The influential empirical papers of Robert Barro (1977; 1978) have been challenged in a number of recent contributions. Barro offered support for the Lucas-Sargent-Wallace (LSW)'policy ineffectiveness proposition'by demon strating, with US time series data, that only unanticipated money growth had any real output or unemployment effects. His results have been replicated in similar studies for other countries1.

The challenge to his and similar results has taken several forms. Mishkin (1982a; b), for example, has shown that Barro's results depend crucially on the length of the lag allowed for in modelling the influence of money on output. Alogoskoufis and Pissarides (1983) challenged the LSW proposition by arguing that price inflexibility is an important feature of the UK economy. In a rather different approach, Pesaran (1982) rejects a modified Barro model in favour of one displaying'Keynesian'features. He argues that existing econometric tests are weakened by their failure to consider'at least one genuine alternative'(p. 535). As the alternative models are not nested in his case, Pesaran was obliged to use non-nested hypothesis testing procedures. In a similar vein, Gordon (1982) has argued that the LSW proposition can only be properly tested by a comparison with a reasonable alternative. Gordon nests two alternative hypotheses in a single equation: the LSW proposition and what he terms the Natural Rate Hypothesis-Gradual Adjustment of Prices (NRH-GAP). Using this approach it is in principle possible to reject both hypo theses or either one: it is conveniently impossible to accept both. In this paper, the real influence of nominal income change-anticipated and unanticipated-is investigated by the application of Gordon's approach to UK quarterly time series data over the period 1963.2 to 1982.2. In general Gordon's findings for the United States are strikingly replicated: the LSW hypothesis is firmly rejected under alternative prior restrictions and under alternative informational assumptions. This rejection arises from the fact that anticipated nominal income growth and lagged inflation have real output effects, contrary to the LSW proposition.