There are two leading models of wage determination for the unionised sector of the economy. The first, which is labelled the monopoly model (MM) by Oswald (I985), assumes that the union sets the wage and the employer chooses the profit maximising employment level. The second, which is labelled the efficient bargaining model (EBM) by Oswald (I985), notes that both sides can improve on the monopoly outcome by jointly bargaining over wages and employment. A key strategy in empirically distinguishing between the two models has been to determine whether the employment-wage pair appears to be off the firm's demand curve for labour.1 For example, Brown and Ashenfelter (i 986) use the significance of a measure of alternative wages in an employment regression as evidence for the EBM. Efficiency wages hypotheses (in various forms) have been used to explain why non-union firms might find it profitable to offer wages above marketclearing levels.2 While paying high wages entails direct costs, this practice may result in benefits such as high worker effort, low turnover, or a large supply of labour to the firm. One motivation for this literature is to explain involuntary unemployment: if the efficiency wage framework is valid, then firms may not lower wages even in the face of excess supplies of labour (Stiglitz, I987). An additional motivation of this literature appears to be the empirical observation that interfirm or inter-industry wage differentials remain even after most possible economic determinants of these differentials have been controlled (Krueger and Summers, I988). Further, these differentials appear to lower quits and raise the length of queues of job seekers attempting to gain entry. They suggest the existence of rents associated with efficiency wages (Katz and Summers, I 989). If any version of the efficiency wages story is correct, then there are reasons for believing that workers' alternative wages will affect employment in the unionised sector even in the absence of efficient bargaining. Intuition suggests that an identical regression to that proposed by Brown and Ashenfelter (I986) would yield a significant alternative wage in the non-union sector of the economy, since the marginal revenue product of labour is affected by alternative wages