Directional Accuracy of Urban Consumers’ Inflation Forecasts in China

Abstract

We evaluate the directional accuracy of inflation forecasts based on the survey data of urban savings account holders in China. By using a new market-timing test, we show that the urban consumers’ expectations of inflation are not a useful predictor of the overall consumer price index (CPI) and the urban household CPI (U-CPI) in China. However, after our in-depth analysis using the inflation rate of each category in the U-CPI basket, we find that the consumers’ forecasts are useful in predicting the movement of the residence component in the U-CPI basket since the third quarter of 2009.

Keywords:

• Chinese urban consumers’ inflation forecasts
• directional analysis
• residence category
• survey data
• urban household consumer price index (CPI) basket

JEL Classifications:

• C53
• E17
• E31
• E37

Notes

^1. The data can be found on the PBOC homepage, http://www.pbc.gov.cn/rmyh/index.html.

^2. Please refer to Ash, Smyth, and Heravi (1998) and Henriksson and Merton (1981) for detail.

^3. The PT test focuses on predictive failure. In general, the null hypothesis of predictive failure is not equivalent to the hypothesis that the predicted and actual values of the variable are independent. However, they are equivalent in 2 × 2 cases as in this study.

^4. For the proof and more details, see the appendix in Pesaran and Timmermann (2009).

^5. In Pesaran and Timmermann (2009), the concern is the dependence among serially correlated multi-category variables, which covers the case of two serially correlated binary sequences. They developed two versions of the test: one based on the trace canonical correlation, and the other on the maximum canonical correlation. We focus on the former because, in the two-dimensional case, the performance of the two versions is similar. See Pesaran and Timmermann (2009) and Chou and Chu (2011) for details.

^6. Since the third quarter of 2009, the PBOC has adopted the new method called the PBOC version of the X-12-ARIMA seasonal adjustment for the use of monthly and quarterly data.

^7. The forecast revision is defined as the difference between two successive forecasts for the same target period. See Isiklar, Lahiri, and Loungani (2006) for more detail.

^8. The survey respondents are urban savings’ account holders in China.

^9. The results are available upon request from the author.

^10. Table 4 presents the urban household wealth composition in China

^11. The results are available upon request from the author.

^12. We do not report these results in this article because they do not change the conclusion. However, the results are available from the author upon request.