Channels & Conflict: Response to Digital Media Distribution, Impact on Sales and Internet Piracy

Uploaded by GoogleTechTalks on 01.09.2010

It's my great pleasure to introduce Michael D. Smith. If your name is Michael Smith, that
D is very important. And this will allow you to Google him if you want to get a copy of
the paper. Shall I zoom this up on your… >> SMITH: Zoom this up, yeah.
>> …website? I've known Michael for, what, seven years maybe, six or seven years. We
met when he was right out of grad school and he's done some fantastic work in this area
of marketing, online activity, intellectual property and a number of other areas of that
sort. So, welcome to Google. >> SMITH: All right. Thank you, Hal. That's
a wonderful introduction. It's a wonderful opportunity to be here. I'm really excited
to talk to you all. And I should acknowledge right up front, thank Google and WPP for providing
funding for this stream of research that we're doing. So the talk I'm going to give is entitled,
"Channels and Conflict: Consumer Response to Digital Distribution Channels." This is
joint work with my colleagues, Rahul Telang, who is co-director with me of the Center for
Digital Media Research at Carnegie Mellon, and our very fine doctoral students, Brett
Danaher, Samita Dhanasobhon, and Anuj Kumar. What I'm going to try to do is present three
papers on the same theme, okay? And so the theme is sort of motivated in part by--you
know, if you can think back to the heady days of 1998 and the speculation about what was
going to happen when you added Internet markets. So there are a lot of wonderful quotes. One
of them, this nice quote in Business Week that says, "The Internet's a nearly perfect
market. Information is instantaneous and the result is fierce price competition, dwindling
product differentiation, and vanishing brand loyalty." And that was sort of part of the
conventional wisdom in 1998, untested by data. And it turns out if you go out and you look
at the data for what actually happened in Internet markets, not all of these hypotheses
hold up. There's a lot of price dispersion, brand matters a whole lot, and if you don't
believe me, you can Google that and, you know, and product differentiation still matters.
What we wonder is whether we're in a very similar time right now with respect to digital
media in the sense that there's a lot of conventional wisdom about how consumers will approach digital
media channels and very little of it is tested rigorously and empirically. So here are two
representative quotes; one from James Gianopulos, chairman of 20th Century Fox, "When people
say, 'Reinvent your business model because of the ubiquitous availability of piracy,'
there's a huge flaw in that. You can never compete with free. That's an economic paradigm
that doesn't work," okay? On the flip side of this is Steve Jobs by no means a disinterested
observer in this debate, but Steve Jobs saying, "You'll never stop piracy. What you have to
do is compete with it." Okay? So this interesting question, "How do I compete with free in general,
and in particular how do I compete with free pirated content? Can I ever get those consumers
back? And if I can, what does that look like?" The second debate is, in my mind, almost as
interesting if not more interesting, the question about, "If I add a digital distribution channel,
what is that going to do to consumption in my existing physical channels?" And here,
I think the conventional wisdom is in the industry that when I add a digital distribution
channel it's going to primarily cannibalize my existing physical channel. So a set of
quotes; Jeff Zucker, CEO of NBC University--Universal: "Our challenge is to monetize these digital
channels. We don't end up--want to end up trading analog dollars for digital pennies,"
okay? Another one from Netflix, also not a disinterested observer, but saying, "When
studios distribute through Amazon Unbox, they're cannibalizing--they're within the DVD window
and they're cannibalizing their own sales." And the last one, unfortunately, I'm not able
to get a good quote on this for soon-to-be obvious reasons, but apparently when Disney
started selling their movies on iTunes, Wal-Mart which makes up something like 35% to 40% of
their sales was so annoyed by this that they sent back crates and crates of Disney DVDs
in protest. Target did a very similar thing; took down Disney encaps and replaced them
with their competitors. And my favorite piece of this is if you believe the press reports,
Wal-Mart sent out emissaries to Hollywood threatening to retaliate if any other studios
followed Disney into the iTunes store, okay? I don't know if horse heads were involved,
but nonetheless, it's a rather strong response. Possibly, yes, possibly unicorns. So these
are the research questions we're interested, "Can you compete with free? And what does
that look like? And then what's the dominant impact of adding these digital distribution
channels? Does it cannibalize your physical sales or might it have an impact on piracy?"
