quality they have to bear a cost C and C is positive does that make sense uh and platforms also just care about how much money they make right so that's the number of transactions that's the fee for a transaction and assume that that that that the platforms can operate uh at at at zero marginal cost um and I'm going to have players discount the

future by Delta this is great out because in fact this doesn't matter at all uh they could not discount the future and everything would be the same except that all the formulas would be nicer so it should probably be that way um okay governance I'm going to introduce later the possibility of governance where governance as I as as I said in my answer to Scott I think um is is just a commitment device so so what I'm going to think of as the governance solution here since I only have one Creator is the platform can hand the creator of veto right and say like I'm going to give you the a veto which says that if I propose a change to business terms which here just means fee um you can if you've already joined my

platform so you're already a member of my platform and I propose changing the government the the fee structure you can say no if you want to in practice what this is going to mean is I can lower my fees but I can't raise them okay I'm going to focus in almost probably all of this talk on one-sided governance that is only one of these platforms has a governance solution not both of them at the end if we get to it um I will show you that in equilibrium that will be true at least in pure strategies that is if this is actually I think some similar to a conversation we had two days ago um if both platforms can can have this governance solution they're going to compete away with each other all of their profits um and

so so two-sided platform uh two-sided governance is is bad for these platforms here there isn't going to be a pure strategy equilibrium where they have it there is however a mixed strategy equilibrium where they both buy buy governance uh with positive probability I'll show you that too if we get to it but um but if if governance is costly and endogenous in pure strategies we're going to have at most one platform implementing a governance solution so I'm going to focus on that case for now and then at the very end I'll show you those results why why don't the creators ever want the fee to go up is there never a case where they think like the platform needs more money to actually so this is what I was alluding to when I was answering uh SC question from online which is I think if you you could

imagine enriching the model such that um like an event Could Happen such that the business needs to do so right needs to change policies in some way to save the platform or to keep the platform viable that the creators would all agree to and that that thing maybe is not contractable and so it's hard to write it into code or it's hard to write it into a contract but you can do it with governance and these guys would be willing to let them do that in the extraordinary event but not be willing to let them raise the fees just in the normal order of business business and in that world governance as a solution would dominate things like Contracting or writing into the code that's not here but that's in the back of my mind as like kind of the obvious answer to well why don't they just write the contract

or maybe there's uncontracted events that governance allow some flexibility um but while still solving the holdup problem I should say as a footnote I think that's a less understood value of governance in these web three communities a lot of the time right a lot of the there's an easy sort of newerk well we can just write immutable code but then of course something happens and you're like oh well maybe that shouldn't have been quite as immutable as it was you know so I I very much like that answer that makes a lot of sense thanks okay so let's talk about what equilibrium look like so this is a simple game to solve right despite the fact that a month ago I solved it wrong um so this complete information game so Ethan just clear so with or without

governments the question is just in Period two can they make the fees bigger than in Period one one or not like is that part of the strategy space that's the key what governance says is they all they can always propose to do so governance says that the Creator can veto it but that that's not explicit in the model or it is that is explicit in the in there I want you to think I guess of there being two games Okay a game without governance where they just can and a game with governance where they can if and only if Al so the Creator says yes and I'm going to compare those two games so you're not my first thought was you just fast forward to the upshot and

just like Define the strategy space differently for the platform where the strategy space is in Period two in one case you can set it whatever you want in the other case you can set it to be yeah no that's not that's not what you're doing so this going to be like those two things will end up being having equivalent equilibria I'm going to I'm going to tack it onto the extensive form okay I'm going to give the Creator uh an actual explicit move uh which is say yes or no but but they're equivalent um but it's a result in it's obvious but a result that they're equivalent right like and in the world in which you could have these like these shocks they no longer would be equivalent okay so we're just going to

look at subgame Perfection of this game so what does that look like so uh we're just going to work backwards so at the end of the game right so what does user demand look like in Period 2 given some quality Choice Q given some Price p on whichever platform um the the good is offered on well you're going to buy the good as long as it's better than your outside option and that's going to give us Demand right and it's going to be uh you know it's going because of uniformity it's going to have the obvious functional form but again there's nothing important about that what's important is it's decreasing in price and increasing in quality okay and so then we're going to be able to say well what are the what are going to be the equilibrium prices right so here is the objective function of the platform

this is is the per transaction um this is the per transaction Revenue this is the size of Demand right because these guys are important in this economy they're not small relative to this economy uh changing the price changing the quality is going to change demand and they're thinking about that they're going to set prices to uh right the benefit of increasing the price is you get more per transaction the downside of increasing the prices fewer people buy uh you're going to get an explicit for solution for prices which is going to depend on the quality of the good as well as on the Fe is being charged right so there's pass through here when I when I get um when the when the platform increases the fee the platform passes some of that fee on to users right they don't pay all of it themselves um in the

