Lost in translation. When sentiment metrics for one market are derived from two different languages

We compare two text-based proxies for the sentiment of investors in the Japanese market. Both proxies are constructed by Thomson Reuters using the same algorithm, and the only difference between them is that the first proxy is derived using only Japanese language items; the second is derived from English language items. The correlation between the proxies is low and this suggests that they measure different aggregate affective states. The English-language sentiment proxy is found to have a positive and statistically significant association with Japanese returns before a key date in the history of Abenomics relating to quantitative easing (April 2013). After this date, the Japanese-language proxy has a positive and statistically significant association with the market. Studies of sentiment are predominantly based on proxies derived from English language sources; our results contribute to this nascent research field and suggest that not using both the English and local-language sources when constructing text-based sentiment indices may lead to misleading or mistaken inferences.