In a joint study between the Universities of Pennsylvania, Baltimore, University College London and Microsoft Research, scientists have looked at how language and behaviour on Twitter could be used to predict a user's income.
Scientists analysed almost 11 million tweets from different job titles across Twitter. They analysed Twitter users with jobs such as factory cleaners and packers, earning approximately $27,679 per year, through to production managers and directors earning over $78,000 per year.
The scientists mapped Twitter users to their income based on their use of certain types of language. Users perceived as 'religiously unaffiliated and less anxious' appeared to have higher earnings. These higher income users were found to have "significantly more followers" and get retweeted more often.
The scientists also discovered that by analysing language use higher income users tweeted more anger and fear when they posted positive and negative general content.
The dataset revealed that African Americans earn much less ($38,571) than Caucasians ($50,442). Higher perceived education played a significant role in having higher income.
Users who were perceived as being Christian earn "significantly less on average than people who chose not to signal their religious belief".
The scientists also discovered that there is a gender pay gap, with females earning less than their male counterparts. Older age groups earned significantly more than younger groups, reaching a plateau after 35 years old.
Less anxious Twitter users had higher mean income ($50,429) than users perceived as neither anxious nor calm ($48,453).
Higher income users posted more about politics, non-governmental organisations (NGOs) and corporate topics.
Lower income users of Twitter adopted more swear words and use more URLs in their posts, showing that these users link to external content such as news, pictures or videos.
Higher income users get retweeted more and retweet much more content themselves. This indicates that high income users use Twitter more for content dissemination. Of course, the larger number of followers these higher income users have, raises the likelihood of a tweet being retweeted.
The majority of Twitter users do not post duplicate content but those who do have lower income. Lower income users also are more subjective online.
The use of anger and fear as emotions are more present in users with higher income. Lower income users express sadness, surprise and disgust emotions when they tweet.
Marketers will be able to use the information gleaned from this dataset to more accurately pinpoint the correct user demographic on perceived earnings - just by the language used in a tweet.