form of increasing the price so a bad thing about increasing the fee from the platform's perspective is that it gets passed on to users and therefore that decreases demand um so are the lambda's uniform the lambdas are uniform okay um and so we can just just write down given the optimal pricing um what the Creator's profits look like in in Period 2 if they're on platform J given the fees right and that's just a number we can a formula we can use what platform are we going to end up on on period two given these these Market decisions right so given given the demand and given the pricing well I'm just going to right A and B are symmetric so I'm just going to think about the thought experiment suppose they started in Period one on platform a

and you should just remember that the right the continuation game they started on platform B is exactly the same you just change all the letters right so I'm always going to index on suppose they joined a um so if they had started on a so if they came into Period 2 working on platform a and they hadn't invested in quality so they they were offering the low product quality product right the platform the the Creator wants to join platform a if they're going to make more money and we've figured out how much money they're going to make here it's their quality minus the fee over this thing right so if the quality is one uh we've got these ones the Q becomes one and it's just this comparison of the two fees they're going to go to whichever platform has the lower fee this is

obvious right there's no interoperability problem in a world where you're offering low quality these platforms are completely identical you go to whoever charges you the lower fee therefore in equilibrium the fleet they compete each other down to zero both platforms have to charge zero okay and we're just going to assume then as you one always does in sub game Perfection that they're going to stick with they're going to stick with a does that make sense to everybody so in a world with low quality now interoperability doesn't matter platforms have to just compete themselves all the way down to not making any money so one thing that you could do as a Creator is invest in on interoperability and that would that would it could be costly to you it's

like equivalent to like having and so I can see why a Creator wouldn't want to do it but one thing that a Creator investing in interoperability would do is create externalities if there were multiple creators because they would be they would become like an INF marginal like uh creator that could that would potentially create if the platforms can't price discriminate across the Creator so like one thing that like like uh creators might want to do is is subsidize Mr Beast to be very flexible about his platform choice because that will induce pricing competition between that's interesting yeah so it's an externality like I only have one Creator but yeah that's interesting things are different in a world in which ex Ane the Creator invested in quality because now

there is lock in right without quality there's no lock in by assumption um and that's not important as long as there's less lock in it's fine um there's locking because of this interoperability now binds yeah and so now the comparison is different across the two platforms because if I switch from A to B right I degrade some of of what I'm offering and that means that even at Price zero right even if B offers no fees at all a can beat them and they can beat them with a positive postive fee so a is going to win here and they're going to win at exactly the indifferent point this is the usual result in a holdup problem that a that the platform here is going to be able to extract from the Creator all of the excess Surplus between what

they're making here and their outside option right and that's just because this is demand bargaining um okay so a is going to win here if there's been an investment in equality and they're going to win at positive profits right they're going to be able to charge fees because B knows they're going to lose some of the quality they've invested in I'm sorry the Creator knows they're going to lose some of the quality they've invested in if they move to B and that's going to come at a cost in terms of profits in terms of demand in terms of the prices they can charge okay so we can just summarize period two um none of this is important other than to see that what's going on here is um I mean it's all important for solving the model but none of it's important to know except for this which

is to see that the Creator is having their second period rent eroded by Alpha right that in the end the profit that the Creator makes in the second period don't don't just right they do respond to the Quality investment but in a way that's degraded by the lack of interoperability precisely because the platform is extracting those rents from them okay so what happens in Period one so in Period one everybody's identical right the two platforms are identical which means that they're going to have to compete each other down to zero right they both want there's positive profits to be made for the platforms if they can get if they can get the the Creator to join them uh in

the first period they're identical so if one of them offered a lower price than the other they would win for sure and so they're both going to have to charge zero I apologize this is a typo this should say f1a fee 1B everybody's going all the fees in the first period have to be zero okay and so the Creator says well what are my choices my choices are I can invest in quality or I can not invest in quality if I invest in quality right I'm going to get a payoff today I'm gonna get that same payoff tomorrow because if I invest in quality fees will stay zero tomorrow right this is just discounted by Delta doesn't matter if instead I invest in quality today I'm going to get the fee of zero today and so I realize all the value of that

