Futures Insight

The Premier League off-season offers fans a chance to reflect on the latest set of 380 matches now consigned to history. The 2016/17 edition was the first of three for which broadcasters Sky and BT have paid a combined £5.1bn for the UK live TV rights to 168 matches per season, up 70% from £3.0bn for 154 matches in each of 2013/14, 14/15 and 15/16. Putting aside our various on-field allegiances, we at Futures are interested in assessing the success of the season from a broadcast perspective; one year in, how will Sky and BT bosses feel about their investments?

The “Premier League TV audiences have fallen” rhetoric has sustained all year, with numerous articles noting a decline in average audience per match from 2015/16 to 2016/17, and drawing a direct line to doom and gloom for Sky and BT, or even the popularity of live football itself. Whilst there has been a noticeable 9% year-on-year fall in average live UK audience per EPL match, this has been driven by a range of factors, and the topline statistic should not be taken in isolation as symptomatic of the death of TV, the Premier League, football, or all three.

More matches = More bad matches

The obvious rebuttal to the fall in audience per match is to point out that the 14 additional televised matches this season must have been allocated to unattractive fixtures, thus lowering the average. This would seem to make sense; Arsenal v Chelsea and other “big” games have always been televised, therefore these new games probably featured the Sunderlands and Burnleys of the world.

However, it’s simply not true. The 10 teams with the lowest average audience in 15/16 were televised a cumulative 111 times, in 16/17 that number fell to 108. Conversely, the number of broadcasts featuring the top 10 most popular teams rose 15%. The most watched team per match this year, Liverpool, were shown live a record 29 times, six more than the previous season. Only Arsenal among the “big six” saw a decline in televised games (two fewer). Overall minutes viewed of live Premier League football fell by 2%, even with the extra matches. Nonetheless, the audience drop off is most stark during the non-premium matches. Audiences for the 10 most watched matches of each season are very similar, in fact the most watched match of 16/17 (Manchester United v Liverpool) attracted a larger viewership than 15/16’s #1 game (Manchester United v Manchester City). It’s a different story at the other end of the scale.

The chart below shows how match audiences over the last two seasons were distributed by decile - the number of matches each season that fall into each 10% step in audience, from the lowest to the highest across both seasons combined:

Two things quickly become clear. Firstly, the majority of EPL matches attract an audience towards the lower end of the scale; there are very few “premium” games. Secondly, 2016/17 has a larger proportion of fixtures in the <30% deciles than 2015/16, similar to the previous graph.

Teams and timeslots

In reality, EPL audiences are a function of not just team identity but broadcasters and timeslots. Sky has more subscribers than BT, and therefore matches typically attract higher audiences on Sky channels. Audiences vary with kick-off time, with Sunday 16:00 games being the most-viewed. Each of these factors interact with each other – it’s far from trivial to know whether Chelsea v Tottenham Hotspur early on Saturday (Sky) would outperform, say Liverpool v Stoke that evening in BT’s timeslot.

This is further influenced by factors external to the EPL – for example Manchester United played just 11 Saturday matches last season (down from 18), because of their Europa League run.

We can, however make a few initial observations of major changes between 15/16 and 16/17, and estimate their impact on TV audiences.

Loss of Newcastle

If Newcastle hadn’t had been relegated, total EPL viewership would have risen season-on-season

The 2015/16 season saw Newcastle United relegated from the Premier League for the first time since 2008/09. They did so having attracted the 9th largest audience per match, and were televised 16 times (8th highest). This is unsurprising; anecdotally the club is one of the best supported in the country. However, these audience numbers are in contrast to other clubs with a similar level of on-pitch performance – Aston Villa and Norwich, the other relegated sides were ranked 13th and 20th in terms of average audiences respectively. Promoted for 16/17 were Burnley, Hull and Middlesbrough, whose TV audiences ranked 11th, 14th and 18th – not terrible but not a replacement for the Magpies.

Let’s pretend there was no relegation in the 15/16 season. If we take the average 15/16 audience for Newcastle, Villa and Norwich, and swap it with those seen for their replacements, 16/17 audiences rise by 5%, and average match audiences would be down just 4.6% from 15/16. If Newcastle hadn’t had been relegated, total EPL viewership would have risen season-on-season. Newcastle will return to the Premier League for 2017/18, with Hull, Middlesbrough and Sunderland (12th) dropping out, suggesting a positive impact on viewing figures to come.

Saturday switch

The main change in the new TV rights cycle from a broadcast perspective was the swap between BT and Sky of the Saturday timeslots; from 16/17 Sky would now televise the early Saturday match, with BT taking the late fixture – the opposite of the previous arrangement. The exact impact of this on TV audiences is difficult to tease out, mainly because of changes to how broadcasters pick which matches to show*. Nonetheless, the overall impact is that Saturday Evening match audiences have fallen by more than Saturday Early match audiences have risen.

