Researchers compared how Twitter sentiment noise relates to next-quarter losses for domestic and foreign stocks. The estimated slope was on average 1.40 lower for domestic firms than for foreign firms, and the relationship was negative in 78% of quarters for domestic stocks versus 48% for foreign stocks. The authors interpret this as evidence that the noise-loss link is temporally more stable for domestic firms.
To explain the gap, they develop a noisy rational expectations model and examine the composition of the contributor pool. Feeds for foreign firms have a higher repeat-author concentration and a larger model-implied nonfundamental component than feeds for domestic firms. The authors attribute this to a producer-knowledge mismatch: the English-language contributor pool underrepresents home-market investors, who tend to hold informational advantages about foreign firms.
They conclude that the foreign signal is not gone but diluted. The findings suggest that who contributes to online discussion matters for the predictive value of social-media data in financial markets.