Arnold Kling  

Bundling II

PRINT
Jobs Come Marching... Social Security and Demographi...

I extended my thoughts on bundling with this essay.


What George Stigler showed is that ordinary intuition about bundling is wrong. Your intuition is that the reason that the seller engages in bundling is to force you to buy something that you do not want. However, as Stigler pointed out, if that were the case, it would be cheaper for the seller to leave out the unwanted good and just charge you for what you want. That is why grocery stores do not bundle milk with broccoli -- it's cheaper for them just to sell you the milk.

...Information goods are almost impossible to sell without bundling. As Carl Shapiro and Hal Varian pointed out in their classic book Information Rules, information goods are characterized by very low marginal cost of production and distribution, as well as by uncertainty on the part of the buyer as to the value (until you've actually obtained the good). Accordingly, for producers of information goods, such as software or music, Shapiro and Varian recommend various strategies that involve bundling.


A few years ago, Varian wrote this article on bundling.

consider a company that has substantial market power in both the word-processor and the spreadsheet markets and contemplates bundling them into an office suite.

By bundling these products and selling the bundle at an attractive price, the seller can assure that most potential customers for either product buy the bundle. This means that anyone who wants to enter the market for either word processors or spreadsheets won't have a significant market left. Here bundling has reduced the reward to potential competitors who want to enter either market.

...Bundling offers a way to extend monopoly, to increase revenue and to raise entry costs. But it may also offer enhanced functionality and lower production costs. Sorting out these benefits and costs of bundling will be particularly challenging for the courts, since the balance of benefits and costs will differ significantly case to case.


For Discussion. In my essay, I point out that

In theory, the government could intervene everywhere it finds bundling. It could stop stores from offering two-for-one specials. It could ban frequent-flyer miles, which are a form of bundling. It could force cable TV operators to offer service by-the-channel instead of as a package, or by-the-minute instead of by-the-month.

How should government decide when to interfere with bundling and when not to interfere?


Comments and Sharing





TRACKBACKS (6 to date)
TrackBack URL: http://econlog.econlib.org/mt/mt-tb.cgi/65
The author at Knowledge Problem in a related article titled MORE ON BUNDLING AT ECONLOG writes:
    Arnold Kling has more to say on bundling, including a new Tech Central Station column and good links.... [Tracked on April 5, 2004 7:52 AM]
The author at fling93 loves fishies in a related article titled Microsoft and Bundling writes:
    There’s been some discussion of bundling and Microsoft lately. Some of it is really interesting and eye-opening, but I thought I’d clear up some misconceptions myself. Warning, this is pretty long.... [Tracked on April 16, 2004 2:25 AM]
COMMENTS (14 to date)
Lawrance George Lux writes:

It would be easier than thought, but it would increase Cost for the suppliers; simply force expense accounting for each individual item, plus an estimated Profit ratio per item. This allows for both individual item and bundle Profit evaluations. Then Government would intervene where item profitability was vastly in excess of the bundle profit share. lgl

Brad Hutchings writes:

lgl... I don't think his question was an accounting one. Bundling is so prevalent that the problem would be bigger than speeding on the highway. I interpreted the question to be one of what moral justification the government would have in a particular case of bundling to step in. While I believe it doesn't ever have a legitimate case for interfering with bundling per se, politicians can find a lot of silly things to upset themselves over.

In the case of basic 70 channel cable, McCain wonders why his constituents should have to pay for the filth they don't watch and don't want to bring into their homes. A background issue is the price growth of cable service since deregulation, but since there are multiple satellite competitors, pure price regulation is probably out of the cards. Frankly, I think McCain sees unbundling as a way to reduce prices. Instead, he was told that it wouldn't be economical.

Taken to the extreme, McCain's unbundling idea might lead to cable service being billed per minute, like telephone service used to be primarily billed. With that billing structure, it's easy to argue about taxes, infrastructure recovery fees, etc. because they seem "market-based". Everyone pays in proportion to what they use. With an unlimited package price, the politics get a lot thicker around those kinds of things, as the charges seem more arbitrary and less based on individual use. Perhaps McCain is thinking along those lines. I doubt it though. I think he's just grandstanding because America had to see a pierced nipple during dinner a couple months ago.

