Hal R. Varian1
December 6, 1996
The advent of low-cost technology for manipulating and communicating information has raised significant concerns about personal privacy. Privacy is a complex issue and can be treated from many perspectives; this whitepaper provides an overview of some of the economic issues surrounding privacy.2
In particular, I first describe the role of privacy in economic transactions and argue that consumers will rationally want certain kinds of information about themselves to be available to producers and want other kinds of information to be secret. I then go on to consider how one might define property rights in private information in ways that allow consumers to retain control over how information about them is used.
The most fundamental economic transaction is that of exchange: two individuals engage in a trade. For example, one person, ``the seller'' gives another person, ``the buyer'', an apple; in exchange, the buyer gives the seller some money.
Let us think about how privacy concerns enter this very basic transaction. Suppose the seller has many different kinds of apples (Jonathan, Macintosh, Red Delicious, etc.) The buyer is willing to pay at most r to purchase a Jonathan, and 0 to purchase any other kind of apple.
In this transaction the buyer would want the seller to know certain things about him, but not others. In particular, the buyer would like the seller to know what it is he wants--namely a Jonathan apple. This helps the buyer reduce his search costs since the seller can immediately offer him the appropriate product. The transaction is made more efficient if detailed information about the consumer's tastes is available to the seller.
On the other hand, the buyer will in general not want the seller to know r, the maximum price that he is willing to pay for the item being sold. If this information were available to the seller, the seller would price the product at the buyer's maximum willingness to pay, and the buyer would receive no surplus from the transaction.
Roughly speaking the buyer wants the seller to know his tastes about which products he may be interested in buying; but he doesn't want the seller to know how much he is willing to pay for those products.
Armed with this simple insight, let us investigate a some more realistic examples.
When many people talk about ``privacy rights'' they are really talking about the ``right not to be annoyed.'' I don't really care if someone has my telephone number as long as they don't call me during dinner and try to sell me insurance. Similarly, I don't care if someone has my address, as long as they don't send me lots of official-looking letters offering to refinance my house or sell me mortgage insurance. In this case, the annoyance is in the form of a distraction--the seller uses more of my ``attention'' than I would like.
In the ``information age'' attention is becoming a more and more valuable commodity, and ways to economize on attention may be quite valuable. Junk mail, junk phone calls, and junk email are annoying and costly to consumers.
In the context of the apple example described above, it is as though the seller of apples has to tell me about each of the different kinds of apples that he has to sell before I am able to purchase.
It is important to recognize that this form of annoyance--essentially excess search costs--arise because the seller has too little information about the buyer. If the seller knew precisely whether or not I was interested in buying insurance or refinancing my mortgage, he could make a much better decision about whether or not to provide me with information about his product.
In the context of the apple example: it is in both parties interest to know that the buyer will only purchase a certain kind of apple. The seller has every incentive to present this information to the seller, and the buyer has every incentive to solicit this information from the buyer.
This is, in fact, how the direct mail market works. If I subscribe to a computer magazine, I will end up on a mailing list that is sold to companies that want to sell me computer hardware and software. If I refinance my house, I am deluged with letters offering me mortgage insurance. In these cases the seller is using information about me that is correlated with my likelihood of purchasing certain products. (See Blattberg and Deighton (1991) for a discussion of some current trends in direct marketing.)
In this context the more the seller knows about my preferences the better. If, for example, I an interested in buying a computer printer, it may well be in my interest and the seller's interest for this fact to be known. If I am only interested in a laser printer, this is even more valuable information since it further reduces search costs for both the buyer and the seller. If I already have a laser printer that I am happy with, the seller may find it valuable to know that since he will not have to incur costs trying in vain to sell me a new printer.
When a mailing list is sold to a third party, the relationship between the buyer's original interests and the seller's interest may become more tenuous. For example, suppose the list of computer magazine subscribers is sold to an office furniture supplier. Some of the people on this mailing list may or may not have any interest in office furniture.
Even though the first two parties in the transaction--the individual who may want to buy something, and the seller who may want to sell him something--have incentives that are more or less aligned, the transaction between the original owner of the mailing list and those to whom it is sold do not have such well-aligned incentives.
Economists would say that an externality is present. The actions of the party who buys the mailing list will potentially impose costs on the individuals on that list, but the seller of the mailing list ignores those costs when selling it.
