The political economy of networks

I’ve had this post in mind for quite a while, and never got in finished to my satisfaction, but it’s been stimulated to a significant extent by reading Clay Shirky, so I thought I’d pop it up now, somewhat half-baked, since Clay is currently visiting Crooked Timber, where I’ve crossposted this.

The biggest single question in political economy is whether and to what extent we can achieve social equality without sacrificing other goods like liberty and prosperity. Neoclassical economics (a project in which I’m a participant) begins with models which imply that, with competitive markets, all factors of production will earn their marginal product. This in turn implies that any intervention that shifts wages or returns to capital away from their marginal product must imply a loss in aggregate income (there are some complex technical issues here which I’ll skip over)

There are all sorts of problems with simple-minded applications of this result. Still, in an economy that fits the standard model of lots of competing firms, all operating in a region where constant returns to scale apply, the standard neoclassical analysis has considerable force. But the growing part of the economy centred on the Internet doesn’t fit this model at all. The Internet is a network and the economies of networks are different, in critical ways, from those of the standard neoclassical model.

There are two largely separate bodies of economic literature on network economies; although the approaches are quite different, the implications of the two are similar in important respects.

One literature is based on the idea of network externalities. In this literature, a network is simply a large collection of nodes, each connected to all the others, like the telephone network. The crucial point is that the benefit of each additional connection is not realised only by the person connected, but by all those already in the network. Hence, in economic terms, networks are characterised by externalities; that is, the productivity of one firm or household is affected by the actions of others. What this means is that market returns won’t produce an optimal network in general.

The other literature, still largely confined to theory, focuses on the stability and topological structure of small-scale networks. (I’m aware of course that economists are coming late to a field in which sociologists and others have been active for a long time). Here’s a typical example, by a couple of the leading figures in the field, Matthew Jackson and Anne van den Nouweland.

The central question is whether there exists a set of payoffs to the participants under which the network can be sustained. The answer is sometimes positive and sometimes not, but to my mind, the really striking feature of this literature is the absence of any concept remotely comparable to marginal product. There’s no general reason to suppose that a network that is stable (in the absence of intervention) is socially optimal in any sense, and hence that there is any particular reason to respect the payoffs generated by the network.

In the context of the Internet, for example, it seems pretty clear that the relationship between the social value of a contribution to the Internet and the capacity to realise a return is tenuous at best. In obvious cases, such as that of spammers, it’s negative. But even for something like Google, which is clearly beneficial in total, there is no clear relationship between contribution and return.

Taking myself as an example, I don’t think I’ve ever clicked on a Google ad, so my contribution to their returns is (or ought to be zero). On the other hand, as a general Internet user, I’ve used Google a lot and, as a blogger, I’ve got lots of readers from Google. But, on the third hand, as a site owner, I spend an inordinate amount of time dealing with comment spam aimed at promoting the Google rank of various scumbags.

So while I’m grateful to Google in many ways, I wouldn’t feel the least bit concerned about a political-economic tax and expenditure regime that limited Sergey Brin and Larry Page to millions instead of billions, and helped others in the network economy (for example, by providing low-cost access to more people).

The more that we see the network economy as the paradigmatic form of organization, the less reason there is to regard the distribution of income thrown up by the market as having any claim to be normative. The topology of a network economy is inevitably a matter of social choice, and the same applies to the associated distribution of income.

11 thoughts on “The political economy of networks

  1. So whilst the network economy might be different from the classical economy, Google is not really a good example.

    Google is a search engine. Almost anybody is free to build a new search engine and one user of Google is unlikely to suffer if some other users switch to the new alternative. Likewise if the new alternative is better at finding webpages then users have little to loose in network terms by switching over.

    Of course users of Google are not the customers. The customers are those that buy advertising. So in this sense there is some real advantage from using Goolge over a start up in terms of the scale of the audience the advertiser can access. However given that they pay per click there is little harm in actually using Google and a similar rival. In fact when I was in business for myself and I was spending around $500 per month on Google advertising (with good effect) I was also using other much smaller providers.

    Microsoft Word would in my view be a far better example. If I choose to use a word processor which does not save documents in a Microsoft Word format then there are huge disadvantages for both me and for those I attempt communicate with. However in this arena we have seen the Wordperfect monopoly broken by Microsoft and now several open source rivals attempting to challenge the Microsoft dominance. There may be network effects but it does not seem that anybody can corner the market for long as a result.

    I wouldn’t feel the least bit concerned about a political-economic tax and expenditure regime that limited Sergey Brin and Larry Page to millions instead of billions

    How would you do that in a general sense without also driving off the private venture capital that made Google possible? And without making innovators stop short of making a really huge contribution to society instead of a merely modest one? In short how the heck would you codify such a tax without killing the golden goose.

    The biggest single question in political economy is whether and to what extent we can achieve social equality without sacrificing other goods like liberty and prosperity.

    The social equality question is interesting but who decided it was important? I’d agree that legal equality is important but not really for any economic reason.

  2. Two comments (I know little about neoclassical economics as previous posts have probably demonstrated, so I’ll avoid that analysis):

    1) In the 90’s, apart from some bad rap and great metal, altavista, yahoo, and a few other smaller search engine/portal websites floated around. These I suggest were the early innovators, exploring the human-technology link for ways of exploiting it commercially. However, they failed to provide search engines that gave a prioritised list of hits which *meant* something to users.
    Google’s founders figured out the ranking by popularity (and other factors), gave hits which matched user’s tacit requirements, and the rest was history.

