Monday, March 17, 2014

The Adults 18-34 Primer, Part 1: Fundamentals of 18-34 Ratings

Coverage of TV ratings has evolved a lot. If you look back at the trade coverage circa 2000, household ratings were the day-to-day currency of choice, even though the industry had changed to a young demo focus decades earlier. Demos soon became a part of the picture, but even just a few years ago, almost all media seemingly weighted total viewership and adults 18-49 ratings pretty much equally. Now, at least in the more advanced circles, it's almost solely an 18-49 world. We know that the correlation with advertising money is massively stronger using 18-49 ratings rather than total viewers. If you're in the cultural popularity business, go with viewership, but 18-49 ratings are far more telling about success and how the industry works.

But it's become clear that there are predictable tendencies even deeper than that single linear relationship. The correlation with 18-49 ratings is good, but beyond that... even younger is better. We see this every year when I line up the ad rates and the ratings; shows like The Simpsons and New Girl skew extraordinarily young and are also the shows that get the most advertising revenue per adults 18-49 rating point. And this ad rates relationship translates to real decision-making. Just last week, the very young-skewing The Mindy Project got early renewed with borderline flop 18-49 numbers.

So today, a three-part investigation into the even younger demo of adults 18-34 begins. Today, we'll use ad rates to lay out the reasons for actually delving into this stuff, and then we'll make some big picture comparisons between the 18-34 and 18-49 demos. The last two posts will then be much more about individual shows.

Longtime friend of the blog Douglas Pucci (SonOfTheBronx) deserves much of the credit for this project, as his finals posts at TV Media Insights are the source of all the 18-34 numbers included here. Though I don't (currently) have any of these 18-34 ratings in my archives, a browse of that forum can get you everything that I'm using here.

Is 18-34 Worth Looking At?

To start, we need to lay out the reasons for actually writing these posts. If 18-34 can't demonstrate some kind of relationship with ad rates, there's no need to do any of this.

Here's what really opened my eyes. Using the three months or so of 18-34 numbers I have from the 2012-13 season, not only did 18-34 demonstrate a much stronger linear correlation with ad rates than total viewers, but it was actually even a bit better than the so-called gold standard, 18-49 ratings!

AdAge Adweek Variety
Linear Correlation (r) w/P2+ 0.561 0.534 0.577
Linear Correlation w/A18-49* 0.939 0.914 0.944
Linear Correlation w/A18-34 0.955 0.935 0.952

*- You might notice these numbers are ever-so-slightly different from the ones in the survey of surveys from a few months back; this is because I excluded the seven shows out of that post for which I didn't have any 18-34 numbers.

So it's possible that 18-34 ratings aren't just a shading that should be added to 18-49. Maybe they should be the standard. It's worth following going forward. But the world can't and shouldn't be overhauled in a day. First, the A18-34 ratings aren't nearly as quickly and widely reported, so in terms of coverage, that's a major obstacle. And second, what I did have from 2012-13 was only about a three-month chunk of the season. So it's possible this chunk was a small sample that happened to be heavily favorable to the 18-34 correlation for whatever reason, and it'd be worse if I had the whole season.

Still, it's probably pretty safe to conclude that 18-34 ratings are of some value, as The Mindy Project can attest. But we won't go much farther than that outside of the bounds of this 18-34 series of posts... at least for now.

But here's another exciting revelation from these numbers: it's also possible that it's not so simple as, "18-34 is strictly better than 18-49, so let's throw out 18-49." In fact, the best reality might be somewhere in the middle, with both numbers adding value. I approached this in literally the simplest way possible: I just added two crude numbers (A18-34 & A18-49) together to get one somewhat less crude number. It's kind of the TV ratings equivalent of the baseball statistic "OPS," which adds together two numbers (on-base percentage and slugging percentage) to get something that's in the middle.

I'll call this A18-34 + A18-49 number TPF, or adults eighteen-to-Thirty-four rating Plus adults eighteen-to-Forty-nine rating.

