Options
UK Viewing Figures Question
PrinceShaun
Posts: 406 Forum Member
✭
So, I was reading the BBC story about how The Queen topped the Christmas Day viewing figures with 7.82 million viewers. In one of the comments somebody was making the argument that the population of the UK is around 60 million, so therefore, even though 7 million people watched The Queen, there were 53 million people (the entire population of England) that did not watch and they were trying to make the point that The Queen is irrelevant.
This strikes me as very strange and not something I had thought about before. Am I right in thinking that the viewing figures are based on households rather the individuals? So if we say the average household is of 5 people then that means 7 million households watched the Queen therefore around 35 million individuals potentially watched?
And also I am guessing that viewing figures are not taken from every single television in the land and that it is only certain households that sign up for it? Kind of like a sample audience?
This strikes me as very strange and not something I had thought about before. Am I right in thinking that the viewing figures are based on households rather the individuals? So if we say the average household is of 5 people then that means 7 million households watched the Queen therefore around 35 million individuals potentially watched?
And also I am guessing that viewing figures are not taken from every single television in the land and that it is only certain households that sign up for it? Kind of like a sample audience?
0
Comments
That's pretty-much it, yes. More info here: http://www.barb.co.uk/resources/barb-facts/how-we-do-what-we-do?_s=4
Basically they use around 5,000 homes which gives figures that are statistically-accurate to +/- 3%.
Despite everything you may hear or read a lot of houses do not have the TV on at Christmas as they believe it's a time for families and the TV is a distraction. A few of my elderly relatives still listen to the Queens Christmas Message on the radio. Also you have to consider not only the UK but the Commonwealth and other UK Nationals not at home such as the Armed Forces. In the UK, many households go out for Christmas dinner whether it is an arranged "do" or to a restaurant or equivelent type of venue.
The point is that you cannot apply the usual criteria to TV watching habits at this time of the year. Whether the Queen's Christmas Message is relevant, and for that matter such messages from any other leader such as The Pope, is a different discussion and cannot be assessed purely on the basis of TV viewing figures which can be picked apart in any arguement.
So the number of people who watched some, but not all of the programme can be substantially higher.
So two people who each watched half the programme will be counted as one viewer.
Commercial channels suffer a lot, as if people leave the room or channel surf during the ads they count as less than one viewer, even though they may have watched the whole programme!!
why is it? Sample sizes for opinion polls rarely go above 1,000. And that's because BARB and polling companies use 'representative samples', not just the first 1,000 or 5,000 people they come across. The sample is picked to reflect very closely the demographics of the entire population. You hear time and again on DS 'yeah but the poll only covered a thousand people and there are 60 million of us so we should dismiss it'. Er, no.
Er, yes. 5000 people is just 0.008% of the UK population. 1 Barb viewer is therefore equivalent to 13000 people watching. It isn't in any way accurate and there's no way you can represent the entire demographic of the UK with such a small sample size.
You clearly don't understand the concept of a representative sample. Of course a sample size of a few thousand can represent the nation if done properly. How do you think exit polls at elections are so accurate? They don't ask millions of people leaving polling stations how they voted, just a representative sample. And if you think the BARB ratings aren't "in any way accurate" I suggest you tell BARB exactly where they're going wrong. And when you've put BARB right, tell the advertising industry who spend millions on the back of ratings information that they haven't a clue what they're doing. Good luck with that!
You clearly don't understand anything about statistics. A representative sample of 5116 households in a population of 24m households provides a margin of error of 2.3 per cent, with a confidence of 99.9%.
To go from 2.3% to 0.5% you would need to increase the sample size 20 times, which clearly provides no cost benefit, because every one is happy with 2.3%.
Just look at forecasts of election results that are done using similar statistical methods. They are usually very accurate - they might get the actual result wrong but the predictions of votes for each candidate or party will be well within the given range of error.
Elections area different beast as they're only asking voting intention on a few parties.
With hundreds of channels a larger sample would be better.
Sky, although a BARB subscriber also do their own ratings using return path data from hundreds of thousands of homes.
They started this when there was concern from multichannels that BARB weren't giving a full enough picture.