quality investment today but tomorrow I'm going to lose some of it and I'm going to lose some of it because of this thing we saw here which is that I get lock in and some of my some of my Surplus gets eroded okay and I have to also I have to pay the cost of buying of buying the quality and so when I'm going to do it I'm going to do it if the costs are low enough relative to um the magnitude of the quality uh value and the size of the lock in effect that's obvious right so if costs are low enough where low enough depends on Theta and on Alpha okay and so what does this tell us this tells us that we're going to get in this world without governance but with lock in uh and with some interoperability we're going to get

quality investment if the quality investment is sufficiently valuable that is if Theta is above some Theta hat okay which depends on the other parameters um if interoperability is sufficiently high so that there's little enough lock in okay that's only that you need some some minimal so if C is too high if costs are too high you'll never get it but um but if interoperability is sufficiently high and as we've already seen if cost are sufficiently low okay so what's true greater interoperability right so this this is where we're starting is interoperability is the source of locking here so if we have greater ER operability we're going to get more investment in quality because there's going to be less worry about lock in and so more willingness to invest we're going to get higher total

Surplus Creator Surplus and user Surplus but for the platform Surplus is non-monotone okay platforms and and and here here's the the picture Okay so the xaxis here is Alpha that is the amount of interoperability okay and these pictures are every player's Surplus and what you can see is everybody's Surplus is in increasing in Alpha some of them discontinuously at the point of right so Alpha hat is the point at which we first invest in quality um but everybody's everybody is is weekly increasing in Alpha here they're flat um and and sometimes there's a for some players there's a jump depending on who gets passed through to but not for platforms platforms are flat before the investment

they jump up discontinuously at the investment in quality because it is only at the investment in quality that they get the lock in effect right and start getting to extract some rents but so there's this discontinuous jump up because oh here now I already I get to extract some rents and then as Alpha gets higher the amount of rents I can extract gets smaller and smaller and smaller why because there's less and less lock in right and so in that second period my bargaining power is weaker and weaker and weaker okay so for the platform the optimal level of interoperability is not zero nor is it complete it's in between it's the it's where you get just enough interoperability to incentivize investment in quality while holding competition to a minimum right as Alpha

gets bigger in some sense a is facing more and more competition from B in the second period and they don't like that so the the optimal level of interoperability that the platform would set is is well short of completed interoperability but it's not none okay and if there was some lock and effect that came along with um with low quality also not just with high quality what would happen is that this thing would be increasing a little bit but there'd still be this discontinuity in this in this non so platforms want in a world without governance platforms want some interoperability they don't want complete interoperability everybody else wants complete interoperability it's good for everybody right it increases creators bargaining power increases which is good for them increases

creators incentives to invest in quality which is good for users in a world with like total interoperability are there still incentives for new platforms to AR r no not here so Platforms in a world with complete oper interoperability platforms are operating at zero profits sorry so no now that's a you know that's a frictionless model where um where where competition is getting us down to zero profit conditions but okay so now suppose we introduce some governance okay so suppose that one of the platforms but not both have the ability to hand to the Creator a veto that says

I can't change the fees without you're saying yes okay so I'm going to think about platform a having having that governance uh solution but platform B not having it um and later I'll talk about endogenous investment in governance in equilibrium as as as Tim was saying uh we could build this straight in but here I'm going to have it be part of the extensive form in equilibrium that's going to impose the constraint that that a could reduce its fees but it can't increase them right its strategy space still includes increased fees but in equilibrium those fees will always be V an increase in fees will be vetoed and so they won't ever do it so it's like a four stage game now or what is it going to be uh I mean the extensive form is it's a long extensive form um but all I mean I think

the only thing to think about if we think about the time if you just think about the timeline of the game um what happens is in the second period when I impose my fees after that there is a move by the creat which is yes or no and if they say no fees stay the same as as they were in Period one and if they say yes they they go to the the the changed in Period two if we had low quality we are still in a world where the two platforms are identical they have to compete down to zero if a didn't offer to revise it if a had a positive fee in the first period and didn't offer to revise it down it would lose right B would win so it's going to offer to revise down the creators will always say yes to a reduction in fees and so

everything stays the same as in the model without governance with with low quality in the second period right so everything you just saw is still true but that's not true with high quality okay so in with high quality if I start if I started on a right at high quality I still face the same the same um comparison yes and a still wants to win right so if they were charging a fee higher than the fee we found before yeah which was the the minimal fee the the maximum fee that allowed you to win when B was charging zero you got to revise down but if you were charging a fee below this you're not going to revise