Here's a comparison of the linear correlations among A18-49, A18-34, and the sum of the two (TPF).

AdAge Adweek Variety
Linear Correlation (r) w/P2+ 0.561 0.534 0.577
Linear Correlation w/A18-49 0.939 0.914 0.944
Linear Correlation w/A18-34 0.955 0.935 0.952
Linear Correlation w/TPF 0.967 0.943 0.968

The TPF correlation is extremely good; in fact, it's even a bit stronger than either A18-49 or A18-34 individually! So it seems possible that some kind of whole may be even greater than the sum of its parts (even if it is literally the sum of its parts). Again, it would be jumping the gun to make massive changes based on this. But if these relationships hold up with next year's data, there may be some real added value in using a TPF or some other fusion of the two numbers.

In the rest of this post and all of the next one, we will step back and introduce the differences between 18-34 and 18-49 as separate ratings. But in the third and final one, we'll return to this crude combination of the two and see if it has anything new to say about this season's marginal shows.

The Big Picture of 18-34 vs. 18-49

Why has this blog been so strictly adherent to Live + SD adults 18-49 ratings over the years, even if it's been clear there are some ways to supplement it? It's pretty simple: it's because all the numbers are out there. When everything is out there, we can develop a rich understanding of what things mean. We know roughly how many L+SD adults 18-49 are watching TV, how many are watching broadcast TV, what a league average entertainment show is, what a hit is and what a flop is. So many other demos and streams tend to hit the presses only when they're noteworthy enough for their networks to point out in press releases. But since these numbers are inconsistently available, it's tough to impossible to put them in that larger context. I see the numbers, but I don't really know what a 2.6 W25-54 rating means, or what a 3.0 T12-17 rating means, or what a 60% DVR lift means. Maybe this is just a me problem, and I'm just not reading and processing the info in these releases enough. But my guess is that it's this way for almost everyone.

If A18-34 is going to become a thing, we're going to have to try to develop the same kind of understanding there that we have in A18-49. So this post will try to lay out some of those same basics.

There's one fundamental point that must be made to start off. If you take away nothing else from this whole post, take away this: not all ratings are created equal.

For example, if you read that an entertainment show got a 2.2 A18-49 rating and a 2.0 A18-34 rating, your gut reaction might reasonably be, "That show is a little weaker in A18-34 than in A18-49." Technically, this is correct; the percentage of the total TV-owning A18-34 population watching was in fact smaller than the percentage of the total TV-owning A18-49 population.

But these are different two different types of audiences. One is much more inclined to watch TV than the other. So grading other ratings by the A18-49 standard will inevitably lead to some misconceptions about strength. To really be able to put relative values on the A18-34 rating, we need to know about its own tendencies.

Here are comparisons of overall viewing and broadcast viewing. Since I don't have the numbers for the whole season, rather than a season-to-date number I just compare them week-by-week. And I'm only using only the weeks when I have all of the 18-34 ratings/shares for the entire week. (Several of the missing November/December weeks I have six out of seven days.)

Week Ending 18-49PUT 18-34PUT 34/49* 18-49bc 18-34bc 34/49*
9 11/24/2013 34.9 30.3 87% 8.2 6.3 77%
12 12/15/2013 34.2 29.6 87% 6.9 5.0 73%
13 12/22/2013 32.1 26.9 84% 5.8 4.2 72%
16 1/12/2014 35.8 30.8 86% 7.7 5.9 77%
17 1/19/2014 34.0 29.9 88% 7.6 5.9 77%
18 1/26/2014 35.4 31.0 88% 7.6 5.8 77%
19 2/2/2014 36.8 31.3 85% 10.1 8.4 83%
20 2/9/2014 35.9 30.3 84% 9.3 6.8 73%
21 2/16/2014 34.4 28.9 84% 8.6 6.2 72%
22 2/23/2014 33.6 28.2 84% 7.7 5.2 67%
23 3/2/2014 33.8 28.9 86% 8.0 6.1 77%
24 3/9/2014 32.8 28.2 86% 6.5 4.6 70%

AVERAGES: 34.7 29.7 86% 8.1 6.1 75%

The biggest, most basic takeaway here is that 18-34 is the scarcer, harder to reach category. The average adult 18-34 watches about 86% as much primetime TV as the average adult 18-49. In these particular twelve weeks, about 35% of the whole 18-49 demo is watching TV in the average primetime moment, but only about 30% of the 18-34 subset.