-Brad

Brad Hutchings writes:

Oh, one more thing to add... If McCain is really about keeping filth out of homes and not (instead) concerned with cranking down the price of cable, the problem is already solved with filtering. Think V-chip. Clinton solved the problem, right?

But seriously, with the Internet, we primarily try to solve McCain's filth problem with filtering. With the telephone, we have the no-call list plus various services (Caller ID, anonymous call block, etc.) and answering machines that can be used pretty effectively to keep unwanted calls to a minimum. In both cases, we keep the entire network (theoretically) available and limit what we don't want.

Perhaps the pool of resources available on the cable network is a size where a politician like McCain sees the filth issue as something the provider should be responsible rather than something customers should deal with. If that's the psychology, then the best thing for cable to do would be to upgrade subscribers to digital cable and 500 channels. Seems to me that would be effectively done by a better bundle offer than standard cable! Ironic.

-Brad

Jeremy writes:

For me, the important point is I do NOT want to fund CNN, MSNBC or other left-of-center news organizations. I am given no option to just buy Fox News. When a provider offers this choice, I will switch to them - no US government involvement required.

fling93 writes:

Your essay made the strange comparison of Microsoft's bundling with auto manufacturers. I see no problem with bundling cable channels and bundling applications into an Office suite. The issue with bundling IE and WMP into the OS is that you are bundling a product with small market share with a product where you have a monopoly, which was not the case with auto manufacturers.

And from a technical standpoint, it is very poor programming technique to integrate applications into the OS. Programs need to be protected from each other, especially the OS. Note that object-oriented programming is based on the idea of hiding and protecting as much information as possible between modules. This reduces the number of bugs (many of which are based on bad inputs into a function or stray pointers causing memory corruption) and also, more importantly, reduces the damage that can be done by a bug by isolating its effects.

Brad Hutchings writes:

fling,

Forget the technical merits, which your IT instructor is/was wrong about in a farcically simplistic way. The world takes care of the technical merits of things. Read some of Arnold's early essays about working with Netscape server products back in the day.

The very complicated issue you bring up though is what constitutes a "product". Microsoft could argue that HTML rendering is a feature that can be leveraged by their and other developers' applications if it sits in "operating system" space with a well-defined API rather than in competitive "application space" where it is more difficult for thrid party developers to hook into. A similar argument can be made about code that plays various media formats and does DRM stuff that the content creators like. One could easily argue that the so-called competitors (Netscape, Real, etc.) in these spaces should have anticipated that they were taking on a "product" that could be best delivered as an operating system service. It may be impossible to defend market share in these kinds of software if you don't own the operating system because the market may choose the vendor it trusts and the vendor who is best poised to deliver the technology in the most flexible way.

-Brad

fling93 writes:

Forget the technical merits, which your IT instructor is/was wrong about in a farcically simplistic way.

I am a software engineer in embedded systems with over ten years of experience. I oversimplified on purpose, because I doubt the audience here is very technical. But the point was valid.

More specifically, the disadvantages to reusing code via libraries is that someone modifying that library code will forget all the different places it gets used and make a change that breaks one of them. You always have tons of different people working on different things, and most engineers are notoriously bad communicators. The less information passed between modules, the better. Which is why object-oriented is such a paradigm shift -- the main breakthrough is how much information it hides.

In fact, the only advantage to reusing code is time to market. Less code to write, thus less time to write it. Of course, that so obviously wasn't the reason they combined IE and WMP with the OS. The applications were already written separately. In fact, they spent effort to combine them.

(Okay, another reason is to conserve memory, but you can't seriously tell me that this was a priority for Microsoft, and the tradeoff is worse performance anyway.)

Microsoft could argue that HTML rendering is a feature that can be leveraged by their and other developers' applications if it sits in "operating system" space with a well-defined API rather than in competitive "application space" where it is more difficult for thrid party developers to hook into.

Actually, it is easier to define standards in the latter method, where everything is open. And note that a lot of the viruses out there actually take advantage of the buggy HTML rendering, which is an obvious reason you don't want that in the OS. It's a serious security risk when an application installation can overwrite a DLL used by the OS. Never mind the fact that DLLs make uninstallation of applications a nightmare.

