Table 14.

Differences in raw annualized returns of long-short portfolios (unique cashtags)

Panel A: Twitter
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Harvard-IV (1)	–	−1.30	−1.24	−2.16	−0.63	−1.01	−0.79	−1.45	−1.61
LM (2)		–	0.13	−0.73	0.72	0.34	0.46	−0.34	−0.29
L1 (3)			–	−1.04	0.59	0.24	0.39	−0.45	−0.49
L2 (4)				–	1.47	1.21	1.37	0.35	0.44
VADER (5)					–	−0.36	−0.21	−0.96	−1.00
Naive Bayes (6)						–	0.22	−0.63	−0.74
Max. entropy (7)							–	−0.74	−0.86
Deep-MLSA (8)								–	0.02
DeepMoji (9)									–

Panel B: StockTwits
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Harvard-IV (1)	–	0.60	−0.31	−1.47	−0.21	0.48	0.45	−0.79	−0.81
LM (2)		–	−0.88	−2.05	−0.82	−0.11	−0.13	−1.38	−1.29
L1 (3)			–	−1.31	0.09	0.83	0.82	−0.54	−0.61
L2 (4)				–	1.22	2.17	2.31	0.43	0.66
VADER (5)					–	0.66	0.65	−0.69	−0.65
Naive Bayes (6)						–	−0.04	−1.28	−1.68
Max. entropy (7)							–	−1.31	−1.76
Deep-MLSA (8)								–	0.07
DeepMoji (9)									–

The table depicts the t-statistics of the average differences in raw returns of long-short portfolios based on daily bullish sentiment estimated from short messages published on Twitter (Panel A) and StockTwits (Panel B) mentioning a unique cashtag. More precisely, we take the difference between returns obtained from methodologies reported in the rows with those reported in columns. The t-statistics are constructed using Newey-West (1987) standard errors. Differences which are statistically significant at the 5% level are highlighted by boldfaced numbers