Not only is the 18-34 demo less inclined to watch TV, but they're also even less inclined to watch broadcast TV. An average member of the 18-34 demo watches the big broadcasters only three-fourths as much as an average member of the broader 18-49 demo.

*- Just a basic semantics note: This percentage is derived simply by dividing the 18-34 number by the 18-49 number, because this is the most intuitive way of comparing the two numbers. I should note that it's incorrect to translate these into saying something like, "there are ___ % as many 18-34 viewers as 18-49 viewers." The numbers describe rates of viewing, not total viewership. The 18-34 population (67.63 million) is about 53% as big as the 18-49 population (126.96 million). So when we say "adults 18-34 are watching 85% as often as adults 18-49," to get an actual viewership comparison we have to multiply that by the 53% ratio. "Adults 18-34 are watching 85% as often" = "46% as many adults 18-34 as adults 18-49 are watching."

These comparisons are informative about the 18-34 demo's tendency to watch TV. But in terms of our main purpose (grading entertainment series), these numbers include a lot of noise. Namely, the broadcast viewing 18-34 ratio spikes for big events like football season, especially the Super Bowl (week 19) and the Oscars (Week 23). These events are big enough to tangibly affect numbers for the whole week, and they tend to have a better-than-average ratio of 18-34 viewers. So the best metric for most of our purposes is their tendency to watch only original non-sports series. So here's a look at the big four original non-sports series "league average" (the number used for A18-49+) both in A18-49 and A18-34 across these weeks.

Week Ending 18-49LeAv 18-34LeAv 34/49
9 11/24/2013 1.90 1.34 71%
12 12/15/2013 1.83 1.28 70%
13 12/22/2013 1.93 1.38 71%
16 1/12/2014 1.76 1.29 74%
17 1/19/2014 1.86 1.40 75%
18 1/26/2014 1.78 1.31 74%
19 2/2/2014 2.08 1.48 71%
20 2/9/2014 2.06 1.40 68%
21 2/16/2014 1.72 1.31 76%
22 2/23/2014 1.80 1.18 65%
23 3/2/2014 2.02 1.45 72%
24 3/9/2014 1.89 1.31 69%

AVERAGES: 1.89 1.35 71%

Week-by-week, the average moment of series programming has somewhere from 65% to 76% the 18-34 rating compared to 18-49 rating. (And the range is tighter - 68% to 75% - if you throw out the two weeks that had very few originals against the Olympics.) The average of the 12 full weeks is 71%. So among entertainment series, there are only two-thirds to three-fourths as many A18-34 ratings points floating around.

Returning to our old example of the series with a 2.2 A18-49 rating and a 2.0 A18-34 rating: if you wanted a show that was the same relative to the league average in both numbers, apply the 71% ratio and that 2.2 A18-49 rating would translate to about a 1.6 A18-34 rating. So the 2.0 A18-34 is a decidedly stronger number in its landscape than the 2.2 A18-49 is in its landscape.

We don't have A18-34 numbers for the whole season, so in the next post (looking at specific shows), instead of directly calculating full season averages for A18-34, we will just apply that 71% ratio to the full season of A18-49 info. It's not ideal, but I think it's a bit better than comparing full seasons and partial seasons (or chopping off all the A18-49 numbers when I don't have A18-34).


One last big picture note. This whole project was inspired by the takeaway that comes out of ad rates articles every year: strong 18-34 shows like The Simpsons and New Girl always score vastly more ad money than their L+SD A18-49 ratings say they "should."