up yeah so they'll be so you're going to end up at this fee in the second period if you were charging more than it and you'll end up you'll stick with the fee you were going to charge before if you were charging less and on the equilibrium path this is where we're going to end up okay on on the equilibrium path what's going to happen is that a is going to offer a lower fee in the first period right and because they're committed to it in the second period right they'll be able to make that smaller fee but for both periods instead of a higher fee a zero in one period and a higher field so this is where we're going to end up on the equilibrium path so now we just want to ask period two is quite quite similar to the to the previous model but period one

is quite different because now when I ask should I join a or b a is making this commitment right at least a partial Commitment if they offer a fee below this right they won't revise it in the second period so with the question we want to ask is can a commit to a fee that's not zero in the first period and induce quality investment by The Firm by the Creator okay so a faces four constraints the creator has to prefer a at that price to B when they invest in high quality in the first period and that's just going to give us a condition right so if I if I do a I'm going to make this payoff and I'm going to make it now both periods because they're not going to

revise that fee right and I'm going to bear the cost if instead I do B well in this period they're going to charge zero in the first period and I'm going to make that same payoff I made before in the second period they're going to behave just like in our old model they're going to have Market power right and they're going to erode some of those rents and so I'm just going to get a a I'm going to get a condition that says like a A can charge a positive fee it has to be below some number which depends on the amount of lock in um but that number is always positive okay so so this is a a real constraint on the fee you can charge but it's always satisfiable so a can always set a fee in the first period that if the if they

invest in quality they're going to prefer a to B and the reason of course is B doesn't have the commitment power so even if B offers zero today we know they're going to ratchet up tomorrow right I can offer something in between for both periods that they like better there's another constraint which is once we get to Period 2 at high quality do they want to switch to B that's just another constraint and the answer is that never binds switching to be in Period two uh is strictly less desirable than switching in Period one so this constraint implies this constraint okay third constraint does when I've invested in quality I have another option if I'm thinking about investing in quality and going to a I could also not invest in quality and stick with B right so that's

another constraint I got to think about I could just live with B right in both periods um that's going to give us constraint that's going to depend on the costs of investment and we're going to get an upper bound on the cost and importantly the upper bound that we're going to get on the cost is higher than the upper bound we're going to get on the costs in the no governance in the no governance World investing in governance is more attractive now right because I don't have because I'm not getting my rent eroded in the second period and finally um you have to prefer to invest in quality um rather than not invest in quality and that's going to again give us a not binding constraint if you're going to not invest in quality go to B Because B can't commit and they're going to have they're going to give you all

the rents whereas a is already committed to this fee they're not going to lower it for you tomorrow if you if you didn't invest in quality so if you're going to not invest in quality go with the guy who didn't commit to sticking with a high fee okay so we're just get these four constraints which boil down to two constraints you got to prefer to invest in quality and stick with A over B and you got to prefer to invest in quality over not invest in quality that's going to give us two constraints one of which is always satisfiable it's a real constraint but it's always satisfiable the other of which is satisfiable if and only if um the costs aren't too high but how high they have to be is a weakening of the constraints relative to the no governance world we could also write

that in terms of of theta the quality Advantage um but what I want you to see here is it is easier to induce quality investment in a world where you have governance than it is in a world where you not because you don't because you can commit not to take all the rents and of course in equilibrium you're going to walk right up to the constraint line and the the um the Creator is going to be exactly indifferent between these two but you're going to get quality more often and that's going to be good for the platform and good for the users even if it leaves the Creator indifferent yeah a nice feature of this too is if you had multiple platforms with governance or with commitment power and just they're exogenously endowed with it then then they're going to be able to the platforms are going to beat each

other up on price all the way down to zero you'll get perfect quality Creator gets all the from their quality kind of kind competition kind of gets you to the right place if you have multiple multiple platforms with that commitment which is why I think it's a feature not a bug I think it's a good yeah I think that I mean that seems it seems right yeah um and it's why of course in pure strategies if governance is costly yeah that's yeah right that that sounds or they no only one of them will buy it yeah but if commitment power is cheap enough then then you can get I see the equilibrium point but then that that seems like good econ abely the platforms aren't really doing anything useful in this model other than being the Conduit on the way to the

