Abstract: This paper studies the insample and outofsample predictability of China's bond market index, term structure premium and exchange premium. Constructing 27 variables from macrolevel, marketlevel and microlevels and generating 6 principal components as prediction variables via principal component analysis (PCA), this paper finds that there is significant predictability in China's bond market on its composite index conditional on different proxies after stripping out special bonds. The prediction variable based on principal component extraction has a stronger forecasting ability, and some macroeconomic indicators and marketlevel indicators can predict the risk premium of bond market. The first principal component extracted based on 27 indicators has a strong predictive ability for term premium and exchange premium in the sample, but a poor predictive ability for outofsample. The conclusion of this paper shows that China's bond market is vulnerable to the impact of macroeconomy, and there is no great difference in market functions between the interbank market and the exchange market. The separation of the two markets deepens with the fluctuation of the bond market.

Key words: return predictability, insample return predictability, outofsample return predictability, market segmentation  ,

摘要：

本文研究了中国债券市场综合收益、长短期债券收益差异、银行间市场和交易所市场收益差异的样本内和样本外可预测性。本文选取宏观、中观（市场）、微观3个层面27个变量以及利用主成分分析法生成6个主成分变量作为预测变量，发现中国债券市场综合收益率在剥离了特殊品种债券之后具有一定的样本内和样本外可预测性。基于主成分提取的预测变量的预测能力更强，部分宏观经济指标和市场层面指标都可以预测债市风险溢价。基于27个指标提取的第一主成分对期限溢价、场所溢价在样本内均有较强的预测能力，而样本外可预测性较差。本文结论表明，我国债券市场容易受到宏观经济影响，银行间市场和交易所市场在市场功能上并没有很大的差异，两个市场的分割随着债券市场的波动而加剧。

［关键词］收益可预测性；样本内预测；样本外预测；市场分割

关键词: 收益可预测性, 样本内预测, 样本外预测, 市场分割