These costs could be mitigated, to some degree, if the individual who is on the mailing list has a voice in the transaction. For example, the individual could forbid all secondary transactions in his personal information. Or, more generally, the individual could allow his information to be distributed to companies who would send him information about laser printers, but not about office furniture.
These considerations suggest that the difficulty in the ``annoyance'' component of privacy could be significantly improved if the communications channels between the buyers and the sellers were clearer, the information conveyed was more accurate, and third-party transactions were restricted only to those transactions that the original consumers authorized.
Let us now consider a more difficult case, the case where the buyer's revealing information about himself is detrimental. Suppose that they buyer wishes to purchase life insurance but knows information about his health that would adversely influence the terms under the which seller would offer insurance. In this case, the buyer does not want information released that would influence the price at which the insurance would be offered.
Suppose for example that the potential buyer of insurance is a smoker, and knowledge of this information would result in higher life insurance premium. Should the buyer be required to truthfully release the information? Since the information here concerns the price at which the service (insurance) is offered, the incentives are perfectly opposed: the buyer would not want to reveal that he is a smoker, while the seller would want to know this information.
Note, however, that a nonsmoker would want this particular information about himself revealed. Hence the insurance company has an easy solution to this problem: they offer insurance at a particular rate appropriate for smokers, and then offer a discount for non-smokers. This would succeed in aligning information incentives for the buyer and seller.
More generally, suppose that the price that the seller would like to charge is higher for people with some characteristic C. Then people who have that characteristic have bad incentives to reveal it, but people who don't have that characteristic have good incentives to reveal it. It is in the interests of the seller to construct the transaction in a way that the information is revealed.
We have seen that several of the problems with personal privacy arise because of the lack of information available between concerned parties. Perhaps some of these problems could be mitigated by allowing for more explicit ways to convey information between buyers and sellers.
For example, it is common to see boxes on subscription cards that say ``check here if you do not want your name and address redistributed to other parties.'' This is a very primitive form of contract. A more interest contract might be something like: ``Check here if you would like your name distributed to other parties who will provide you with information about computer peripherals until 12/31/98. After that, name and address will be destroyed. In exchange you will be paid $5.00 for each list to whom your name and address is distributed.''
Although it would be hard to fit this sort of contract on a subscription response card, it would be easy to fit it on a Web page. The contract that is being offer implicitly assigns property rights in an individual's name and address to him or herself, unless the individual chooses to sell, or more properly, rent, that information.
This particular legal policy seems quite attractive: assign a property rights in information about an individual to that individual, but then allow contracts to be written that would allow that information to be used for limited times and specified purposes. In particular, information about an individual could not be resold, or provided to third parties, without that individual's explicit agreement.
This idea appears to have been most thoroughly explored by Laudon (1996). He goes further than simple contracting and suggests that one might sell property rights in personal information on markets. As Laudon points out, there is already a large market in personal information. But the property rights are held by those who collect and compile information about individuals--not by the individuals themselves. These third parties buy and sell information that can impose costs on those individuals, without the individuals being directly involved in the transactions. In economic terminology, there is an externality.
The personal information industry in the US is primarily self-regulated, based on the so-called Fair Information Practices.3
The European Community has more explicit privacy regulation; for more on international regulations, see the Electronic Privacy Information Center's page on International Privacy Standards.
It is worth observing that the Fair Information Practices Principles would automatically be implemented if the property rights in individual information resided solely with those individuals: secret information archives would be illegal; individuals could demand the right of review before allowing information about themselves to be used; and those who wanted to utilize individual information would have to explicitly request that right from the individual in question or an agent acting on his behalf.
Laudon goes on to propose that pieces of individual information could be aggregated into bundles that would be leased on a public market he refers to as National Information Market. For example, an individual might provide information about himself to a company that aggregates it with 999 other individuals with similar demographic and marketing characteristics. Such groups could be described by titles such as ``20-30 year old males in California who are interested computers,'' or ``20-30 year old married couples interested in home purchase.''
Those who wanted to sell to such groups could purchase rights to use these mailing lists for limited periods of time. The payments they made would flow back to the individual users as ``dividends.'' Individuals who found the annoyance cost of being on such lists greater than the financial compensation could remove their names. Individuals who felt appropriately compensated could remain on the lists.
Although they are many practical details of implementation that would need to be solved to implement Laudon's market, it is important to recognize that information about individuals is commonly bought and sold today by third parties in market-like environments. The National Information Market simply gives individuals an economic stake in those transactions that they currently do not have.