Creator they I think that's pretty that's true quite generally in they shouldn't get paid yeah EX in this model yeah another like interesting that's not that's not here but like with multiple cre like as the number of creators grow grows you might think that the governance cost throw because now you have like a collective action problem among the of course here that's not here all the creators would agree on everything like in this like they all just say like yeah don't raise fees right um you could imagine a richer model where that's not true but here actually you could just delegate one person to vote and it'd be fine although you might you might still have the collective action problem that

nobody wants to be responsible for the delegation or something if you need like a quorum of a certain number of votes or delegation approvals or something yes and I mean that's interesting like if if voting itself is a little bit costly now maybe like there's actually governance doesn't work quite as well as this and there is still some room to exploit lack of interop of loot to exploit lock and that's interesting okay so what can we say uh when we compare the world's with one-sided governance to no governance um what's true is that there is more quality with governance okay so on the x-axis here we have the value of the investment that's Theta and on the y- AIS here we have the amount of interoperability and this is just showing you right the worlds in which we're comparing the no governance and

governance model uh everywhere here we're getting govern we're getting quality if we have governance but there's this this this I don't know what that shape is uh where we get it if and only if we have governance right so governance slackens the constraint for getting quality by allowing us to um allowing uh a space where there's valuable there's valuable um uh investment to be made in terms of total Surplus um but if we couldn't commit they the the amount we would extract in the second period is so big that they that the Creator wouldn't wouldn't be willing to invest and we can hold that back with the commitment device um and incentivize them to okay so we're getting more quality Ethan this is just s of an observation I

don't think it's like relevant for anything in particular but just I'm just thinking about applying this to web 3 and so one thing that seems interesting web 3 is like I think people generally think and the many more expert people than me in the room can correct me on this but like I think we think interoperability is higher he's usually like more baked in in in web 3 and there's also more governance but one of the thing that strikes me looking at the governance in web3 is like sometimes the governance is determining the level of interoperability like there are votes over decisions about like bridging and stuff so it seems like there's some interesting potential interaction between the two and I'll show you I mean there is an interaction here I'll show you what it is but it's different than that um so this is just to give you a

sense of okay so we saw what the willingness to pay or what the platform's preferences induced preferences right indirect preferences over interoperability were and let me show you now what those same what those that willingness to pay or induce preferences are for governance okay so this is what the platform's rents look like in a world with one-sided governance this is obviously the winning platform this is the platform that that gets governance and gets the market um what are their what do their payoffs look like in a world without governance and with governance okay so down here where Theta is really small you're not getting investment in quality no matter what and both platforms the platforms making zero profits no matter what in this space you are getting where where now this is we on the x-axis here I have the the qual the value of the investment

so in this moderate moderate value to the investment right you get you get quality investment if and only if right if and only if you get if you you have governance and so now there's this jump up in Surplus to the platform it's now making profits from the quality investment its ability to extract that Surplus from the Creator when it has governance but it can't do it without governance because it can't incentivize uh quality investment and then when we get even higher now we're getting quality investment no matter what still the platform is is um extracting a little bit of extra rent through this commitment power because it can it can it can U right it can bargain the it can

bargain the Creator all the way up to indifference right leaving the Creator with absolutely no um right all the way up to indifference so the willingness to pay right looks non-monotone it's a zero at very low levels of the importance of the of the quality investment it's increasing and increasing rapidly right uh when when you can get quality investment if and only if you have government and then it jumps back down and then continues to increase once you get both so there's a little bits of rents to be made even if at in a world where every you're going to get quality investment no matter what but the big the big rents are to be had from governance in a world in which governance is what allows you to get people to invest in quality we can do the same thing and this is to come to

Andy's point in terms of if we change the x-axis here to be in terms of interoperability so now for inter as long as C is not above that that that upper Bound for any level of interoperability with commitment with governance you can get quality investment why because interoperability no longer really matters it matters for the scale of the price you have to pay but is no lockin problem once you have governance right interoperability was blocking in lack of interoperability was blocking investment in governance precisely because it created lockin but once you have governance you don't have to worry about lockin so you are always for any Alpha here as long as C is below that upper Bound for any Alpha you're getting you're getting quality

investment with governance but for low for interoperability too low you are not getting it um without governance and then there's a jump up for one of them but again there's still a little bit of space and so so if we think about uh willingness to pay right at very low levels of interoperability governance is super valuable at very high levels of at higher levels of interoperability it's less valuable these things are kind of substitutes for each other um I guess the only other thing I want you to see here is that when you have governance interoperability is all bad from the platform's perspective from everybody else's perspective it's good but from the platform's persp it's all bad right you want low interoperability if you

have governance because interoperability just weakens your bargaining power there's no interoperability no longer matters for your ability to get quality it just as a reduction in your in your operting P so so whereas without governance you wanted moderate interoperability with governance you want none so platform's willingness to pay for governance is highest when Innovation occurs if and only if there's governance and that's for moderate values of the innovation for moderate cost of the Innovation and for low levels of interoperability that's when you want governance we can also I mean we can talk about sort of all the pass through and the incidents of governance um so you know it it does lots of things right it affects the fees that get charged I

