Home

/

GitHub

/

vilaweb: Reproducible data analysis for www.vilaweb.cat

73.269 piulets esborrats

“Tenían preparadas bombas”, va piular el diputat del PP Alejandro Fernández al matí del 23 de setembre del 2019. “Cuando advertí de la ‘batasunización’ del movimiento separatista a través de los cdr, me dijeron de todo. O se corta esto de raíz o tendremos un grave problema. Repito: tenían bombas”.

Mitja hora després, va esborrar el piulet.

El mateix dia, la diputada de Ciutadans, Inés Arrimadas va piular lo següent: “Torra llama a apretar y los radicales obedecen. Comandos separatistas habrían estado organizando acciones violentas con explosivos mientras millones de catalanes siguen desprotegidos por el Gobierno de España. Gravísimo. ¿Esto es normalidad, Sr. Sánchez?”. 20 minuts després, va esborrar el piulet. Però una hora després va tornar a piular el mateix, substituent la paraula “comandos” amb “radicales”.

“Condeno las cargas, faltaría más” va piular el diputat socialista Raúl Moreno en referència a l’1 d’octubre a la nit del 9 d’agost del 2018 com a part d’una conversa al Twitter amb l’usuari @xevialde. El 22 de maig del 2019, va esborrar el piulet. El mateix dia va esborrar centenars d’altres piulets.

“¿Alguna cosa a decir?” va piular el 25 de setembre del 2019 el regidor del PP de Badalanoa, Xavier García Albiol, etiquetant al President català Quim Torra i citant un article de la Vanguardia sobre les suposades confesions dels CDRs detinguts. “La presidencia de la Generalitat está en manos de un personaje indigno del cargo que ocupa capaz de justificar la violencia” va continuar Albiol. Menys de mitja hora després, va esborrar el piulet.

“Cita aquest tuit amb l’insult en català que mes t’agradi” va piular l’usuari ‘CatalanThings’ a la vespra del 3 de gener del 2017. Una hora i quatorze minuts més tard, el diputat Gabriel Rufían va citar el piulet afegint ell el texte següent: “Referèndum pactat amb l’estat”.

El 10 d’agost del 2019, 2,5 anys després, Rufián va esborrar el seu piulet.

“Hem de resistir la inhabilitació de @QuimTorraiPla. No resignar-nos”, va escriure el diputat de la CUP Albert Botran a la tarda del 3 de gener. El piulet va sobreviure 7 segons abans de ser esborrat. Uns minuts més tard, Botran va tornar a piular, però amb un missatge amb més matissos: “Si el President @QuimTorraiPla decideix plantar cara hem d’estar al seu costat totes les forces partidàries de l’autodeterminació, ens agradi més o menys el seu govern…”

In today’s era, social networks are an integral part of politics. Politicians say what they think and react to the news (and one another) in real-time. What they do on social media gets recorded in an easily searchable format forever. Unless, of course, it gets deleted.

Deleting a social network post, such as a tweet, is fairly commonplace. Normal people do it, and so do politicians. Most deleted tweets occur in the minutes immediately following the original post, often because of an error in spelling or grammar. Some were simply accidental touches of the “retweet” button, corrected by un-retweeting.

But there are other deleted tweets which are of far more interest. That is, sometimes a politician goes back to a tweet from 1, 2 or even 5 years ago and removes it. This usually goes entirely undetected by the public. In fact, this is how Twitter (and other social networks) want it to be: if users can’t control (delete) their content, users might be more hesitant to generate that content which, ultimately, would be bad for business.

But public figures, such as politicians, are a different case. What they say (and tweet) is of public interest, and so is what they delete. Politwoops is a platform dedicated to keeping tabs on politicians on Twitter, and capturing their deleted tweets for the historical record. The Politwoops data is publicly available and searchable: for example, through Politwoops’ website, one can easily and quickly see the deleted tweets of Lorena Roldán, Carles Puigdemont, and others.

The questions

Which politicians delete their tweets? How many do they delete? How long after the original tweet do they take them down? What kind of tweets get deleted?

The method

Unlike most of the articles I write, there is no “hypothesis” to test here. Instead, this is exploratory data analysis, meaning that the structure of the analysis is not guided by a pre-designated plan, nor by a falsifiable premise, but rather by a sort of “wandering” around in the data. For the sake of transparency, and as a check on my own selection and confirmation biases, I’ll do my best to document the process here.

Important note: politwoops does not keep track of all politicians. The deleted tweets of those who are not tracked by the platform are, unfortunately, lost forever.

The results

Defining the politicians: First, I compiled a list of 485 politicians of interest and retrieved their Twitter usernames. This included (a) all of the representatives in the Catalan Parliament, (b) all of the representatives in the Spanish Congress, and (c) a few dozen other political figures of relevance to Catalan and Spanish politics. The full list of those politicians is available here.

Retrieving the data: Next, I retrieved all of the deleted tweets for all the politicians in my list. Unfortunately, the Politwoops platform only tracks some politicians, so the final dataset ended up being only 85 politicians (that list is here). I wrote some code to fetch the data from the Politwoops platform, and voilà, 12 hours later I had 73269 deleted tweets in a local database. The code for this entire process is publicly available (here) as well as the data itself (here).