Roughly adjusting for this change reduces the fall in average match audiences to 6-7%, not quite as important as the loss of Newcastle but definitely significant.

A deeper approach

Whilst we’ve identified a couple of plausible explanations for the decline in EPL audiences beyond a genuine fall in popularity of football on TV, it’s clear that there are too many interlocking pieces of the puzzle to manually calculate the impact of each. We need a way to codify the drivers of EPL TV audience figures, such that we can assess the two seasons in these terms and either explain the audience decline, or leave some of the decline unexplained and pointing to a genuine drop in consumption or interest.

With this aim in mind, I’ve built a regression model to analyse how these factors contributed to the game by game audiences in the 2015/16 season. The model starts with the timeslots and broadcasters as variables, and introduces each team as a factor stepwise. This means we will only consider the identities of teams that have a significant impact on the audience.

The resulting model has an R2 = 0.71, i.e. it explains 71% of the variance in the data. Below are the teams the included, with the suggested audience uplift from average (beta value):

Note the teams that are here. As well as the big six we have Newcastle, confirming our previous theory that they played a part in driving EPL TV audiences in 15/16. The model’s selection of Leicester City hints at an explanation that we may have missed – their historic run to being crowned champions of England meant that matches in which they appeared attracted audiences to a greater extent than all but three other sides. Finishing a less remarkable 12th last season was surely less of a draw to viewers. Creating a similar model for 16/17 produces the below team betas:

Liverpool’s audience influence increases on the back of a strong on-pitch performance and many televised matches, whereas Arsenal’s drops for the opposite reasons. Overall, the cumulative beta, or total influence of team identities of TV audiences, falls by 18% - with the majority driven by the absence of Newcastle and Leicester.

To test the performance of last season, we can apply our 15/16 model to the 16/17 season, and compare the TV audiences that the model predicts versus those that each game actually received.

The model predicts that accounting for timeslots, broadcasters and team identities (as well as the loss of Newcastle) the total viewership last season should have been 10% higher than it was. This would put the average audience per match 1.2% lower than last season – suggesting a slight disadvantage for 16/17.

So where did last season underperform?

Leicester’s televised matches on average underperformed by 371,000 [viewers in 16/17], almost exactly their uplift in 15/16

Watford v Manchester City and Liverpool v Middlesbrough were the clashing televised fixtures on the last day of the season, and therefore cannibalised each other’s viewership. Noticeably six of the ten worst performing matches involved Leicester City, in line with our prediction. In fact, Leicester’s televised matches on average underperformed by 371,000, almost exactly their uplift in 15/16 – suggesting the team fell completely back into the pack both on the pitch and on the television.

Had Leicester’s heroics come last season instead, the total TV viewership for the Premier League would have been 4.3% higher, with an average per game 5.5% lower than 2015/2016.

Conclusion

Whilst average EPL audiences have fallen year-on-year, a large chunk of it is explainable by the factors outlined above.

Furthermore, when you take away the impacts of the Saturday Switch, the loss of Newcastle and Leicester’s decline, what is left almost certainly doesn’t point to a less popular product. A number of other factors are harder to put a numeric value to, but are much more likely to explain the remainder. Sky has lost subscribers to BT over the last two years - adding only 40,000 subscribers in the three months to the end of March, compared with 70,000 in the same period last year. Other discussions of the matter often cite the “Olympic hangover” effect; a saturation of sports viewership in the summer reducing TV audiences for the following year. Finally, digital streams of matches are estimated to receive audiences of 5-10% of those through traditional TV, through legal channels (Sky Go etc) alone. The UK audience bureau, BARB, has begun what it calls "Project Dovetail", to try and understand the size of online TV viewership, but currently these figures are largely unreported. This drain on traditional TV audiences will continue over time, but this is more of a case of viewers lost to the semantics of measurement, rather than non-sporting distractions.

In all, Sky and BT, who have presumably undertaken a similar analysis, do not have abject reasons to worry about the quality of their £5bn product. One strategic development from Sky for the new season will be the rebranding of their channels; scrapping the 1-5 labels in favour of sports specific monikers; Sky Sports Football, Sky Sports Golf etc. This will allow Sky to offer a range of subscriptions prices, with the cheapest starting at £18pm. It remains to be seen how this will affect EPL viewership, although the premium matches will almost certainly only be available to those paying full whack. Nonetheless, audiences for the less glamorous ties – those driving down average viewership in 2016/17, could see a boost. Look for EPL audiences to rebound somewhat next season, driven by Newcastle’s return, especially if the title race is closer.

---

*Broadcast rights packages are sold with the promise of certain fixture picks for a weekend – e.g. Sky’s Saturday 12.30 package comes with six 1st picks, 14 2nd, and eight 5th picks. A first pick means Sky would get first choice of EPL fixture to broadcast in that timeslot on the given weekend.