None of this addresses the main point, which is that all of you are ignoring the fact that Microsoft abused its monopoly power by leveraging its monopoly to increase its market share in other markets. Which is not what auto-makers do when they bundle.

shamus writes:

Bundling should only be an issue to the extent that monopoly is an issue, and most monopolies exist through government fiat. Cable TV is a monopoly mandated by government, so it seems reasonable for government to redress the problems it created. Microsoft may control standards, but consumers actually can use other products if they wish. In theory, government should refrain from interfering with markets unless there are clear public policy grounds. In practice, government acts to maximize its power. Government promises the garden of Eden, but delivers the snake.

fling93 writes:

Bundling should only be an issue to the extent that monopoly is an issue

Yes, my main complaint is that the only issue about bundling was completely ignored by Kling's article, and I can't really see how it could have been an unintentional oversight considering the context that Microsoft has been in the news lately.

Cable TV is a monopoly mandated by government, so it seems reasonable for government to redress the problems it created. Microsoft may control standards, but consumers actually can use other products if they wish.

By that argument, Cable TV isn't a monopoly either because consumers can choose broadcast TV (nevermind satellite). In defining monopoly, more important than market share is barrier to entry (which, for Microsoft, is the fact that nobody wants to buy a PC that doesn't run MS Office) and monopolistic behavior.

Indeed, 100% market share does not define a monopoly if it cannot be defended, and indeed, having a 100% market share isn't illegal as long as you don't use it in uncompetitive ways, like bundling a product into your monopoly product to win market share, or providing different pricing to OEMs based on how much they cooperate with you in maintaining your monopoly (e.g. charging OEMs for Windows regardless of whether the PC had a different OS installed). Whether or not they also engage in monopolistic pricing is still up for debate (remember that software has extremely low marginal costs).

But everyone here already knows this stuff already. The only part that was muddy was whether there was any good reason to bundle IE and WMP into the OS, and I've already shown why it is obvious that it was actually a disadvantage, except to leverage the monopoly to gain market share in another market.

Brad Hutchings writes:

fling wrote:

In fact, the only advantage to reusing code is time to market.

OK, my pedigree first, before I tear into this... 15+ years consumer application development experience, with forrays into user and developer component software, and OS experience with Mac OS, Windows, Palm OS, BeOS, and Linux/FreeBSD. I call myself a product guy who can write code at a lot of levels. Think of me as an interior designer who can paint and lay tile with the best of them.

Your whole post indicates to me that you come from the open source world and have a distaste for Microsoft. So your quote above doesn't surprise me, because the open source world has yet to get a grasp on what good user interface is. One important ingredient is consistency, and one way to promote it is to have shared code (usually provided and promoted as the "right way" by the OS vendor) with solid, flexible APIs. The APIs are like a contract between the OS vendor and third parties that define the platform. Some application developers have succeeded in developing good APIs that make their applications sustainable platforms on their own. Photoshop and even Netscape Navigator come to mind as examples. But for the most part, the OS vendors have traditionally been the best at making kits, defining APIs, and ensuring enduring compatibility. These are the "product" part of the operating system and always involve a wider circle than the software engineers. Your quote indicates that you may be qualified to argue architectural details, but a little out of range discussing product issues, especially when normal people have to use the products.

Apple makes its web kit (the power behind Safari) available in Mac OS X now. Is that anti-competitive behavior? It's really no different than Microsoft packaging IE as an embeddable control that other developers can use in their applications. Perhaps the best engineered (and most ironic) example of bundling a web browser with an operating system was the BeOS. "NetPositive" exported a thing called a "Replicant", which could be embedded in any 3rd party application and had a very high level event interface to control it. That event interface was exposed in greater and greater detail in revisions to the BeOS. Oh, and it was ironic because Jean-Louis Gassée whined to the DOJ about Microsoft bundling IE!!

-Brad

fling93 writes:

Brad: Your whole post indicates to me that you come from the open source world and have a distaste for Microsoft.

