How to Intelligently Regulate Microsoft

What would you do about Microsoft?

Aw, one of my favorite subjects. As someone who thinks he knows a lot about software, the computer industry, microeconomics, industry and competitive analysis and law, I believe I can say something useful on this subject

Government regulation of private enterprises — whether administrative or antitrust lawsuits — should be based on microeconomic theory. Classical microeconomic theory, such as the works of Alfred Marshall, are based on diminishing returns — products or companies eventually run into limitations, so that a predictable equilibrium of prices and market share is reached. Classical microeconomics is based on the bulk-processing (e.g., metal ores, pig iron, coal, lumber, heavy chemicals, soybeans, coffee), smokestack economy of Marshall’s day. If a coffee plantation expands production, it is driven to use land less suitable for coffee — it would run into diminishing returns. The market would be shared by many plantations, and a market price would be established at a reasonable level. Classical microeconomics works fairly well in describing and explaining these kinds of companies. Intelligent government regulation of these industries is and should be based on classical microeconomic theory.

A new type of enterprise and economy has developed, primarily starting after World War II. It is based on increasing returns. Increasing returns are the tendency for that which is ahead to get further ahead, for that which loses advantage to lose further advantage. Increasing returns generate not equilibrium but instability: If a product or company or a technology gets ahead by chance or clever strategy, increasing returns can magnify this advantage, and the product or company or technology can go on to lock in the market.

The leading scholar on increasing returns is Brian Arthur of the Sante Fe Institute and Stanford, whose Harvard Business Review article on this subject provides a good summary.

Mechanisms of increasing returns exist alongside those of diminishing returns in all industries. But roughly speaking, diminishing returns hold sway in the traditional part of the economy, which increasing returns reign in the newer part, knowledge based industries. Since World War II, knowledge based companies have become an ever increasing part of our economy. (This is particularly true because the classic processing industries have mostly become hyper efficient — we can grow enough food to feed our country and to export food with less than 3 percent of our labor force working in agriculture.)

Operating systems are the classic example of increasing returns. You’re probably using the Microsoft Windows operating system, and you’re doing so probably not because you consider it to be the paradigm of well designed software, stability or good security, but because almost everyone else uses it. Since almost everyone else uses it, software developers primarily write for the Windows platform, making it increasingly attractive to users, and thus increasingly attractive to developers, and then to users, and then to developers, and so on. This is the classic positive feedback cycle.

The government’s primary attempt to regulate Microsoft has been to file antitrust lawsuits against Microsoft. Antitrust law is based primarily on two statutes: the Sherman Antitrust Act (1890) and the Clayton Antitrust Act (1914). These statutes were passed approximately 100 years ago, when the economy was based almost entirely on diminishing returns and thus the economic theory behind such statutes was based entirely on classical microeconomics. Subsequent case law interpreting these statues has also been primarily based on classical microeconomics.

Therein lies the fundamental problem with how the government is seeking to control Microsoft — it is using laws and methods based on antiquated and inappropriate economic theory. The people suing Microsoft (the Department of Justice and the Federal Trade Commission) and the people adjudicating the dispute (the trial judge and the justices on the appellate courts) need to understand increasing returns. Regulating a knowledge based company is completely different than regulating a steel company. In addition, they need to understand technical concepts in computer science as well as the computer industry. The number of lawyers and judges in the U.S. who understand computers can be counted on your fingers, with lots of fingers left over.

Given that the prosecutors and judges need to understand this new form of economics as well as computer science and the computer industry, and they clearly don’t, it should not be surprising that they got it wrong. Given their lack of knowledge, what would be truly amazing would be if they got it right.

One needs to start with an assessment of what benefits Microsoft has provided to our society and economy. Microsoft has provided extraordinary benefits. The most important is that Microsoft has provided a common ubiquitous platform — the Windows operating system — for which developers can write software. This has huge advantages for consumers, who can purchase tens of thousands of software packages which would not exist if there were 100 operating systems, each with 1 percent market share, creating a world where it didn’t make economic sense to write a software package for each small market.