Part of the recent carriage deals Sky did with C4 and ITV was access to this valuable data.
Sky's own ratings system is more to get demographic information for those who are watching so they can be sold targeted advertising.
Interestly, regarding the amount of people watching, it tallies with BARB for the most part.
For the major broadcasters though I have no reason to doubt the accuracy.
I thought that with the minor channels they usually say that the figure is below some level because they recognise that they cannot measure it accurately. Advertisers will make their own decision based on the advertising being extremely cheap or specific statistics on particular interest groups by other means.
Exactly.
5% of the population are British Asian, and there are around 50 Asian channels on Sky.
Assuming 5% of the 5,100 BARB boxes are in British Asian households, that's only around 250 homes being monitored. And no doubt not all of those watch/subscribe to the Asian channels.
This information is far less accurate though as it doesn't know:
A) How many people are watching or their age or demographic information
b) anything about viewing in households where there is no return path
c) what people are watching not through their Sky box
D) whether the TV is even switched on
Sky's data is only really useful when it comes to time shifted viewing habits. For anything else the BARB data is more accurate and more reliable.
There was a good TV movie by Danny Devito about how easy it could be to "fix" the ratings.
He discovers that the most "floating" TV viewers carry most weight with the ratings company (a bit like political parties in the UK obsessing over floating voters).
He arranges for these families to "win" a round-the-world cruise, and while they are away his mobster friends sit in their houses watching the TV shows he tells them to.
If the sample selected is actually a true random sample - which means that every single household in the whole country had an exactly equal chance of being chosen, then 5000 is q
If the sample selected is actually a true random sample - which means that every single household in the whole country had an exactly equal chance of being chosen, then 5000 is quite a large sample size actually.
Look at it this way. Suppose we think of every coin toss that will happen from now until doomsday. That is billions of billions.
Select a sample of 6 such tosses and it is fairly likely that the number of heads and tails will not be close. But when you take, say, 100 tosses, the chance that heads and tails differ by say 10% is rather low. If the sample size gets to 1000 then the chance that the heads and tails differ by 10% is very low. It can be worked out what the odds are. That is a small proportion of the total. The accuracy actually depends on the sample size, not the population size.
http://www.dummies.com/how-to/content/how-sample-size-affects-the-margin-of-error.html
That's fine for a yes or no or tossing a coin, but how can for instance a county like Cornwall have only around forty BARB homes to measure the viewing of 200+ channels across five platforms over two sexes, five age groups and four social classes?
How many 65+, c1, male viewers with Virgin Media are there likely to be found in forty Cornish homes?
It doesn't matter how many there are as long as the number is representative of the population that fills the same criteria.
BARB doesn't measure areas as small as Cornwall anyway - the closest you would get is the ITV Westcountry West subregion, which includes Devon and a bit of Somerset too.
It also doesn't measure C1s, they would be lumped in with As and Bs.
And I am pretty sure, though don't know for certain, but I don't believe they measure platforms separately: Only pay TV or non-pay TV homes.
So the number of 65+ ABC1 men with pay TV in the subregion will be pretty accurately measured.
In a sample survey, the accuracy (strictly speaking, that is called the standard error) of a figure depends, as I said above, on the number of cases in the sample. So if a sample for judging how many watched BBC1 or ITV is say 1000, then the standard error will be very low i.e. the figure is extremely likely to be accurate.
But, for the case of say the numbers watching Sky Arts 2 in Essex, the sample size is likely to be small, say 13. As an estimate of the correct actual number watching Sky Arts 2 in Essex, you are quite right, it will be nearly useless.
However, it is also true that the results tells us it is extremely likely that very few Essex viewers watched Sky Arts 2 - we are unsure whether it was 1300 or 500 or 2500 but it was low.
That is, if the sample size is low, the figure can be taken but can not be considered accurate to a few per cent, but might be perhaps 50% out, either way.
It is the same with opinion polls. The figure for Labour will be only a few per cent out, while the figure for the Greens might be a much larger percentage wrong. But be careful about that because although the Green vote might be out percentagewise, in actual numbers it is still only a small number out.