关键词: CPI预测,食品,非食品,高频数据,春节因素

Abstract: The accuracy of CPI growth prediction based on traditional methods has somewhat declined due to the influence of many factors, so a new predication method is needed. The differences of the new methods are as follows: firstly, the mainstream forecasting framework is changed from the bisection method of“food + non-food”into the trichotomy framework of“food + industrial consumer goods + services”, which has a more comprehensive content; secondly, the weight system of CPI sub-items is adjusted, according to such data as the urban residents consumption expenditure structure, the input-output table and so on; thirdly, it takes into account the factors of Spring Festival objectively and specifically, dividing them into the three situations, i.e., pre-festival, in-festival and post-festival, then each of them is quantified, so as to examine their impacts on the price changes of each sub-item. The retrospective test shows that the result of the new method is notably superior to the traditional methods in predicting the CPI growth. The distribution frequency of forecast error within 0.1 percentage point is close to 80%, and the forecast accuracy is nearly 15 percentage points higher than the results of the existing methods.

Key words: CPI prediction; food; non-food; high-frequency data; factor of Spring Festival