Almost all colorectal cancer patients who are treated with pharmaceuticals receive multiple drugs in the form of a regimen rather than a single drug, simi- lar to anti-retroviral “cocktail” treatments for AIDS patients. For example, the regimen with the greatest market share in 2005 contained four separate drugs: bevacizumab, oxaliplatin, fluorororicil, and leucovorin. The 12 regimens in our sample are reported in Table 1. Most of our analysis, therefore, is conducted at the level of a regimen rather than a drug.
To account for changes in quality of new goods we use a hedonic price regression and the construction of two quality-adjusted price indices. The quality-adjusted price indices are constructed from the estimated benefits of innovation that were derived from the estimation of a discrete choice model of demand for colon cancer drugs under two extreme assumptions. The technical details of each approach are described in this section.
Hedonic Price Regression
Hedonic price regressions were introduced by Court (1939) and formalized by Griliches (1961) as a way to account for the new goods problem. In essence, newer goods usually contain more desirable characteristics, and therefore, failing to account for the value of these characteristics will overestate the true change in prices. The hedonic price regression is motivated by the “hedonic hypothe- sis,” which states that goods can be viewed as aggregations or bundles of lower order variables that the literature calls characteristics or attributes. These char- acteristics are the variables that consumers care about and that are present in their utility functions. The hedonic function provides a dissaggregation of the observed transaction prices into the variables that affect the economic agents’ behavior. Court and Griliches proposed estimating a surface that relates prices