Forecasting activity using sentiment indicators : the case of Chile

Abstract

In this article we evaluate, in several dimensions, the ability of two sentiment indicators, consumers and business, to forecast year-on-year variation of Chilean activity. We do in-sample and out-of-sample exercises to evaluate predictive capacity. In out-of-sample exercises, when we predict activity using a constant we find that the business confidence indicator (BCI) have the capacity to improve forecasts, the economic perception index (EPI) do not shown predictive capacity in a naïve context. Adding a univariate structure, the results continue to show predictive ability for the BCI, the variable improve forecast for horizons 1 to 12 month ahead, for the EPI the results show predictive ability in medium term forecast, 9 and 12 months ahead. When we use the model proposed by Urrutia and Sanchez (2008) (USM), BCI continues to show predictive ability, but by itself, on average, does not deliver better forecasts that USM. For EPI we find predictive ability in out-of-sample exercises for medium term forecast. The hit rate exercises shows that BCI and EPI correctly predict changes in direction of activity in most horizons. We conclude that business confidence indicator can be use as a leading indicator of Chilean activity. A contribution of this paper is use Clark and West (2007) test in iterative method of forecast.