No, as I already stated, I work in embedded systems, which nowadays means vxWorks from WindRiver, who does not compete with Microsoft. That you don't know what embedded systems means indicates you're not a programmer (at least, not any more than I'm an economist). My development platform is Windows 2000, because at the time our company bought our machines, WindRiver didn't offer a cross-compiling development platform for Linux. I use Windows XP at home for consistency. I don't use or develop for Linux. I have a distaste for bad programming practices (and monopolistic behavior). The stringent stability requirements for embedded systems is similar to that for OS programming. The GUI part has to be handed off to a browser on a host machine, but faces similar issues. Plus, my first job was actually in OS kernel development in a high-security system.

As a small-l libertarian, I acknowledge that dealing with monopolies can sometimes be worse than the monopoly itself, but nobody here seems to be arguing that. Instead, you're trying to defend indefensible technical practices from a product standpoint when your requirements can easily be met by doing things the right way.

It's very bad to share code between application and the OS for obvious security reasons. But also, as I mentioned before, because a programmer who has to modify the shared code is more apt to make a mistake breaking one or the other, and breaking the OS is serious. Moreso, programming for the OS is very different from programming applications, making this kind of mistake even more likely. Furthermore, the ideal library function is something simple and static (which is why the standard C library includes stuff like math operations and string comparisons, etc.). Stuff that will never change, exactly because changing a library function is a very risky deal. But an application library function is more likely to change much more often than an OS library function.

Especially something like HTML rendering, given that the HTML standard constantly changes.

One important ingredient is consistency, and one way to promote it is to have shared code (usually provided and promoted as the "right way" by the OS vendor) with solid, flexible APIs.

The more robust and secure way to do this is to provide libraries separate from the OS, or for services that only the OS can perform (like access to hardware) provide an API of kernel system calls, and require a minimum level of trust before a program is allowed to do this. For consistency of look, you can have a UI layer above the kernel layer, where the UI is only allowed to access user devices, like the display, mouse, etc., and have that UI layer provide an API for things like displaying windows and dialog boxes (I'm pretty sure Microsoft provides this API, although I'm not sure about the layering).

None of this requires sharing libraries between the OS and applications, but merely having the OS provide services, and tiering those services and access to those services to minimize risk. Allowing somebody to ask you for a service is much more secure than sharing code you both call. The main difference is that the "sharing" is one-way (note, the OS doesn't ever call anybody else's function) so there isn't much of a security issue beyond the damage the service can do (which again, is a trust issue, addressed by tiering and limiting access). Note, Microsoft already offered consistency in look and feel with Windows 95.

For things that the OS doesn't need to do (like rendering HTML), you can provide DLLs or statically linked libraries available to applications without having the OS call functions in those same DLLs. Otherwise, this is an easy point of attack for a virus: overwrite a DLL with a malicious one, and you can take over the OS and do whatever damage you want to do, anywhere on the machine. The security issues and bug-causing problems of shared code is less crucial when you share them only between applications.

There's absolutely no technical problem with Microsoft selling Windows with IE preinstalled along with DLLs available for other applications to call (other than the fact that the DLLs provide more installation headaches than the amount of resources they save vs. statically linked). But that's not what they did, which belies why they did it: that they wanted to confuse non-technical people to get around the charge of bundling for the purposes of leveraging a monopoly to increase market share in another market.

shamus writes:

By that argument, Cable TV isn't a monopoly either because consumers can choose broadcast TV (nevermind satellite).

Linux and Windows both operate on the same platform (a PC) and one can be interchanged with the other, whereas exactly one provider controls he RF cable coming into your house. Comparing computing and video markets, satellite and broadcast would be more analogous to heldhelds and cel phones. There are different markets for different technological platforms. The market for RF cable communications is mandated to be delivered by a government-approved monopoly. If the government made it illegal to load any OS other than Windows on a PC this would be an exact analogy for what government has done with cable TV.

fling93 writes:

Seems like the same argument would indicate plasma TVs don't compete with LCD TVs, or that Apple does not compete with Microsoft. The utility a consumer receives from a product is dictated by what the product does, not how it does it.

Gray writes:

TV monopolies will soon be forced to follow the "a la carte" business model...look what happened to the music recording industry! Consumers Unite!

Comments for this entry have been closed
Return to top