International comparisons of pharmaceutical prices: what do we know, and what does it mean?

In their article in this issue of the Journal of Health Economics, ‘‘Cross-National Price Differences for Pharmaceuticals: How Large, and Why?’’, Patricia Danzon and Li-Wei Chao report on research suggesting that, contrary to conventional wisdom and previously published research, prices of pharmaceuticals in the US are not much higher than in other developed countries. The findings reported in previous studies, they argue, are biased because they rely on unrepresentative samples of drugs and employ unweighted index number procedures. Ž . The Danzon–Chao hereafter, DC study is complex and deals in great detail with the relatively esoteric nuances of price index construction. A thorough assessment of the issues raised explicitly by DC, as well as those suggested implicitly, could easily take twice the length of the original DC article. Any editorial must, by necessity, be much shorter than such a detailed assessment. Thus, in interpreting the provocative findings from this important, useful and informative article, I believe it useful here to focus on but four sets of issues. The first concerns price index construction procedures. The economic theory of index numbers provides a set of criteria that allows us to distinguish and identify preferable from less preferable price index number construction procedures. DC are correct in pointing out that most previous research on international comparisons of pharmaceutical prices has employed procedures that do not stand up to close professional scrutiny: the products sampled are unrepresentative, the samples are much too small, and frequently the price indexes are simple unweighted means. However, DC find that when one employs several alternatives among the set of more preferable price index construction procedures, results vary dramatically depending on what weights one employs. Recall that in the context of bilateral price index construction, a Laspeyres price index with country i as the reference country weights prices in countries i and j by country i’s quantities, whereas the corresponding Paasche price index weights both countries prices by country j’s