And not to spoil the punch line for you but what I'm hoping to convince you is that, number
one, you can compete with free. Distributing content for free actually can stimulate sales
of paid product for the same content. And number two is that the dominant impact of
adding a paid digital distribution channel seems to be reducing the demand for piracy
and having no effect on existing DVD sales, okay? And I'll try to--I'll try to get through
the talk and leave some time at the end where we can interact on these questions and I hope
we can have a healthy, fun discussion around these, okay? This contributes to a couple
of different academic literatures. I won't spend a whole lot of time on these. The first
one is there've been a set of very interesting papers looking at the impact of, primarily,
music piracy on sales of CDs, primarily. And typically, these papers I'm going to oversimplify
them but typically, these papers take Napster as an event and look at what the introduction
of Napster did to sales of CDs. And these papers acknowledged the important question
which is, you know, "When they find that Napster resulted in a reduced--a reduction in CD sales,
is that because Napster was free or is that because Napster was digital?" That is, "Is
that because I wanted to get the content for free," or is that because, "Napster was just
so darn easy for me to use compared to going down to BestBuy or, you know, HMV or what
have you and browsing through the sales of--browsing through the racks of physical CDs?" There've
been a couple other studies on video. The one I want to highlight partially selfishly
and partially topically is a paper Rahul and I wrote where we looked at what happens to
DVD sales around the time a movie is broadcast on television. And what we found is that when
a movie is broadcast on television, there's almost an immediate uptick in demand for DVDs
on Amazon and in demand for piracy, that that the availability of content, we sort of used
that as an exogenous shock and then say that the availability of content on piracy whether
the movie is available or not doesn't seem to affect the uptick in DVD sales, okay? Which
kind of had us scratching our heads because we really do believe piracy hurts. But here
we were seeing, you know, here we were seeing something that sort of, I don't know, challenged
that view where the presence of piracy on the digital channel wasn't affecting DVD sales.
And so what we wanted to come back and do is try to gather data that would help us explore
that question in a little bit more detail, so. And the way we're going to do this is
with exogenous shocks in the market. So there's some classic challenges associated with measuring
the impact of piracy. You know, if I regress demand for piracy on demand for sales, you
have an obvious endogeneity problem. How do I break that endogeneity? Okay, so let me--let
me talk about the data sets that we used for the first two studies. So for the first two
studies, we went out and gathered data. The first place we gathered it was
And I had a very bright doctoral student who went in and realized that if you indexed the
Amazon top 100 selling DVDs and added 200 in place of 100 in the URL, you could sort
of scrape the whole site. What was fun is that 48 hours later, that ability went away.
So somebody was watching, but fortunately, she had all the SIN. So she has been scraping
every DVD sold by Amazon on a daily basis; collecting the price, you know, characteristics
of the DVD, and importantly, the DVD sales rank, okay? So we're going to use that sales
rank as a signal of changes in absolute sales. So it turns out if you assume a Pareto relationship
between the ordinal sales rank and the cardinal number of sales that you can--that you can
impute a change in sales rank as a percent change in sales. We have that for all DVDs
at Amazon from November 2006, hopefully, through yesterday if things are going well. And, you
know, these numbers seem so much more impressive when we present them to academic audiences.
They're much less impressive when you present them here at Google. So you'll forgive me
for my humble data count here. The flip side of this is how do you measure piracy? So we
scratched our heads for a long time about how do you get a reliable measure of, you
know, a reliable signal of piracy demand. And we finally decided to go to the oracle
of all piracy knowledge. I asked one of my undergrad classes how they steal movies. And
they cheerfully answered back that at least at that time the primary tool they use was
BitTorrent. So we found the most popular BitTorrent tracker site, Mininova, at the time, indexed
all of Mininova's television content, movies, and videogames, and that we're going to use
television content for this. So we have a daily scrape--and, again, this sounds much
more impressive in other context, but we have a daily scrape of every torrent posted to
Mininova including the number of seeders, number of leechers, and cumulative number
of downloads. So we're going to--we're going to use that. We're going to use primarily
the downloads as a signal of changes in demand for these pieces of content. We're going to
go through a fun exercise of trying to code the names of the torrents. Sadly, pirates
aren't all that disciplined about adding good metadata to their torrent names, but we're
going to go through an exercise of trying to code these back to the right pieces of
content and we're going to use that. We've got data from November 2007 up through last
November when Mininova started filtering their content. And, again, we've got what we believe
to be a fair number of observations, at least for somebody working for pizza and on an old
computer. Okay, so I totally--let's start gathering this and let's just wait for fun
things to happen in the market, all right? So the fun thing number one was NBC kindly,
presumably to support the academic research endeavor, decided to pull all of their content
from Amazon, so--or from Apple. So NBC got into a fight with Apple about a variety of
things; some of them revenue, some of them pricing, and some of them content protection.
