The Death of Television 2 – Blame it on the Weather!

Heatwave_2A much-anticipated Royal Wedding and successful 2018 FIFA World Cup (with England reaching the semi-finals for the first time since 1990) have resulted in some notable mass audience success stories for schedule-based television in recent months. The 4th series of Love Island has also managed to break records for ITV2, attracting droves of younger viewers to the channel, and additionally helping to make the ITV Hub more popular than the BBC iPlayer in recent weeks. Indeed, the combination of the World Cup and Love Island resulted in monthly ad revenues for ITV being nearly 23% higher in June-2018 versus June-2017. The continuing popularity of schedule-based television has also been highlighted by the widely reported unhappiness of Virgin Media customers following the removal of the UKTV channels from the platform on 22-July, after the two failed to agree on a new carriage deal.

While no serious commentator would deny that the TV industry is facing significant challenges and needs to evolve to avoid being eclipsed, these recent events do appear to attest to TV’s underlying resilience and continuing relevance, and the fightback against the FAANGs does look to be gaining some traction, with both Netflix and Facebook experiencing notable share price drops after announcing lower than expected second-quarter subscriber/user growth. It may therefore come as a surprise to many readers that the BARB measured consolidated (i.e. Live + 7-Days catch-up) Individuals 4+ Total TV audience was 5.8% lower in 2018-Q2 than it was in 2017-Q2, falling from 8.07 million to 7.60 million. In fact, if we eliminate the impact of population growth and look at quarterly Average Daily Minutes of Total TV viewing, the decline is even more pronounced, with a 7.1% drop from 195.6 minutes in 2017-Q2 to 181.8 minutes in 2018-Q2 (down 13.8 minutes). Inevitably this will be cited by some commentators as evidence that TV is in terminal decline, and that the notable success of recent months can do nothing to prevent its inexorable demise. Is this, however, really what is going on?

It is no coincidence that we have been in the grip of a heat wave approaching the legendary status of that of 1976, and the notion that the weather has a significant impact on TV viewing levels is well established within the industry, with the rationale behind this being that, whatever the season, viewers are less likely to stay in and watch television if the weather is unusually warm and sunny, or vice versa if it is unusually cold, wet and dark. There is of course a strong underlying seasonality to television viewing habits in the UK, with Total TV being consistently lower in the spring and summer than it is in the autumn and winter, but the point here is that if, for example, one were to compare winter viewing levels in two different years, the expectation would be that, other things being equal, the more severe winter would have higher levels of TV viewing. With such a significant (7.1%) drop in 2018-Q2 coinciding with a heat wave, it is therefore noteworthy that (despite strong competition from the SVOD services) the year-on-year quarterly Average Daily Minutes of Total TV only fell 2.8% (down 6 minutes) in 2018-Q1, a quarter during which the weather was particularly inclement with the likes of ‘The Beast from the East’ and ‘Storm Emma’. In fact, declines in excess of 7% in quarterly year-on-year Total TV Minutes are not unprecedented, with 2014-Q1, for example, a particularly mild quarter weather wise, being down by a very notable 7.2%, a decline that was considered so significant at the time that BARB included an article in the 2015 edition of The Viewing Report outlining the findings of an analysis it had conducted on the impact of the weather on TV viewing levels. It is worth quoting the conclusion from this article (ibid., p 12) in full:

“…in recent memory, some commentators have tended to assume that if TV ratings are down year-on-year, this proves that people must be dropping out of the broadcast stream in order to watch content on new platforms like Netflix or YouTube. Our analysis here gives the lie to that non sequitur, showing as it does that there may be simpler (indeed, more elemental) factors in play. Our weather formula offers an important new method for analysing what can, at first sight, be perplexing variations in year-on-year viewing patterns.”

Given the well documented impact of the SVOD services on schedule-based TV audience levels in recent years, it is important to remember that BARB’s analysis was based on 4 years of weather and TV audience data ending in early 2014, so the underlying downward pressure on TV viewing that has characterised the last 5 years was still in its early stages. As such, it is perfectly understandable why no allowance was made for the impact of market fragmentation in BARB’s model, with any potential adverse effect on the model’s reliability over its analytical timeframe being limited. This is, however, something that will need to be addressed in any weather-based TV viewing model when looking at more recent audience data (starting from 2013 onwards), as failure to do so would result in significant model misspecification, making any associated predictions unreliable.

Our key objective is to assess to what extent the two weather variables: ‘Maximum Temperature’ and ‘Daily Sunshine Hours’, that were identified by BARB as having the most correlation with television viewing at the time, are still relevant in the more fragmented TV landscape of today, and if so, to what extent they help explain the very notable 7.1% drop in quarterly year-on-year Total TV Minutes in 2018-Q2. We have therefore adopted a quarterly timeframe for our model, and with the relevant weather data being available from the Met Office up to the end of June-2018, we are also able to cover the quarter of interest. What remains is to account for the impact of audience fragmentation, and while one option would be to incorporate a simple (effectively catch-all) time-trend into the model to help factor in the underlying downward pressure on schedule-based TV viewing levels, there is a more compelling (and statistically speaking also more robust) alternative.

From July-2013 onwards, BARB has been capturing 8-28 days catch-up viewing as well as the time spent on other TV-based activities ranging from SVOD (Netflix, Amazon, Now TV, etc.) to gaming, and as these activities cannot currently be reliably distinguished from one another, they are collectively categorised as ‘Unknown’ viewing. As discussed in my previous research note in the ‘Death of TV’ series (aimed at dispelling some of the misinformation around this topic and highlighting that, while facing significant challenges, TV is evolving rather than dying), the decline in the consolidated (Live + 7-Days) Total TV audience is closely correlated with the growth in 8-28-Days and (most notably) ‘Unknown’ TV viewing, and we now have 5 years (20 quarters) worth of relevant data to work with. The results of our final weather-based TV viewing model are outlined in the Figure below.


The main takeaways are that the weather variables identified by BARB are still highly relevant, and that by additionally factoring in the underlying impact of market fragmentation, our model can explain nearly 99% of the variations in the quarterly Average Daily Minutes of consolidated Individuals 4+ Total TV viewing over the last 5 years. The model also suggests that around half of the 7.1% decline in 2018-Q2 is likely to have been caused by the unusually warm and sunny weather, dispelling the notion that the underlying downward pressure on schedule-based TV viewing levels has suddenly accelerated.

Being able to isolate the impact of such occasional seasonal factors from true underlying structural ones is, of course, the basis of any robust forecasting model, and now more than ever it is important to gain a reliable understanding of the trajectory in which schedule-based TV viewing is heading. It is only by building robust forecasting models, that we can gain a better understanding of underlying vulnerabilities and test/inform strategies designed to avoid worst case outcomes. To end, as seems appropriate, with a weather-based analogy, while the forecasts linking accelerated global warming to rising greenhouse gas emissions are of scientific interest in their own right, their true value is in the role they have played in first persuading people of their veracity, and then informing strategies to help us actively counteract the worst of the predicted outcomes.

If you would like to receive the associated research notes to Dr Farid El-Husseini’s blog posts please email him directly on:

This entry was posted in BARB, Death of Television, Facebook, Forecasting, Heat Wave, Impact Assessment, Model, Netflix, Regression, SVOD, Total TV Viewing, Weather. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *