关键词: 偏好变化, AIDS模型, Hamilton/Costa方法, CPI偏差

Abstract: Theoretical studies by Economists suggest that the existence of substitution bias, quality change bias, new product bias and outlet bias in CPI causes overestimates of the real cost of living index. In addition to these biases above, our analyses suggest that there exists another kind of bias caused by changing preference, and the bias direction is different from the others. Based on the assumption of the changeability of preference and the modified AIDS model, this article, using the household expenditure survey (HES) data in 36 cities from 2002 to 2012, revises Hamilton/Costa method to re-estimates China's CPI bias via methods of rolling regression and weighted regression. The results show that after considering changeable preference, the yearly CPI bias (overestimation) significantly reduced from 2.93% to 0.95%. This proves that the existence of changeable preference, unlike the other existed bias makes CPI tend to underestimate the true cost of living index, that means CPI is not always overestimate the true cost of living index in general.

Key words: Changeable Preference, AIDS Model, Hamilton / Costa Method, CPI Bias