won't we're running short on time so I won't belabor these points but it affects everything right it affects the fees that get charged it affects the prices that gets charged it affects the level of demand um and so I guess if you just want like the quick Tak homes uh the average fee is higher without governance um it when quality happens within if and only if uh governance the average price is higher without governance in those circumstances and the average demand is lower without governance you and who who benefits the users benefit uh from the greater quality investment the creators are indifferent because they're always getting bid all the right right up to uh right up to indifference it's not that they're making zero profits right they are because they have some Market power

relative to the users so they're making profits but um but they're getting bid to indifference by the platforms and the platforms uh benefit the platform benefits from governance uh and benefits the most when uh governance is necessary for creating quality investment and it's good for total Surplus so the last thing I want I want to see before I talk very briefly about endogenous governance is thinking about the social value of governance relative to the platform's willingness to pay for governance okay so we can write down the social value of governance that is just this total Surplus calculation okay so why is governance to valuable it's valuable because it gets uh it gets quality investment quality investment is good for everybody because it right

quality investment as long as C is not big quality investment just creates value that gets passed through some of it to users some of it to the platform um so this is the social value of governance we can compare the social value of governance to the platform's willingness to pay for governance okay and we'll see that it is that that they the platform willingness to pay for governance is strictly below the social value of governance only in right if and only if we're in this world where governance is necessary for Quality investment so this is kind of bad news in the sense that um platforms are less willing than the social planner to buy governance if and only

if governance is important yeah so governance is important when it's necessary to incentivize quality investment and that's exactly when the platforms are less willing than the social planner to do so and that's because there's some pass through here the platforms don't capture all of the social value okay so I'll just say one word about we can think about I've studied one-sided governance you can imagine two-sided governance right I've studied no governance and one-sided governance you can imagine two-sided governance we can write that model right so just think about extending this game even extending the extensive form game even one more step backwards so there's imagine there's a beginning of the game where each of these platforms can buy a governance solution right at some

positive cost K and now we know what's going to happen in each of the possible continuation games right if if we get one-sided governance we get the outcome we just studied which I'm going to call PG the profits for the platform right that's you know this if we get no governance by anybody we get the thing I studied at the very beginning of this talk right the no governance model and if we get two governance it's just as as we were talking about it's kind of obvious if you just think about this model for a second the platforms at the beginning of the game have to compete each other down to zero profit right if I if I have if we both have governance and I offer a slightly lower price than you of course they come to me and so therefore we're going to compete all the way down to

offering fees of zero okay so in a world where we both accidentally buy governance we both make profits of zero and we bear cost of K that's very bad for us and so that's just this simple game right we can each buy governance or not and um if I buy governance and you buy governance we both make negative K because we compete each other down to zero profits and we have to pay the cost of the governance solution if neither of us do we get the model I solved at the beginning of the G of the talk and if only one of us does we get the model that I solved at the be in the middle of the talk minus the cost of governance right and this game has three equilibria right at least if K isn't too big if K is really big this is the only

equilibrium right but if K is smaller then there's pure strategy equilibria where one of us buys governance and the other doesn't there is not a pure strategy equilibrium where we both buy governance because if I know you're going to buy governance I'd rather not and make zero instead of minus K but there's also a mixed strategy equilibrium right there's a mixed strategy equilibrium where we both randomize we're both indifferent the the the randomization probabilities look like our willingness to pay for govern a ratio of our willingness to pay for governance minus the cost of governance over the willingness to pay for governance that's what the mix strategy in this game looks like and in that mix strategy equilibrium everything's in the support right we can get two-sided governments one-sided governance no governance we can think about the comparative Statics that's what that

probability looks like as I mean this this is just chicken straight up yeah exactly this is just a game of chicken um and so we can also right we also know lots of things because it's just the game of chicken about purifying this and what this would look like um if we say purified it by making K random um so we can we can say things about that we know how to study that game but um I don't I don't know whether I should believe the mix strategy equilibrium of this game or not um but you know one could study the global game version of this and and say comparative static things about the cut off um and they have the natural comparative Statics um so I'll I'll leave it at that and just conclude with saying to sum up sort of the take-homes I think of the model are both

interoperability and governance are ways of addressing lockin um platforms at least in this setting prefer one-sided governance over interoperability as a solution if platforms are going to do interoperability they want to do it only partially right whereas everybody else wants complete interoperability um governance is good at least in a world of partial interoperability governance is good for users and platforms and total Surplus but leaves creators and different platform's willingness to pay for governance is below the social value of governance in particular in the range of parameter Valu is where governance actually matters for outcomes that is whenever governance would Inspire