But the upshot is that in--on December 1st of 2007, they removed all of their television
content from Apple iTunes and restored it later in September of '08. I'm going to focus
on the December event. We can talk about what happened in September, okay? So this gives
us the nice sort of event that we want to have. We've got a set of consumers cheerfully
buying iTunes content from NBC, and on December 1st, they can't anymore. Where do they go,
okay? So we're going to extract again the piracy in DVD data and we're going to use
this as a sort of a difference-in-difference or control treatment group model. We're going
to use--we're going to try to measure NBC--changes in NBC's piracy and changes in NBC's DVD sales
with ABC, CBS, and Fox as the control group, okay? All right. What I need to convince you
though is that ABC and CBS piracy looks like NBC's piracy in the before period. That is
ABC, CBS, and Fox are a decent control group for NBC in the before period for this to work.
So let me see if I can convince you of that with a pre-treatment falsification test. The
model we've got is the log of downloads on to date dummies, episode dummies, and then
an NBC indicator for the actual event. And what I'm hoping to find is that in the period
before NBC took their content off of iTunes that that dummy variable for NBC is insignificant
and in the period after it, it's significant. And you already know the answer to this question,
this is sort of the chart of the conditional values and what you find is that NBC seems
to map non-NBC pretty well in the period right up to December 1st. And then after that, we
see that NBC piracy levels are higher than ABC, CBS, and Fox, okay? In percentage terms,
we can run this and just stare at the coefficients, we interpret these coefficients as percentages.
And what this is saying is that the demand for NBC piracy increased by about 11% in the
period after they removed their content. And I should point out this is--we have data up
through Christmas. I'd love to tell a wonderful story about how we decided to exclude the
Christmas holiday period. The fact of the matter is my very talented doctoral student
went home for Christmas over the holidays and right about Christmas, Mininova changed
the way they indexed their pages and so I'm not entirely sure anymore what happened after
Christmas. But I can say, you know, with a reasonable amount of confidence that up until
Christmas this was an 11% increase. The other interesting thing going on here is this variable
says what happened to non-NBC piracy during the same time period. It's statistically,
you know, marginally significant and it's not well-identified. I don't know what would
have happened had NBC not done this. But if you just--if you just interpret that on the
face of it what it's saying is, not only was there an 11% increase in NBC piracy but there
was an additional five to six percent increase in non-NBC piracy. Which might mean that these
consumers who were buying on iTunes and decided to go pirate their content through BitTorrent
not only pirated the NBC content but also pirated some ABC, CBS, and Fox content while
they were there, okay? Does that make sense? All right. Oh, I'm sorry--go ahead.
>> Just one clarification question for the thoughts on CBS data, they're all--they were
all on iTunes the whole time? >> SMITH: They were all on iTunes during the
whole time, so it's a proper control group, yeah.
>> And they are available on the computer [INDISTINCT]
>> The nice thing about this is Hulu was still in private beta during this time frame so
the number of Hulu users was relatively small and there was no big changes in distribution--digital
distribution for any of these folks. So a good question. So the other thing you can
do with this data is you can carve it up by genre. So if you wanted to characterize the
typical pirate as a young male, this data would be consistent with that view. So what
we see is a much bigger uptick in demand for pirated content for sci-fi shows, comedy,
and to a lesser degree, action, than for drama, okay? I don't know--I don't want to push this
too hard but you could draw your own conclusions from this. So then we can go and do exactly
the same test on DVD sales at Amazon. So here the dependent variable is the natural log
of Amazon sales rank. And we're measuring sales of NBC DVD box sets for television versus
ABC, CBS, and Fox DVD box sets for television. And what you find is no statistically interesting
change in sales for NBC content relative to ABS, CBS, and Fox, okay? So you scratch your
head for a little while about this. Second event is ABC and Hulu, okay? So here we have
NBC, CBS, and Fox are showing programs on Hulu. Disney announces that they're going
to add there content to Hulu. And we can do basically the same thing just in--just in
reverse, all right? So May 1st of 2009, ABC announces they're going to add their content.
