Using nationally representative workplace data for Britain we show that over the last quarter century union voice -- especially union-only voice -- has been associated with poorer climate, more industrial action, poorer financial performance and poorer labour productivity than nonunion voice and, in particular, direct voice. On the other hand, union-based voice regimes have experienced lower quit rates than non-union and "no voice" regimes, as theory predicts. Over that time, while the workplace incidence of voice has remained constant, with roughly 8 workplaces out of 10 providing some form of voice, there has been a big shift from union to non-union voice, particularly direct employer-made voice. Thus employers are prepared generally to bear the costs of voice provision and manifest a reluctance to engage with their workforce without voice mechanisms in place. The associations between non-union voice mechanisms and desirable workplace outcomes suggest that these costs may be lower than the benefits voice generates.