In addition, by separating the operating system from computer hardware, Microsoft has commoditized the hardware industry, creating a hyper efficient industry that continually provides extraordinary improvements in price-performance. Every year you can purchase more and more of a computer for less and less money. The trend will continue for the foreseeable future. (Some believe that at some point we will hit a wall as we reach certain fundamental physical limits, such as the speed of light. For a variety of technical reasons I believe they are clearly wrong.)

People with little understanding of the history of the computer industry underestimate these two advantages provided by Microsoft. I used to work for Digital, which manufactured minicomputers (in fact, they manufactured very good minicomputers). Although Digital was the market leader in minicomputers, there were numerous other minicomputer manufacturers, such as IBM, Data General, Prime, Wang, Tandem and others. Numerous software packages were simply not written because a software publisher would have to write for so many platforms. There is radically more choice in software packages today than when I worked at Digital.

And if you were a Digital customer, Digital controlled everything. They made the CPU, the memory, the disk drives, the printers AND the operating system. (The same was true if you had an IBM minicomputer, or a Prime.) If you wanted to upgrade memory, you paid through the nose. Today, you can purchase memory from 100+ different companies, thus ensuring a very low price.

Accordingly, Microsoft has provided an enormous benefit to you and me in what it has done. We can purchase computers more powerful than supercomputers of a decade ago for less than $1000, and we have an almost infinite choice in software packages to use.

The problem is that along with these enormous benefits, Microsoft has caused considerable harm, and does so more and more each day. Some of this is inevitable in an increasing return industry. Much of this harm, however, is caused by business practices that are clearly unethical and immoral, and as adjudicated by the trial court and affirmed by the appeals court, illegal. (Although that case was primarily concerned with Microsoft’s illegal conduct under the browser market and particularly with its actions toward Netscape, that conduct is a very small part of business conduct by Microsoft that is illegal.)

Microsoft’s dominance over the computer industry is hard to exaggerate. It is a monopolist in desktop operating systems and desktop productivity software. It is a significant player in local area network and wide area network operating systems (Windows Server). It is the most important publisher of development tools and languages, such as Visual Basic and C++. It is the most important developer of Internet software and tools, and its Internet browser has dominant market share. Its database management system (SQL Server) is one of the three leading DBMSs and its market share is growing. As of January, 2004, Microsoft is generating more than $1 billion a month in excess cash, after taxes. As of July, 2004, Microsoft had more than $60 billion in cash (this was before they declared their $30 billion dividend). The interest on its cash is greater than the revenues of all but a few of its competitors. It has so much cash on its balance sheet that under normal SEC rules, it would be classified as a mutual fund rather than an operating company and it must obtain an exemption from these rules every year.

The issue is what to do about this, particularly going forward. My goal is not to punish Microsoft, nor am I looking to enrich Microsoft’s competitors. My only concern is what is best for consumers. In medicine there is a saying, “First, do no harm.” If you are going to regulate Microsoft, your first concern should be not eliminate these two enormous benefits provided by Microsoft. So let’s look first at what not to do. Putting Microsoft out of business clearly makes no sense. There are tremendous advantages in having a monopoly operating system that provides a common platform for developers and which commoditizes the computer hardware market. And whatever you think of Windows, it is clearly becoming over time a better operating system. Stability in the current versions of Windows — XP for the desktop and Windows Server 2003 for the server — is no longer much of a concern. Security is now the primary concern in Windows, and I’m confident that Microsoft will over time solve that reasonably well, if for no other reason than a good chunk of the Fortune 500 companies will start to consider alternatives such as Linux on the desktop if Microsoft doesn’t.

Breaking Microsoft into two separate companies — the proposal accepted by Judge Jackson, which was overturned on appeal — makes little sense if you understand microeconomic theory. Judge Jackson accepted the plaintiff’s theory that since Microsoft was primarily a monopolist in two large markets — operating systems (Windows) and desktop productivity software (Microsoft Office) — it would make sense to have two separate monopolists, one for Windows and one for Office. The problem is that these two monopolies are interdependent on each other and thus in the current system, Microsoft has a strong economic incentive to keep the price of these two monopoly products lower than they otherwise be. Microsoft finds it economically rational to price Windows lower than it would otherwise be because it knows that if you purchase Windows, you will likely also purchase Office, and thus Microsoft gets paid twice. If there are two separate companies, the operating system monopolist would increase the price of Windows, because it would no longer care whether a sale of Windows also meant a sale of Office. And the Office monopoly company would also raise the price of Office, since it would no longer care about what happens to Windows’ market share. The fact that Judge Jackson accepted the plaintiffs’ proposal to create two separate monopolies clearly indicates that he is clueless about microeconomics and the computer industry (as are the plaintiffs).

