The use of real-time information in Phillips curve relationships for the euro area

The dynamics of the Phillips Curve in New Keynesian, Expectations Augmented and Hybrid forms are extremely sensitive to the choice, timing and restrictions on variables. An important element of the debate revolves round what information decision-makers took into account at the time and round what they thought was going to happen in the future. The original debate was conducted using up to date, revised estimates of the data as in the most recent official publications. In this paper, however, we explore how much three aspects of the specification of the information available at the time affect the performance of the various Phillips curves and the choice of the most appropriate dynamic structures. First we consider the performance of forecasts, published at the time, as representations of expectations. Second, we explore the impact of using 'real time data' in the sense of what were the most recently available estimates of the then present and past. Finally we review whether it helps to use the information that was available at the time in the choice of instruments in the estimation of the relationships rather than the most up to date estimate of the data series that has been published. Thus different datasets are required in the instrument set for every time period. We use a single consistent source for 'real-time' data on the past, estimates of the present and forecasts, from OECD Economic Outlook and National Accounts. We set this up as a panel for the euro area countries covering the period since 1977. The OECD publishes forecasts twice a year, which permits a more detailed exploration of the importance of the timing of information. Our principal conclusions are (1) that the most important use of real time information in the estimation of the Phillips curve is in using forecasts made at the time to represent expectations; (2) real time data indicate that the balance of expectations formation was more forward than backward-looking; (3) by contrast using the most recent, revised, data suggests more backward-looking and less well-determined behaviour.