关键词: 投资者关注, 本地偏差, 网络搜索, 信息优势

Abstract: Classic asset portfolio theory suggests diversification is helpful for reducing portfolio risks. However, it is well documented that despite the benefit of diversification, both professional investment managers and individual investors display a strong preference for investing in local stocks. Two competitive theories are suggested to explain the causes of the local bias. The information advantage theory argues that local investors get information advantages from local social networks with local agents such as executives, employees, customers, suppliers and local media et al. Meanwhile, the behavioral theory attributes the local bias to investors' irrational biases such as familiarity preference, herding behavior, overconfidence et al. However, prior empirical studies show inconsistent results to the effectiveness of these theories.
One of the difficulties to empirically study the causes and effects of the local bias is the lack of appropriate proxies for the extent that investors over allocate their assets towards local stocks. Based on the close interaction between investors' asset allocation and attention allocation, the ratio of stocks' Baidu search volume (which is expressed as Baidu index) initiated by local netizen is used to that initiated by all netizen nationwide to measure on what extent investors over allocate their attention to local stocks so as to indirectly measure investors' local bias on asset allocation. With a sample of A-share stocks from 2007-2016, it is found that on average the percentage of Baidu search volume initiated by local netizen is 5.73 times of their population percentage, which indicates that investors in A-share market over allocate their attention and asset to local stocks notably. Investors in less developed provinces pay more attention to local stocks. Stocks with smaller sizes, lower book-to-market ratios, less turnover, lower debt-to-asset ratios and fewer shareholders are more attractive to local investors. The impact of investors' local attention on A-share stock risk premiums, price synchrony and pricing efficiency is empirically tested so as to test whether the local bias in A-share market is better explained by investors' information advantage or behavior biases. It's found that stocks with more local attention show higher risk premiums, stronger price synchrony and lower pricing efficiency. All these findings are consistent with the irrational behavior hypothesis, indicating that local investors in A-share markets do not show significant information advantages.
Three aspects are contributed to the literature. First, since Baidu indices are officially released by Baidu and could be easily and freely obtained online, a convenient and effective proxy of the local bias for related researches is provided. Second, with empirical findings from three different perspectives, it is provided that the local bias of A-share investors is better explained by the irrational behavior theory rather than by the information advantage theory. Third, it is shown that investors' geography preference on attention allocation makes significant impact on the process and efficiency of A-shares pricing, which is meaningful for A-share market participants and regulators.

Key words: investor attention, local bias, internet search, information advantage