quality investment and the lack of governance would not in equilibrium that's when platforms are under willing to invest in governance and then I've got lots of questions about what what possible next steps might be you can think about a model with lots of creators you can think about a model with multihoming you can think about a more properly Dynamic model I think that's actually straightforward to do here you can think about other sources of lock in like search costs um or network effects open to other ideas that's what I got so Ethan I'm sure there'll be questions from the audience but maybe just before you get to those questions it might be helpful now to sort of connect this to web 3 I think some of the connections are pretty obvious like the these questions about governance and lock in are pretty core to people's you know interests in web 3

but just might be helpful to hear from you like what parts of this you've looked at web 3 quite a bit what part where are you thinking about Connections to web 3 here more explicitly yeah so I mean and Andy's like being like the the straight man or something so Andy and I have written a bit about um holdup problems in platforms and think that I I do actually believe in a way that's somewhat hard to convince uh them of but I believe that web 2 platforms are suffering substantially from the holdup problem I think that their creators do know that their profits are getting eroded and are under investing and that they uh need solutions to that problem and I think that um you know governance is Trivial here in the sense that

there's only one Creator but in a world with lots of creators like buying a governance solution is not trivial for lots of reasons it's hard to get participation but also it's like hard to implement and I think that um the kind of whole Spirit of thinking about well what are robust systems of decentralized governance online is like that's a first order question for implementing something like a solution so I think from my perspective like the connection to web 3 is like there's lots of innovative work going on I'm thinking about how might we Implement secure robust governance Solutions and I think this is Maybe from my perspective at least or my my my view an underappreciated use case for online governance I think like just lots of

lots of online life has platformy like things and has lockin and there's lots of opportunity to improve business outcomes by with a good governance solution and I think one has to First successfully make the case to some sort of e-commerce company that in fact the holdup problem is real um in order to convince something that they should think about governance but but it seems obvious to me that the I mean you just like you watch creators for these platforms talk about it themselves and they they are all now aware I they may not have been when if like they they were not perfect game theoretic agents X anti and maybe it didn't occur to them that the App Store was going to increase you know inapp purchase fees or that Etsy was going to start taking all of

their rents but they know so now and I think like if you think about who's building cool apps now or who's building whatever like they are thinking about which platforms are more or less likely to let them keep some profits in the future once they become dominant I think Network effects and interoperability and all these things mean lock in is forever uh and so I think like there's just going to be na there's there should be natural demand for governance solutions that comes out of that set of problems the I so you had a different distribution non-uniform Distribution on lambdas everything here I think would still hold qualitatively I I mean so um it couldn't be an arbitrary distribution but like the mlrp or

something like that would be enough right you got you you you you don't want like um I think although I you know I'm not a price theorist so my intuitions about this come from thinking like a game theoretically and I think like price theorists would have much more to say about like what makes for a well- behaved demand function um so I'll have to have more to say about that but it's completely obvious to me that uh that in a world where um where I had the mlrp in the distribution then I'm just get right I'm just going to get like a nice decreasing function of of p and an increasing function of Q in exactly the way I want for my demand function uh if the distribution had weird humps and bumps and stuff maybe maybe not I don't know I just kind of a request as a next

step is I I think thinking through um how the analysis evolves if the platform um creates value in a more meaningful way um so so as a something where the platform and the Creator together you know where the platform does something that creates value that it's more than just a a pass through or conduit because in in your model the platform is really just a conduit between the customer and the Creator and in that model all you want to do is get prices beaten down to zero and actually the stock exchange work I talked about on Monday is is is I think a a good example where in that in that

market a few decades ago the exchanges had a lot of Market power that came from to trade stock you had to be together in the room together that gave them Network effects and lock in and then as the market got more electronic interoperability was then regulatorily mandated by the SEC and fees got beaten down to zero I I showed you there like 0.0000 one cents per share so they are the take the take rate in the US Stock Market is you know one 100,000th of the take rate in a lot of a lot of other uh platform markets um but as soon as you have the exchange the the platform doing more than just being a conduit and then it then it strikes me

as a much harder like a lot more economic moving pieces because then you worry Tim used the word aifi earlier in the talk you worry about a platform where governance is completely decentralized not being able to flexibly move with the time yeah it's interesting and just thinking through those tradeoffs strikes me as really interesting and promising so at least what I know of the platform Market literature the theoretical literature it all has this characteristic right like it like the I think the value in the models like the implicit value of the platform is exclusively some assumption that the that the user that the two sides of the market couldn't find each other without the platform and there's no other like

