Is there a market for work group servers? Evaluating market level demand elasticities using micro and macro models

This paper contains an empirical analysis demand for “work-group” (or low-end) servers. Servers are at the centre of many US and EU anti-trust debates, including the Hewlett-Packard/Compaq merger and investigations into the activities of Microsoft. One question in these policy decisions is whether a high share of work servers indicates anything about shortrun market power. To investigate price elasticities we use model-level panel data on transaction prices, sales and characteristics of practically every server in the world. We contrast estimates from the traditional “macro” approaches that aggregate across brands and modern “micro” approaches that use brand-level information (including both “distance metric” and logit based approaches). We find that the macro approaches lead to overestimates of consumer price sensitivity. Our preferred micro-based estimates of the market level elasticity of demand for work group servers are around 0.3 to 0.6 (compared to 1 to 1.3 in the macro estimates). Even at the higher range of the estimates, however, we find that demand elasticities are sufficiently low to imply a distinct “anti-trust” market for work group servers and their operating systems. It is unsurprising that firms with large shares of work group servers have come under some antitrust scrutiny.