This paper analyses the impact of productivity developments in the United States and the euro area on the euro-dollar exchange rate. The paper presents a new measure of relative average labour productivity (ALP), which does not suffer from the biases implicit in readily available relative ALP data. Importantly, the patterns of these series differ widely. Employing the Johansen cointegration framework, four Behavioural Equilibrium Exchange Rate models are estimated using four different productivity proxies. Our results indicate that the extent to which productivity can explain the euro depreciation varies with the productivity proxy used: readily available measures explain most, our new, preferred measure least. If foreign exchange traders used the former to assess productivity developments, this might thus have contributed to the weakness of the euro in 2000/2001. In all models, however, productivity can explain only a fraction of the actual euro depreciation experienced in 1999/2000.