There may be information about me that I don't want revealed just because I don't want people to know it. For example, many people are very touchy about personal financial information being revealed. They don't want other people to know how much income they make, or how much they paid for their house or car.
In some cases there is a social interest to making such information public. Consider the following two examples.
In each of these cases there is a public interest in having this information publicly available. Making information available about owners of motor vehicles may help ensure safer operation. Making sales prices of houses available may help ensure the accuracy of tax assessments. My neighbors may care about the assessment of my house, not because they particularly care about my tax assessment, but because they care about their tax assessment.
Whether or not such information should be publicly available would ideally depend on an individual benefit-cost analysis. If I am willing to pay more to keep my assessment private than my neighbors would be willing to pay to see it, we have a potential way to make everyone better off: I pay my neighbors for the right to keep my assessment private. If they value seeing my information more than I value keeping it private, then they pay me for the right to see it.
This sort of transaction is not really practical for a variety of reasons, but the same principle should apply in aggregate: one has to weigh the ``average'' potential benefits from making this sort of information public to the potential costs of keeping it private. The presence of a market where individuals can sell information about themselves helps provide a benchmark for such benefit-cost calculations.
Certain kinds of information can be collected and distributed without revealing the identity of individuals. Froomkin (1996) explores some of the legal issues involving anonymity and pseudonymity; see Camp et al. (1996) for a computer science view. Karnow (1994) proposes the interesting idea of ``e-persons'', or ``epers,'' which are serve to provide privacy while conveying a relevant description of the individual.
Many sorts of public information have been available at some transactions cost: in order to find housing assessments, it has typically been necessary to travel to a city or county office and look up the information. Now that increasing numbers of consumers are computerized it is possible to acquire this information much more inexpensively. Information that was previously deemed useful to be publicly available under the old transactions technology, may now be deemed to be too available.
This, it seems to me, has a reasonably simple solution. The information could be made available in digital form, but at a price that reflected the transactions costs implicit in acquiring the information using the old technology. The price paid for the information could then be used to defray the cost of making it publicly available.
For example, suppose that, on average, it took a citizen one hour to go to the country records department, look up a tax assessment and photocopy the relevant material. Then a reasonable charge for accessing this information online might be on the order of $25 plus 20 cents or so per assessment requested.
This sort of charging schedule essentially restores the status quo, provides some funds for local government, and offers an additional choice to individuals. People who didn't want to pay the $25 could make the trip to the county records office and access the same information there ``for free'' (i.e., paying no monetary cost.)
I have argued that an appropriate way to deal with privacy issues is to determine a baseline assignment of rights, but allow individuals to trade those rights if they desire to do so. If there are no transactions costs in trading or negotiation, the initial assignment of privacy rights is arbitrary from the viewpoint of economic efficiency.4
To see this, suppose that it is worth 50 cents a week to me to have my name omitted from a junk email list, and that it is worth 20 cents a week to the owner of the junk email list to have my name on it. If the owner of the email list has the right to put my name on it without consulting me, then I would have to pay him some amount between 20 and 50 cents to have him remove it. On the other hand, if he has to seek my permission to use my name, it would not be forthcoming, since the value to him of having my name on the list is less than the value to me of having it off. Either way the property rights are assigned, my name would end up off the list.
If there are significant transactions costs to making contracts such as these, the standard Coasian arguments suggest that an efficient allocation of rights would be on in which the transactions and negotiation costs are minimized. In this case, the appropriate comparison involves the transactions cost to the individual to having his or her name removed from the list to the cost to the mailing list owner of soliciting permission from individuals to add them to the list.
When phrased in this way, it appears that the current practice of adding someone's name to a list unless they specifically request removal probably minimizes transactions costs. However, the rapid advances in information and communications technology may change this conclusion. The development of social institutions such as Laudon's market would also have a significant impact on transactions costs.
Privacy becoming a very contentious public policy issue. The danger, in my opinion, is that Congress will rush into legislation without due considerations of the options. In particular, a poorly-thought-out legislative solution would likely result in a very rigid framework that assigned individuals additional rights with respect to information about themselves, but did not allow for ways to sell such property rights in exchange for other considerations.
In my view, legislation about rights individuals have in information about themselves should explicitly recognize that those rights can be ``leased'' to others for specific uses, but cannot be resold without explicit permission. This simple reform would lay the foundation for a more flexible, and more useful policy about individual privacy.
In addition it would enable business models that would potentially allow for reduced transactions costs and better matches between buyers and sellers.