July 6th, ABC starts streaming their TV shows to Hulu. We're going to use the same difference-in-difference
model. We have eight ABC shows and 42 non-ABC shows, and these are just the shows after
eliminating categories where that, you know, that might be unusual. So game shows, reality
shows, talk shows, these are sort of the, you know, might be systematically different.
And we're going to look at a two-week period before and after July 6th of 2009, okay? So
here again, we're going to interact the ABC variable with the after variable and what
you find is about a 37% decrease in piracy in the two-week period after ABC adds their
content to Hulu, which is larger than I was expecting but it's statistically significant.
I've asked Samita to beat this up in a variety of different ways and it just doesn't go away.
So a huge decrease in the demand for piracy. And then if you look at DVD sales, you know,
nothing statistically significant. If anything, it's directionally negative. Which again,
because these are sales rank, a decrease in sales rank actually means an increase in sales.
So if anything, there's a slight increase in sales but not in any statistically significant
way, okay? Yes? >> I would like to [INDISTINCT]. In this case,
[INDISTINCT]. >> SMITH: Yeah. So this is--this is--so, as
I understand the question, what you're saying is we're looking at the two different contexts.
iTunes is one where I pay for the content and I get it, whereas Hulu is one where I
get it for free or free with commercials, so they are different contexts. Yeah, and
I'd buy that completely. And that's part of the reason I think, you know, it's kind of
interesting that the DVD sales changes directionally positive, right? I think broadcasting--so
if you go back to what I said about the paper Rahul and I did with what does television
do; what's the dominant impact of broadcasting the whole movie on television? It's to encourage
people to go out and buy the DVD. Here we've got a setting where you put it up on Hulu,
I can get the whole piece of content and it might be encouraging some people, although
not a statistically significant number of them, but it might be encouraging some people
to become interested in that content and possibly buy it on DVD. Yeah?
>> But could you make sales for exactly the same episode or same season or?
>> SMITH: Exactly the same series, yeah. So, typically, the DVDs are released in season
long so it was exactly the same series and the same seasons that were being put up there.
Good question. Any other questions on this? >> For that series, was it the most recent
one or you were looking at the previous series? >> SMITH: We were looking at whatever they
were putting up. So that's a good point. I have to ask Samita. I would suspect what she
did was a match set, because I know that's what we did in the NBC study, a match set
of DVDs and whatever seed, but they had to be available on both sides. Whatever DVD was
available matched with the season that was also available. I think that's right. Yeah?
>> Well, were you able to look at [INDISTINCT], were you able to look at the whether there
was impact on ratings of the actual broadcast in terms of movies you're concerned of?
>> SMITH: I haven't done that. I would love--I would love to do that, right? There's a bunch
of anecdotal evidence and speculation that, you know, shows like, "It's Always Sunny in
Philadelphia" gained a following through Hulu and then became popular on broadcast. It's
a natural spillover. I just haven't found the data to identify it. It's a cool question.
