News engagement under threat on Facebook?

Earlier this year, Mark Zuckerberg announced that the Facebook news feed would put more emphasis on social relations and less on news. This is the result of several controversies involving Facebook, especially the publication of fake news during the 2016 US elections. One of the concerns of news organizations is that changes to the algorithm may threaten their reach and the audience’s news engagement.

Having collected all posts by news organizations (n=356,188 from 59 Dutch news titles), I am able to analyze how much engagement these posts had and how this may have changed over time. I will look at the number of likes for each post over a 16 week period. We do not know exactly when the new algorithm will be in effect, but sometime after January 12, when Mark Zuckerberg posted the news on Facebook.
Instead of the mean number of likes for posts, I compare the median number of tweets for each week over the 16-week period. I prefer the median over the mean number of likes, because the distribution of likes is extremely skewed: a few posts receive extremely many likes and most of posts only receive a few likes (see for instance De Echo van Nederland). The median is a better measure for central tendency measure because it is incentive for these extremely popular posts. For now, I aggregated the data on a weekly basis to eliminate intra-weekly fluctuations (such as weekends versus weekdays).

Reviewing the first figure, we see that the weekly number of posts on Facebook is quite constant over the 16 week period.

in the second graph (below), the median number of likes fluctuates strongly. We particularly see a drop in week 9 in 2018. Could this be the result of the new news feed algorithm?

For a more detailed view, I prepared the same graph for different types of organization (below). Although all types of platforms have an online presence, they differ in terms of their origins. I distinguish newspapers, television, RTV (mostly radio stations and regional and local broadcasters) and online-only platforms. We see that newspapers publish the most posts, simply because there are many newspaper platforms. Radio, TV and online-only platforms roughly publish the same amount of posts.

Looking at the median number of likes for each type of platform, we see that TV posts on Facebook receive the most likes, while the other platforms receive considerably less likes for their posts. As for the variation across these 16 weeks, we see the strongest decrease in likes for TV news and current affairs programs. The online-only and RTV platforms also shows a slight decrease, while newspapers remain constant over time. Again, in week nine the number of likes declines. At the end of week 16 though, there seems to be some recovery as indicated by a slight increase likes. Whether this increase is persistent, time will tell.

In conclusion, if I were to make a guess when Facebook’s news feed algorithm changed, it would be week nine. Furthermore, it mostly affects news posts that are extremely popular, in this case posts by TV organizations and current affairs programs. Then again, there are many factors that affect the number of likes. So, this might be a continuing story….

Looking into the future of newspapers

Last week there was a lot of discussion about Ross Dawson who predicted the time of death of the newspaper in several countries. Piet Bakker devoted a blog to it as did Trouw (there are probably many more sources so let me know in the comments). It’s important to know when the newspapers get extinct, so then we know when to have cleared our agenda’s to go to the funeral or at least have found a new job.

All kidding aside, we love predictions (especially weather forecasts)  and we need them. However, it’s one of the most difficult things to do. All predictions are uncertain. However some are more uncertain than others. Having no additional data, predicting the end of the newspaper is similar to looking into a crystal ball, humming some unclear sounds while burning incense: rubbish in gets rubbish out. Then again, if one would actually have additional data things brighten up. What data for instance? Well, first of all the hard data the publishers already have, even the data we researchers have little unconditional access to. For instance,  (1) detailed subscription data (paid, non-paid, discounts) (2) financial data (costs,  revenues), (3) marketing strategies  (4) changes in population composition, (5) competitors’ actions and developments, and (6) autonomous developments such as general economic growth and level of unemployment, but also technological developments. Some other suggestions have been made in comments on Piet Bakker’s and Ross Dawson´s blogs, such as the writing off of printing presses, the culture of reading for each country. All more or less relevant to be taken into account.

The prediction made by Piet Bakker (i.e. time series analysis) are, in my opinion, equally uncertain (see my posted comment). Especially because he used very little observations (six years) to predict far into the future (40 years). So, the further away the prediction is the more uncertain the prediction is. Not only that, a possible bias in a short term estimate is most likely increased the further away the prediction is, as such biasing the final prediction. Also, I suspect there will be an excellerated decrease in subscriptions at the end, because newspaper publisher (or the banks) will see the end is near and will pull the plug before they loose even more capital.

There are alternative approaches to the moving averages approch, though. The first one is using multiple time series, where the dependent is the number of subscriptions and the others are the independent ones (some of which are listed above). Alternatively, a system dynamics model could be developed taking account of feedback loops, because every action may have positive effects on the dependent variable (in this case the number of subscriptions) but can have an effect on the causes as well in the future. These complex problems are called messy (cf. Vennix, 1996) because of the large numbers of unknowns. There is alo a more qualitative version called group model building. This used when hard data do not exist. In this approach experts from different background are set around the table and chart the problem at hand. Then different scenario’s are explored to foresee possible futures. Subsequently, the aim is to develop a business strategy that is robust for many possible futures. These approaches have been used for instance by Shell and the Dutch government. I’m not sure it’s too late for publishers to undertake such an exploration, what what I hope they’ve learned that it’s best to prepare for the worst case scenario when you can. For the publishers that would have been sometime in the eighties of the last century. Then again, like quiting smoking, it’s never too late.

On a final yet important note, and this may be depressing for the publishers, the readers of physical newspapers may be a dying breed. It’s in large part a generations issue. People who grew up with (their parents) having a newspaper at home are the ones that were more likely to have a newspaper subscription as well. Until the Internet took of in the late nineties of the last century: youngsters that grew up in the age of the Internet are less likely to subscribe. A similar thing happended with television: young people that grew up with only bublic broadcasting continued watching public broadcasting while youngsters that grew up with commercial television as well didn’t watch public broadcasting. So, it seems inevitable to see newspapers as we know them disappear. If they survive it’ll be not in this fashion. In the technical sense, no medium has ever really disappeared, it’s function merely changed. This is likely to happen to the newspaper as well. What is more important is how journalism will evolve the coming decades. But that’s an entirely different discussion.

Newspaper markets and the generation gap

Piet Bakker has made it a large part of his academic career to monitor the developments on newspaper markets. This week he posted the latest analyses on recent and distant Dutch developments. The title leaves not to guess about : “Down, down, deeper and down”. As such, nothing’s new.

One interesting fact  however is that no one made a comment on the musical reference the title has. The title is of a song by the British rock band Status Quo (see the video clip of their performance on the famous Dutch show Toppop at the end of this post). Is this a generation gap between Piet, myself and the rest on the Nieuwe Reporter blog? Who knows…
Regarding a status quo and the newspaper market: As such there seems to be a large discrepancy between the status quo and the development of the numbers of newspaper copies sold in the Netherlands: i.e. down, down deeper and down. However, one could argue that there is a status quo in the decline of newspaper copies being sold. If so, the end is nigh for the printed newspapers. Which dont think will happen: no medium ever has disappeared in history. Merely their function may have changed. For now, we’ll have to wait and see patiently how the newspapers’ function will change in the coming decades.