Another important thing to note is that these two shows, while each strong in 18-34, are quite different in terms of gender skew. In fact, they're almost complete opposites. A quick spotlight: the last time each show hit a 1.9 A18-34 rating, New Girl had a 2.4 in W18-34 (implying roughly a 1.4 M18-34), while The Simpsons had a 1.4 in W18-34 (implying roughly a 2.4 M18-34).

So while these 18-49 overachievers share huge 18-34 ratings, the similarities in audience pretty much end there. There are two potential ways to read this:

1) There is not a strong preference toward either gender, so the generic adults 18-34 rating is a good catch-all.


2) Ad rates shouldn't be defined in terms of one linear relationship. These shows each overachieve because of their incredible reach in their individual demos.

At least for now, I'm operating under #1. Two reasons: a) It's a lot less work; and b) The correlations with generic A18-34 and the hybrid (TPF) are pretty damn strong anyway. But I'm certainly not sure that it's the best way to go. Another thing to note in #2's favor is that New Girl and The Simpsons don't just get the most $ per 18-49; they're also among the biggest outliers on the $ per A18-34 chart (though by much smaller margins). They may do this because of their extreme scores in their respective genders. This might warrant further investigation down the line, but just doing the generic A18-34 stuff is more than enough of an effort at this point.

Coming up: Part 2 compares the A18-49 and A18-34 ratings of some of TV's top shows. Who overachieves and underachieves if we were using A18-34 rather than A18-49? And Part 3 will combine the two numbers and see if the bubble landscapes are changed with a new kind of True Power Rankings.


Spot said...

Bravo, Spot. That's one fascinating and intelligently written analysis of the ratings demographics. I've always wondered how the variables of age, gender and income are weighted in setting ad rates. I look forward to your further analysis. My guess is that the 18-34 demo will eventually become the gold standard w/r/t importance in determining ad rates and consequently whether a show lives or dies. Will a 35-49 demo become the silver standard? And the broader 18-49, bronze, followed by the 25-54 in fourth?

Spot said...

New Girl 1/21/2014
A18-49 1.9 , A18-34 1.9 , W18-34 2.4, M18-34 around 1.4.
Fox is surely selling that 2.4 W18-34. Because they'd be dumb to sell 1.9 in A18-34 (or in A18-49), that's half a point less of eyeballs.
They would sell that 1.9 if there would be no buyers for W18-34. But we know there are buyers! How? From your great posts "lining up the ad rates and the ratings".

Ditto The Simpsons, but M18-34. In short, generic adults 18-34 rating is *not* a good catch-all.

Next, majority (I'd say more than 80%) of 5 broadcasters' shows are selling A18-49. So one can not simply replace A18-49 with A18-34. Each show has its target demo, and only ratings in that demo are relevant for show's success.
For example, The Vampire Diaries, The Originals, and Reign all target W18-34, that's their all-important demo. But Supernatural, Arrow and The Tomorrow People target A18-49, and CW doesn't give a rat's ass what those are doing in W18-34. Or A18-34.

About "they're also among the biggest outliers on the $ per A18-34 chart" :
It might be caused by Fox network getting better CPM than other broadcasters. But probably you calculated in.

Spot said...

I have a feeling M18-34 are more valuable than W18-34. Due to their somewhat higher scarcity.
And I think that for any show, generally someone 18-34 is more valuable than someone 35-49, but there's enough importance in factors like gender/income that it's not always the case. Plus, I'm guessing that, say, a 34 year old is not way hotter property than a 35 year old.

Spot said...

I'm loving that proposed TPF metric (and its indirect baseball origins - Nate Silver would be proud).

I can't wait to see how this shakes out. I'm sure there'll always be mysteries and complexities simply because, as Silvio implies, different advertisers seek different demos, but this is the most important step to take in being able to do what drives most traffic for the tellymetrics sites like this one - make more informed guesses on which shows are(n't) getting renewed. It's pretty obvious that Mindy is much more able to escape the bear than 18-49 can capture, and if we have the numbers to reflect this...

Post a Comment

Newer Post Older Post Home
© 2009-2016. All Rights Reserved.