We offer the first direct evidence of an implicit contract in a goods market. The evidence we offer comes from the market for Coca-Cola. We demonstrate that the Coca-Cola Company left a substantial amount of written evidence of its implicit contract with its consumers-a very explicit form of an implicit contract. In general, observing implicit contracts directly is difficult because of their implicit nature. To overcome the difficulty, we adopt a narrative approach. Based on the analysis of a

more »

... rge number of historical documents obtained from the Coca-Cola Archives and other sources, we offer evidence of the Coca-Cola Company not only saying that it had an important implicit contract with its consumers, but also acting on it. This study makes an additional and unique contribution by exploring quality as a margin of adjustment available to Coca-Cola. We present evidence that the implicit contract included a promise not only of a constant nominal price but also a constant quality. We document the dedication to a 6.5oz serving of the "Secret Formula." Indeed, during a period of over 70 years, we find evidence of only a single case of true quality change. By studying the margin of adjustment the Coca-Cola Company chose in response to changes in market conditions, we demonstrate that the perceived costs of breaking the implicit contract were large. In addition, we are able to offer one piece of direct evidence on the magnitude of these costs by studying the events surrounding the failed introduction of the New Coke in 1985. "Sellers . . . can influence the shopping behavior of customers by pledging continuity of an offer . . . Yesterday's offer has a strong influence on today's demand." Arthur Okun (1981, p. 28) "Implicit contracts strike many economists . . . as quite plausible. Yet their existence is-almost by definition-not subject to objective verification, at least not from conventional data sources." Alan Blinder, et al. (1988, p. 149) "[Narrative approach] allows a vast body of information that cannot be employed in conventional statistical tests, to be brought to bear on [the]... question." Christina Romer and David Romer (1989, p. 167)