Data Shows that Negative Remarks Lead to Fewer Followers

Continuing my series of TweetPsych based data points, this is based on analysis of over 100,000 accounts and looks at the “Negative Remarks” category. Negative remarks include things like sadness, aggression, negative emotions and feelings, and morbid comments.

As it turns out, nobody likes to follow a Debbie Downer accounts with lots of followers don’t tend to make many negative remarks. If you want more followers, cheer up!

If you liked this post, don't forget to subscribe to my RSS feed or my email newsletter so you never miss the science.


Alex Guest February 11, 2010 at 4:16 pm

Hi Dan,

The data shows a correlation between negative comments and fewer followers but it does not reveal a causal relationship. Is it that people with fewer followers make more negative comments or that negative comments lead to fewer followers? or something else?

My hypothesis is that people who have little skill in developing relationships are also less happy hence you will find more negative comments amongst those with fewer followers.

The straightline trendline is also problematic because the uneven spread of users – decreasing as followership increases – means the data at the large following end is less reliable, hence the shape of the graph. Putting that aside a better fit would be a power curve.

Tony February 11, 2010 at 5:10 pm

Now that's just plain ludicrous!

Nate Davis February 11, 2010 at 11:08 pm

Thanks for the chart Dan; I'd be interested to see a follow-up post detailing methodology/context, because while of course people like @scobleizer and @guykawasaki and @zappos tend to be generally positive, I would suspect that there are plenty (especially in the entertainment sphere) who use negativity for comedic effect. Also, did you end the graph at ~8k followers because of insufficient sample size above that? I know little about statistics, so pardon me if for such reasons @aplusk and so forth can't be included, but that's unfortunate because those folks have such a significant influence on the medium.

Also, I have to agree with Alex's point about correlation ≠ causation. Now of course nuanced, qualified statements like “Data show correlation between negativity and fewer followers) are not the way to get pageviews and retweets, sadly (I think the internet encourages sloppy logic), but coming from someone who calls himself a scientist, that sort of semantic fudgery is disappointing.

But hey, if a bunch of casual retweeters conclude from this that they should be more cheerful on Twitter, I suppose the world's a better place.

devonjordan February 11, 2010 at 11:50 pm

Interesting chart Dan, but i have a few questions. Forgive me if you have answered them elsewhere.
First, who is the 100k sample? In talking this out with a friend of mine, and looking at tweetpsych, we came to the conclusion that your sample comes from people voluntarily submitting to tweetpsych. I was wondering if this may create a bias in your data, as tweetpsych came from you, therefore most of your data would come from people who are at least social media saavy.
Second, and this is based on an assumption that a computer is deciding what a “Negative Remark” is, did you account for sarcasm/dark humor/irony, I cant imagine a computer can pick that up.

The message behind your graph is a sound one, happy tweets make for happy/more followers and “nobody likes to follow a Debbie Downer.” Its just the how you came to that conclusion that irks me a bit.

But like Nate said, if this just leads to more cheerful tweets, I'm happy with that.

Bertil February 12, 2010 at 10:37 am

Great to see that yours comments sound more and more like an academic seminar. Because we are there, let's adress causality: try to work with a unfollowing tracker to see what were the latest comments before someone goes away. (And try to compare that with how was mentionned someone getting new followers.)

I'd love to have more categories in “negative”: sad, bitching, dark humour, etc. I'm assuming that twitter offers statistically significant proportion of each of those — hence my surprise to see the jagged line. You might want to pool twitterers by categories, especially beyond 4000 followers, to smooth out the curve — and use a log-scale to match. I'd love to see ratio of negative, negative to positive, and total number of negative twits — or at least you saying which is most significant.

michaeldaehn February 12, 2010 at 5:58 pm

Anecdotal evidence:

I started a blog called Marketing that Sucks. I thought it would be fun to lampoon bad marketing. I also started a counter blog called Marketing that Rocks, just in case I found people doing it right. Much to my surprise the latter was much more popular.

It is easier to be negative. Also people are usually looking to learn how to be better and search for the positive when gathering information. I try to show what people are doing right now and find I get a better response.

Zack Covell February 12, 2010 at 9:56 pm

Tweets a very different from Emails which are very different from Facebook.

So; regardless of the emotion associated with the message or webpage, the platform in that peeps read the msg. will determine what is acceptable for emotional messages…

I find whenever I send out an email about Mike Dillard I get more click than anything. I've even sent out free ebooks (that I wrote) to my list and had zero clicks to it (and it provides REAL value), but no one found the ebook in the first place. Then on the same day my Mike Dillard emails were clicked like crazy.

Sometimes I feel like using the Internet to share my knowledge and excitement is a popularity contest.

Mike Stenger February 14, 2010 at 5:11 pm

Hey Dan,

This is some great data, however, to me this is pretty obvious. Yeah, the majority of people don't like negative people and those who tend to be more negative. Some do but most of us prefer to follow someone else. Seems like common sense to me but good data nonetheless Dan.

Promotional Products February 15, 2010 at 2:36 pm

I'm in the same boat on this one… I like the idea and it makes a whole lot of sense to me, but almost too much sense. Of course no one wants to be peppered by negative ideas all the time. But I am appreciative of the information to back it up.

kara February 22, 2010 at 11:49 pm

common sense applies again…??

Maciej Stachowiak February 23, 2010 at 12:48 am

we did platform that analyze Tweets constantly for certain phrase – like company name. Our analytical model takes into consideration 4.800+ sentiments describing emotions [not only bad :)].

I do publish analysis for phrase 'Buzz' in my Google Buzz profile:

john143 June 10, 2010 at 11:58 am

Loved Reading this post, many thanks for the information.

john143 June 10, 2010 at 11:58 am

thanks for posting this, was really useful and interesting to me.

Manjunath D S July 3, 2010 at 10:25 am

It is not a surprise how online behavior follows offline behavior. Most do not like negative response and put people away from those making negative personalities. Thanks for sharing.

{ 22 trackbacks }