There was even a dumber proposal on the table — create, say, five companies that each owns the rights to Windows. Each would be free to develop Windows, creating five different versions of Window, each incompatible with each other. Very quickly, there would be no common platform for which to write software, so many if not most software developers would not be able to afford to develop software packages. (This is what happened in the Unix world, where there are 10+ flavors of Unix, each slightly incompatible with each other. Linux offers the chance to remedy this absurdity, although some worry that even inux may become fragmented. (Some are concerned, for example with what Red Hat is doing.)

If one wants a structural remedy (i.e., breaking Microsoft up), I would divide Windows into, say, six companies: one R&D company and five marketing companies. One of these companies would own the rights to Windows and the right to develop and enhance Windows; this would be the R&D company. It would not sell Windows, however. Windows would be sold by the five marketing companies. Each marketing company would have the right to sell the Windows operating system (“OS”) that is being developed by the R&D company. The R&D company would be a non-profit company. It would receive revenues from the marketing companies to fund future Windows development, but not to earn a profit. The five marketing companies would sit on the Board of Directors of the R&D company, along with various computer experts appointed by a federal commission.

Let’s assume that the Board of Directors concludes that $1 billion a year going forward was needed to develop Windows, and that the five marketing companies estimated they could sell 100 million copies of Windows a year. Each of the five marketing companies would pay $10 a year to the R&D company for every Windows license they sold. The five marketing companies would then be free to sell Windows for whatever price they could charge, to cover their marketing costs, a profit, and whatever services they wanted to provide. Since these five marketing companies are selling the same identical version of Windows, they will be selling a commodity product with all of the advantages that provides for consumers, while at the same time consumers would enjoy the advantages of having a platform that software developers want to write for. If one marketing company charged too much for Windows, the other four marketing companies would undercut it. These five marketing companies would compete with each in providing service. You may choose to buy a more expensive version of Windows from marketing company III because it provided better technical support than the others. (The same is currently true in purchasing commodity computer hardware — you can purchase a high quality computer from Gateway and get lousy technical support, or you can pay a little more to Dell and get great technical support.)

The same would be done for Microsoft Office: one R&D company and five marketing companies. One could also do this for Windows Server, but there is less need to, since the server market is fairly competitive and Linux’s market share is growing faster than Windows Server’s share.

Putting aside structural remedies, there are numerous clearly unobjectionable remedies that should have implemented if the courts knew what they are doing. (Unobjectionable assuming you’re not Microsoft, that is.)

The first is that Microsoft should be required to provide much more technical information so that people can use its software better and so that potential competitors could develop competing products.

1. Microsoft should be required to publish its source code. Not some of it, but all of it. (In doing so, its legal ownership of this source code would not be changed. You still could not legally pirate its software, it’s just that you could see what is going on.)
2. Any customers who has paid the standard license fee to Microsoft would be allowed to modify the source code (and enhance it) for its own internal use. That way, if there was a huge security hole that really bothered you and that Microsoft would not fix in the foreseeable future, you could always fix it yourself, or hire a consultant to fix it for you. (Richard Stallman, founder of the open source software movement (he prefers the term “free software”) asks, “Even if you’re mechanically inept, would you ever purchase a car where you could not open up the hood?”)
3. Microsoft should be required to adhere to the coding standards espoused in its own technical books published by Microsoft Press. If the source code provided to the pubic is not properly commented, for example, it is much harder to understand what is going on.
4. Microsoft should be required to publish all of its internal technical manuals and specifications for all of its software packages. For Microsoft Office, for example, Microsoft has over 500 notebooks of technical specifications used by its development team. One notebook, for example, covers solely the topic of how to convert 1-2-3 spreadsheets into Excel spreadsheets, while another covers how icons should be standardized across Office applications. This information would be extremely valuable to technical customers trying to understand why Office is causing certain problems.
5. Microsoft should be required to publish all of its standard Application Programming Interfaces (“APIs”) and be forbidden to use any “hidden” APIs that it does not publish. APIs are the mechanism by which an application program communicates with the OS. There have been many allegations that the Office developers have an unfair advantage because they use secret, unpublished APIs.
6. Microsoft should be required to published the data file formats of all of its software packages. It’s very hard to develop a product to compete with Word since one of the features a competitive word processor must offer is to be able to open all Word documents flawlessly. Since the data file formats of Word .doc files are not published, it is basically impossible to do this flawlessly. At the same time, Microsoft should be required to convert its data file formats into ASCII files, rather than binary files (ASCII is much easier to read), that are understandable, based on XML, making it easier to understand its file formats and making repair of corrupted files much easier.
7. Microsoft should be required to publish all of its technical specifications, such as the layout of its file systems.

