Fig 1.
The number of cryptocurrencies in our dataset created each month between 2019 and 2021.
Fig 2.
Plot of the in-sample estimated engagement coefficients.
Table 1.
Interaction coefficients estimated using all data.
Note that the interaction coefficient for liking is set equal to one.
Fig 3.
Plot of the cryptocurrency engagement coefficients versus creation date.
Table 2.
Engagement coefficients and returns for cryptocurrencies with top returns in different time periods.
Fig 4.
Plot of the evolution of the returns of different cryptocurrencies as a function of the logarithm of their engagement coefficient (estimated using prior data).
The returns are measured relative to one month after the cryptocurrency was created.
Fig 5.
(left) Histogram of the cryptocurrency mean bot probabilities. (right) Plot of the median returns in each bot probability cluster versus time (relative to one month after the cryptocurrency’s creation).
Fig 6.
Plot of the (left) AUC and (right) absolute value of the Spearman correlation coefficient for predicting returns versus time (relative to one month after the cryptocurrency’s creation) for the social media features.
Fig 7.
Plots of returns of investment strategies versus holding time for different social media features.
The investment strategies are based on fixed cryptocurrency holding times and minimum feature thresholds. Each plot is titled with the feature considered.
Fig 8.
Plots of returns of investment strategies versus feature threshold for different social media features.
The investment strategies are based on fixed cryptocurrency holding times and minimum feature thresholds.
Fig 9.
Plots of one month returns of investment strategies for different investment dates based on thresholds for different social media features.
The thresholds are quartiles of the relevant feature for cryptocurrencies created before the investment date.
Fig 10.
Plots of 12 month returns of investment strategies for different investment dates based on thresholds for different social media features.
The thresholds are quartiles of the relevant feature for cryptocurrencies created before the investment date.