A spurious regression occurs when a pair of independent series, but with strong temporal properties, are found apparently to be related according to standard inference in an OLS regression. Although this is well known to occur with pairs of independent unit root processes, this paper finds evidence that similar results are found with positively autocorrelated autoregressive series on long moving averages. This occurs regardless of the sample size and for various distributions of the error terms.

We analyze the properties of multiperiod forecasts which are formulated by a number of companies for a fixed horizon ahead which moves each month one period closer and are collected and diffused each month by some polling agency. Some descriptive evidence and a formal model suggest that knowing the views expressed by other forecasters the previous period is influencing individual current forecasts in the form of an attraction to conform to the mean forecast. There are two implications: one is that the forecasts polled in a multiperiod framework cannot be seen as independent from one another and hence the practice of using standard deviations from the forecasts' distribution as if they were standard errors of the estimated mean is not warranted. The second is that the forecasting performance of these groups may be severely affected by the detected imitation behavior and lead to convergence to a value which is not the "right" target (either the first available figure or some final values available at a later time).