There are several other things Microsoft could be required to do.

The first is to publish versions of Office for the Linux and Macintosh operating systems, releasing such versions at the same time or before it released the same version for Windows. These versions would be required to work as well, or better, than the Windows version. (Microsoft currently publishes a version of Office for the Macintosh OS, but it is only the basic edition of Office, which excludes packages such as Microsoft Access.)

Microsoft should be required to distribute competing versions of software that are currently provided for free by their publishers. Thus, when you purchase Windows, you would not only receive the Internet Explorer browser, but also the current version of Netscape, Opera, and other browsers, all already installed so you don’t have to find them, download them and install them. When you install the OS, the default browser would not be Internet Explorer, but another browser chosen at random. (You would always have the choice to make Internet Explorer your default browser.)

Microsoft should be required to distribute as part of Internet Explorer a version of Java that is acceptable to Sun.

Given that the most serious problems of Windows is its terrible security, Microsoft would be required to fund an independent security foundation, at a cost of $100 million a year. ($100 million is about a week’s worth of interest on the $50+ billion in cash that Microsoft has.) This foundation would hire the smartest, most diabolical teenage hackers in the world and pay them well to find every conceivable security loophole in Windows and other Microsoft products. Microsoft would be required to fix these problems as they were found within 90 days.

Some consider my proposed measures too harsh toward Microsoft. My response is that Microsoft has been found guilty of illegal practices and that verdict was sustained on appeal. (The appeals court overturned the proposed remedy (namely the court-ordered breakup of Microsoft) but not the finding of liability.) A federal court adjudicating an antitrust court sits as a court of equity and thus has very broad powers to structure remedies it sees fit.

I should add that in many ways I admire Microsoft. Unless most companies that have become fabulously successful, they are rarely complacent. Their hiring standards are exceptionally high. Their managers are extremely smart; even the middle managers I’ve met there are extremely impressive. Microsoft learns from its mistakes and rarely makes the same mistake twice. Bill Gates is one of the smartest and most forward thinking people in the world. His idea to step down as CEO of Microsoft in favor of Steve Ballmer, so that he can do what he does best (product design and thinking about the future) while Ballmer does what he does best (day-to-day administration), has created one of the most successful Chairman/President relationships in the history of capitalism. (The Michael Dell – Kevin Rollins partnership is also unusually productive.) If properly regulated, Microsoft could continue to provide enormous benefits to our society and economy.

(If this subject interests you, I highly recommend the last few chapters of “High St@kes, No Prisoners: A Winner’s Tale of Greed and Glory in the Internet Years” by Charles H. Ferguson (New York: Random House, 392 pp., 1999). It is a captivating account of how Ferguson started Vermeer Technologies, which published FrontPage (a Web authoring tool) which Microsoft purchased for $130 million. Ferguson is perhaps the most insightful analyst of the computer industry.)

Read James’ essay on minimizing total cost of ownership in information systems.

List of other essays written by James Mitchell  |  Copyright notice

Cite as “How To Intelligently Regulate Microsoft ” by James Mitchell. December 12, 2003, version 1.4.
www.jmitchell.me/essays/intelligently-regulate-microsoft.