that's it like if they weren't there they would there would be no transactions and and so there's no value other than absent them um absent it the market wouldn't exist um I guess that's there's some things where that seems more or less right to me but I take it like so if I think about like what is Apple App Store what value is it adding other than matching app users to app developers I guess what it's providing some content moderation kind of function um works well with iPhones right the apple does the iPhone is a special product right that gives them some some Market power you know Uber will do like trust

and safety stuff and the ratings and the app they're do they're doing stuff that's hard to imagine a fully and and I'm very curious to talk about this with this room but like a fully decentralized web 3 thing doing some of the like oh a driver got in a crash let's investigate whose fault was so if if you were to think about this in this model is the right way to think about that that the platforms themselves should be heterogene like one platform should be better than the other is that what you have in mind I don't I honestly don't have in mind a particular way to model it me as a hard just modeling judgment will be required but just something something where you think through where there's genuinely got to be great if the platform could

commit not to raise its price but it would also be great if the platform had the flexibility to innovate circumstances change and those two things are kind of intention with each other and I don't have an Insight how to model it exactly I think I think it's super I agree I think it's super important because I think it's the strongest counterargument like I'm not saying it's correct in reality but I think it's the absolutely strongest counter if you talk to people who work at large platforms today they'll tell you like we expend enormous amounts of resources shaving like an eighth of a second off the load time on the next you know piece of content in your feed because other otherwise we'll lose that user to another platform and and we do all this fancy AI to serve you like the

stuff you really like that's what you know Tik Tok would say or something and they had to invest a ton to figure out how to do that and so forth and so and then their view is therefore like they think that's the entire game and so they think of governance as like a pointless expense that would make the service slower or less agile and that's definitely and you've heard that argument before I know and so I think that's definitely like their Baseline view I think on this model would be like the costs of governance are extremely high okay uh because it's it would prevent the platform from swiftly implementing what's necessary to remain like competitive yeah so in a I mean the model captured that in the sense that if the costs to governance are high they

won't buy governance and maybe that's the world we're in that they believe that and maybe they're wrong wrong and maybe they're right um they are if there was like a I could invest in the quality of the platform they are making some rents here um like they're not right they're not competing themselves all the way to zero um so sometimes there are decisions like that that change the algorithmic feed right so like YouTube has constantly changed like what type of videos they're going to surface and that has like distributional effects on certain types of creators like certain types of creators just went bankrupt others made tons of money as a result so there's like different there there's at least a politics between creators in their agreement it's not pricing like with

pricing they all creators kind of agree that better but like the the prank videos were bad for YouTube in general I think because they made a lot of money but I think they like created like policy risk you know they they're encouraging people to do dangerous things and so they tried to downweight them as a result all the prank video guys did poorly but all the cooking shows did well or something you know like that there was a there was a there was a winner and a loser in that change so like another definition of interop ability is like interacting with applications on other platforms like for instance in web 3 if I think of like two rollups as like two like uh ecosystems and applications on roll up are these

contents like these whatever these creators do and roll up itself is a platform now when I do interoperability I not only make it like now easier to retain quality when I move to the other side because I retain my connections with my friends on the other side but I also like increase my inherent value where I am so this might be another effect to incorporate into the analysis MH because even without moving I cannot talk to the other apps so I will be increasing my uh like quality yeah and for that you'd want at least a model where users could multi-home yeah okay

thanks um I'm wondering if you had some intuition on a model where like a new platform is joining so for Simplicity let's say you only had one platform at the start that kind of incentivizes high quality um always for the Creator and now if like a new platform is joining in and it has two basically two AES that it can make decisions on how interoperable it wants to be with the first platform and how um whether it chooses governance or not and if uh what like the room looks like there what like the best decision yeah so that's interesting so what that second platform would do and mind you this is a model without switching costs right so switching costs are going to move this around a little bit um in ways that

might be interesting might interact with governance interestingly I'm not sure um that entering platform if they could choose anything would choose zero interoperability and governance right because they can come in they can win if they have governance um and that's going to maximize their rents as long as governance isn't too expensive if governance is expensive they'd want to come in and offer well it's a little complicated so as I was saying to Pablo earlier like interoperability it seems to me is in lots of ways a two-way street um so if I come in what can I do

I can build my thing interoperable with you but if that also means you're interoperable with me which it seems like it must um then that doesn't really help me I'm still going to lose um I I think what I want to do is I want to enter and I want to if all I have is the tools I've modeled here I want to enter and I want to have governance and then I can win and I'm willing to enter because I am going to make positive rents as long as entry costs aren't too high all right thank you thank you all I really appreciate it