The starting point for any analysis of insurance classification is an obvious but fundamental fact: insurance is only one of a number of ways of satisfying the demand for protection against risk. With few exceptions, insurance need not be purchased; people can forgo it if insurance is too expensive. Indeed, as the price of coverage rises, the amount purchased and the number of people purchasing will decline. Instead of buying insurance, people will self-insure by accumulating saving to serve as a cushion in the event of loss, self-protect by spending more on loss protection, or simply use the money not spent on insurance to purchase other goods and services. An insurer must compete against these alternatives, even in the absence of competition from other insurers. One method of competing for protection dollars is to classify potential purchasers into groups according to their probability of loss and the potential magnitude of losses if they occur. Different risk classes may then be charged different premiums, depending on this expected loss. Were it not for the need to compete for protection dollars, an insurer could simply charge each individual a premium based on the average expected loss of all its insureds, without incurring classification costs. In constructing risk classes, the insurer's goal is to calculate the expected loss of each insured, and to place insureds with similar expected losses into the same class, in order to charge each the same rate. An insurer can capture protection dollars by classifying because, through classification, it can offer low-risk individuals lower prices. Classification, however, involves two costs. First, the process of classification is costly. Insurers must gather data and perform statistical operations on it; marketing may also be more costly when prices are not uniform. Second, classification necessarily raises premiums for poor risks, who purchase less coverage as a result. In the aggregate, classification is thus worthwhile to an insurer only when the gains produced from extra sales and fewer pay-outs outweigh classification costs plus the costs of lost sales. Even in the absence of competition from other insurers, an insurer who engages in at least some classification is likely to capture more protection dollars than it loses. When there is not only competition for available protection dollars, but competition among insurers for premium dollars, the value of risk classification to insurers becomes even clearer. The more refined an insurer's risk classifications, the more capable it is of "skimming" good risks away from insurers whose classifications are less refined. If other insurers do not respond, either by refining their own classifications or by raising prices and catering mainly to high risks, their "book" of risks will contain a higher mixture of poor risks who are still being charged premiums calculated for average risks. These insurers will attract additional poor risks, and this resulting adverse selection will further disadvantage their competitive positions.