Two experts in human behavior have recently warned the scientific community and the marketing industry that data collected from social media networks are not to be trusted when issuing statistics. The two scientists compared this situation with a historic failure of political polling – the “Dewey Defeats Truman” headline on Tribune newspaper a day after President Truman had won the elections.
Open access to social network data bases has changed the face of on-line marketing for good. These data can be seen as a never-ending gold-mine that can turn overnight into every marketing executive worst nightmare.
Scientists say that social networks abound in personal revelations that on a larger scale don not help accuracy that much. Information on these platforms may be misleading since demographic differences are scarce; many messages are posted by bots that pretend to be humans and the posts’ content is often influenced by site design.
Also Facebook, for instance, that lacks a “Dislike button”, fails to provide accuracy – many users may have a negative reaction to a particular content, but aren’t allowed to express it in a poll, researchers noticed.
The two researchers have also analyzed the main drawbacks of some social media platforms with unexpected results.
Facebook is mostly used by females, and people with low income. Study shows 71% of American Facebook users are young females with a low income. Seventy-six percent of overall Facebook users are female, while 66 percent is male. Eighty-four percent of the Facebook users are people between 18 and 29, while 76 percent represents people with less than $50.000 per year income.
On Twitter it seems that there are more Afro-Americans than white or Hispanic. Instagram has more women than men (about 28 percent more) but not as many as Pinterest where there’s a 5 women to one man ratio. LinkedIn is more of a male network, with more black people than white. Its users are mainly in the middle ages group (30 to 64 years old), are highly educated and 38 % of them earn more than $75.000 per year.
The two researchers also revealed that on-line companies that provide social media information for their advertisers often put users in easy to spot groups (by age, gender, income etc.) making the individual user hard to be recognized as a person. Other social media networks report highly optimistic accuracy levels of their data which in fact don’t exist. For instance, Twitter reported political affiliation of its users was at a 90% level of accuracy; in reality, the accuracy level barely reached 65 percent.
In conclusion both scientists suggest their colleagues doing real social science to be aware of these issues and truly “know their data”.
Latest posts by Anne-Marie Jackson (see all)
- SF Hospital Slaps New Parents with $19K Bill for Baby Treatment - Jun 29, 2018
- Furious Trump Blasts Harley-Davidson for Moving Production Overseas - Jun 28, 2018
- Warning! MRI Machines Could Poison You - Jun 27, 2018