Let’s start by examining the number of tweets deleted by each politician.

Of course, it should come as no surprise that politicians who tweet more also delete more. So, let’s examine the number of deleted tweets as a percentage of the number of all tweets sent by that politician (at the time of their last deleted tweet).

Interesting: Roger Torrent has apparently deleted 99% of his tweets. But a manual look at the account being tracked by Politwoops reveals that they are tracking his old account (rogertorrent11), rather than his current one (rogertorrent). So, not so interesting after all (I warned you, this is exploratory analysis).

What about “raulmorenom”, Raúl Moreno, a PSC parliamentarian in the Catalan Parliament? He has deleted 8591 tweets. And what about ‘susanabeltranga’, Susan Beltran García, a representative in the same chamber for the “Ciutadans” party - she has apparently deleted more than half her tweets. Pol Gibert (PSC) and Inés Arrimadas (Ciutadans) also stand out as having high deletion rates.

When did they delete these (thousands of) tweets? What were these tweets about? Let’s dig in further.

We don’t know exactly why a politician chooses to delete a tweet. But when that tweet is deleted gives us a clue. Generally, a tweet deleted after just a few seconds was due to an error. A tweet deleted after a few months or years might be due to a change of opinion. And a series of many tweets being deleted at the same time suggests a systematic social media “clean-up”, ie an effort to erase the person’s previous online presence.

The best way to see these patterns is to literally see them. So, let’s have a look. On this page we’ve generated one chart for each politician. How do you interpret these charts? Each dot is a month; its positioning on the x-axis reflects when the tweet was created (more recent tweets to the right, older tweets to the left); its positioning on the y-axis reflects when the tweet was deleted (recently deleted tweets towards the top, tweets deleted a long time ago towards the bottom); its size tells you how many tweets were deleted for that time combination; and its color refers to whether they were tweets or retwets. You’ll note a diagonal line going from the bottom-left to the upper-right: this is the high rate of tweets that are deleted immediately after their creation.

Perhaps going through one hundred charts is not your idea of an enjoyable Saturday morning. So, let’s highlight just a few that are interesting.

Albiol is a prolific tweeter, with 15070 tweets on his timeline at the time of writing. But he is also a prolific deleter (1061 deleted tweets). His deletion behavior is similar to most people’s: he sometimes makes a mistake, and deletes the tweet very soon afterwards (hence the pronounced diagonal line). Most of his deletions are retweets.

The calendar of Sánchez’s deleted tweets is notably different from Alibol’s. Ones eyes are immediately drawn to the horizontal lines: one at the beginning of 2019, and one at the beginning of 2020. Horizontal lines in this kind of chart means deleting many tweets at once. Like Albiol, nearly all of Sánchez’s deleted tweets are simply removals of retweets.

On January 6, 2020, at about 8:00PM, it appears that the former secretary of communication for the Ciutadans party deleted his twitter account entirely, along with 2201 tweets.

Rufián tweets frequently, and also deletes frequently. Most of his deletions are retweets. There are no clear patterns in his plot, suggesting that he randomly goes through and cleans up old tweets.

From 2016 until mid-2017, Inés Arrimadas appears to have used an automated tweet deletion service, which deleted here tweets approximately 3 months after they were written. In fact, all of her tweets prior to July 22, 2017 have been deleted (in total, this makes for 3517 deleted tweets). Since then, she now rarely deletes tweets (apart from those which are tweeted and then immediately retweeted).

Relative to Arrimadas’ 3517 deleted tweets, Junqueras deletes very little (only 437), mostly retweets.

Like Junqueras, Puigdemont has deleted relatively few tweets, most of them retweets.

Iceta is a very prolific tweeter and retweeter (59922 tweets!). Which makes it all the more surprising that he has only deleted 651, mostly retweets. It is curious, however, that he appears to regularly go back to mid-2015 and delete retweets from that period.

Nogueras used to delete more tweets than she does nowadays. And almost all of her deletions are retweets. Her plot shows a clear pattern of random, rather than systematic, deletions.

Gibert’s chart is one of the most unique. Two clear lines are visible: a horizontal line in early 2017 (the horizontal line means that he deleted many tweets at one time), and a diagonal line above the “real-time” line (meaning that he had likely set up a software which automatically deleted tweets after one year).

The horizontal line: May 24, 2017, when he deleted 1892 tweets in one day.

Moreno also appears to have undergone a social network cleanup on May 22, 2019, a day in which he deleted 7834 tweets.

More interesting than when tweets were written and deleted is what: that is, what were the deleted tweets about.

Unless you have a spare 610 hours (the time it would take to read all 73269 tweets at a rate of 30 seconds per tweet), we’ll need to search. And manually exploring deleted tweets by myself is, simply, too big of a job.

So, rather than explore the 73269 tweets myself, I’ll simply hand them over to you, the public. You can explore them through the politwoops website interface if you’d like, but that interface is built around scrolling through specific politicians, rather than looking for specific text. So, to faciliate the latter, here’s a simple web application built for quickly searching (for any text) and filtering (for the Spanish/Catalan politicians in this analysis): https://datacat.cc/piulets.

Enjoy!