    Keeping ads separate from the general searching instilled the necessary sense of trust to maintain early repeat users and to build up a solid database of link weights, which reflected user’s interests. By giving companies some limited access to this information (eg by listing company ads on the side of a search hit list relevant to their business), the companies had a way of tapping into the central nodes of their target market. When users followed any of these ads, their searching of the company product list (as I understand it) implicitly added those links to the google links database, to be found by other user searches. By that mechanism, a small number of ad clicks could embed the company web site into the hit lists of future searches. The beauty of this is that the company has many different searches linked to it, over time. This was worth a lot I would guess.

    In today’s google, note that when you do a search the list often has some book extracts in it. Click on these and you are taken to the google books viewer. From there, if you follow one of the links to purchase (say Amazon, Dymocks, Google), then you have just participated in google’s revenue making.

    2) In the 90’s myself and several rather more knowledgable people did IP/FR/ATM/SDH data network analysis for Telecom/Telstra. The idea was to ensure that the network was optimal, given the statistical nature of the data transfers among individual customers, and the growth rate, and real engineering constraints.

    The big constraint of interest was that if the classic “backhoe through the Melbourne-Sydney Fibre cable” scenario happened, at least say 75% (not the figure used in the original analysis) of traffic could be rerouted around the link failure for the core links, and say 90% for edge links.
    The connection to the economic network problem is that google, in a manner of speaking, forms a supernode linking many concepts (via user search keywords and links followed) in a highly redundant fashion. While that might not look optimal, the robustness means that users will find products of interest from multiple paths – kind of analogous to the data network single point of failure constraints. Back in the day of the non-ranking search engines, it was pretty much hit-or-miss, or word-of-mouth (ie a mate gives you the URL to go to). The old engines were simply too fragile in their search results – if your search words were even a little bit off, bad luck, no link to your search target.

    PS: good work on Garnaut draft. Well done all.

  3. I think you are confusing natural monopoly with network effects in the case of Google. If Google is a network good, what is the exclusive network? Isn’t a network good something where the utility of the network increases with the number of users, not necessarily its profitibility? Why would Google’s utility change if it had less or more users? A search engine will work just as well with 100 users as with a billion. Nothing stopping it trawling the web.

    Facebook and ebay are network goods.

    If I was to replicate Facebook, I could build the best functioning social networking website in the world, but it would have no utility without users. A social networking site with one member has zero utility, but a search engine with one user has just as much utility as a seach engine with a billion users.

    The same is true of ebay. The utility of ebay comes from the number of buyers and sellers. If I build a better functioning website than ebay it will have zero utility without lots of users.

    The more people that use Google the more money Google will make and the harder it will be for competitors. But that’s the same in many normal industries and is a result of natural monopoly, not network effects. Google’s profitibility is tied to its size, but its utility is not.

  4. I don’t even think Google even qualifies as a natural monopoly. It just has a good search algorithm and a high degree of brand recognition. What are the barriers to entry other than patent law?

    I would agree that ebay is a network good. If it has more customers then it is more appealing to other potential customers. However it still has competitors such as Amazon in the second hand book sector.

    I don’t think Facebook is a network good. More advertisers does not make the product necessarily more appealing to other potential advertisers. Although there are certainly economies of scale I don’t think they make it a network good. Advertisers can take their business to Google for instance without much penalty. Clearly it’s utility for social connection increases with the number of users, however users are not customers.

    I think paypal and e-gold are network goods. The more they are used the more appealing they become.

  5. Users of Facebook are not customers, but they are deriving utility (ie benefit) from Facebook. Something given away for free can still be proviting utility. Even though most users of Facebook pay nothing for its service they still benefit from it, and the amount of benefit increases with the number of other users.

    Or put another way, from an advertisers point of view it is not a network good. But from a users point of view it is.

  6. Perhaps. However why should we worry if somebody corners the market in giving stuff away for free?

  7. p.s. I suppose that the government might feel like such a provider is encroaching on government territory. 😀

  8. re: barriers to entry

    1) For a search engine to be seen as useful, there is some minimum amount of coverage needed, and it requires some amount of server hardware and network bandwidth. That bar has steadily risen from the old days, i.e., when one could start with a handful of servers and do something useful.

    2) Google has *always* been good at designing software that not just gets useful results, but *scales huge* and does it *cost-effectively*. Google routinely deals with scalability issues far beyond those of most companies, and they’ve hired an amazing set of world-class computer scientists to keep on doing this.

    Although invisible to the casual user, it’s important, because if you grow fast and are successful, but your hardware/software architecture has a scale limit in the wrong place, or costs too much, you run into a brick wall, your performance tanks, and you’re dead.

  9. John Mashey,

    All fair points. However I don’t think any of them make Google a “natural monopoly” or a “network good” in economic terms. The issues you outline are ones that competitors like Microsoft are free to try and replicate. All you have really said is that Google has been innovative and well run.

  10. “The topology of a network economy is inevitably a matter of social choice…”

    I’m not so sure that choice has anything to do with it – unless in the sense that ‘choices’ are path dependent, interdependent, etc and basically pretty lacking in meaning.

  11. I’ve worked for a big natural monopoly, and Google isn’t one.

    But you *asked* for barriers to entry, just like my VC employers always do, and I listed a couple, for which the scalability is one, and the awesome hiring might be another, although the latter is sometimes more transient.

    Establishing barriers to entry is *not* the same as being well-run. There are 4 combinations, one of which is really profitable, one of which tends to put you out of business, and two of which are OK, for different reasons. VCs are very fond of barriers they can believe.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s