Any other questions on this? All right. So the summary here; ABC's availability on Hulu
is correlated with a 37% decrease in piracy and no change in DVD sales. So we see a similar
effect here to what we saw before. Let me--and what I'd like to do is spend the last half
hour focusing on the last paper. This last paper is the most recent so this is very much
work in progress. So I'm hoping we can crowd-source some, you know, some other cool analysis we
should be doing on this. And this is also the paper that Google, and Hal, and WPP are
supporting, okay? So we found a studio; let's call them studio X for a moment. And studio
X has a problem that all studios do. When they license their movies to HBO, HBO asks
for an exclusive broadcast license, which means that you can't also show it to other
competing channels but it also means that you've got to take it off your other digital
distribution channels. So studio X has to take it off Video on Demand and they have
to take it off of iTunes. So we have a very similar structure. We have, you know, movie
enters the theatrical window. About 28 weeks later, it goes into the DVD window and it's
typically released DVD, iTunes, and Video on Demand all at the same time. And then when
it enters the HBO window, they have to take it off of Video on Demand and EST. It stays
up on DVD primarily because there's just no way to get it out of the channel once it's--once
it's in the physical channel. And it stays up on piracy, obviously, pretty much through
this whole period. Now, the first--the first challenge we had was how do you isolate the
window effect from the broadcast effect? Again, this previous paper suggests when you broadcast
a movie, DVD demand increases. The fun thing about this is that when you--the contract
with HBO starts on the first of the month that the movie is broadcast on HBO and the
first broadcast of the month is almost always the first, second, or third Saturday of the
month. And so we're going to use that variation between the first of the month and the first,
second, or third Saturday in the month as a way to identify the window effect from the
broadcast effect, if that makes sense. Okay, all right. So very similar questions, "Where
do these customers go?" Except here, we get to look at hopefully both former Video on
Demand customers and former iTunes customers separately. So where do former Video on Demand
customers go when they can't get on Video on Demand anymore? And where do former iTunes
customers go when they can't get it on iTunes anymore? And also, you know, what does this--what
does the broadcast do as well, another example of competing with free, okay? All right. The
data we got from studio X was their DVD, VOD, and EST sales for ten movies that entered
the HBO blackout window from December '09 through March of '10. We have each of these
data add a DMA level for the top 153 DMAs in the United States. DMAs are sort of regions
of the country. Is everybody familiar with--okay. And we have the--we have weekly observations
of peer-to-peer activity through one of their vendors, also broken out by DMA based on the
IP number of the person committing the piracy, okay? All right. So we're going to use the
HBO window--entering the HBO window as an exogenous shock to demand for both VOD and
iTunes, okay? So first thing I need to convince you of is that entering this window is exogenous
as news we can tell it is. So talking to studio X, they negotiate with HBO sort of collectively
for all their movies and determine a period of time after theatrical where it's going
to enter the HBO window and that's set before they ever see what the DVD demand is, okay?
It's also comforting that when you go and look at the lag between when it leaves the
theater and when it enters the HBO window there's no systematic difference in that lag
as a function of either DVD sales or theatrical sales. This really does look like an exogenous
shock, okay? And as I mentioned before, we're going to try to exploit the lag between the
introduction of the window and the broadcast to try to isolate out both the broadcast effect
and the window effect--or broadcast effect, the entering the window effect from the broadcast
effect. Does that make sense? Okay. If you do this--so the simple model for this shows
that there's an increase in both DVD and piracy demand when it enters the window, although
not particularly large and not statistically significant, and then there's a big increase
for both of them when you broadcast. Let me show you the model that's going to tease out
the two effects I want to see. So this model, I've got "S" is either the log of DVD sales
of piracy for movie "i" in DMA "j" in week "t" after DVD release date, or piracy for
movie "i" DMA "j" in week "t." I've got a set of movie fixed effects of DMA fixed effects
of dummy variables. At the end, I'm going to try to dummy out the Christmas weeks for
those that were released over Christmas. And then what I'm going to try to do is I've got
a dummy variable for the HBO window, which is here. I've got a dummy variable for the
broadcast period. So a separate--I want to try to separate out the broadcast period.
And then what I'm going to try to do is I'm going to take a prediction of what the lost
sales on VOD would look like for movie "i" DMA "j" in time period "t." So we just use
a couple different exponential decay curves. It's not particularly sensitive to this, but
I'm going to try to predict what would VOD sales have been for this movie in this DMA.
And I'm also going to try to predict what would iTunes sales have been for this movie
in this DMA and I'm going to interact those two variables with a dummy variable for the
broadcast period, okay? And what I'm hoping here is if this tells me, for DMAs with high
levels of VOD sales, did they have a higher than normal change in DVD sales and change
in piracy? And DVDs with higher than normal iTunes sales, do they have a higher than normal
change in DVD sales and piracy? Okay? Does that make sense? All right, cool. And so this
is what we see. So if you--if you look at changes in DVD sales, you can see an increase
in sales during the blackout period, although not statistically significant. You can see
an increase on sales during the broadcast period, consumers see this on HBO and go out
and buy the DVD. You can see an uptick in DVD sales for DMAs that have high--that have
high VOD sales. So lost VOD sales seem to be moving towards DVD sales. I see no statistical
change in DVD sales as a function of DMAs that have high levels of iTunes--of lost iTunes
sales. And on the flipside for piracy, I see similar coefficients over here, showing the
movie on HBO leads to an increase in the demand for piracy as we've seen them in other papers.
And DMAs that have a large number of VOD sales have no statistically significant increase
in piracy. And DMAs that have a high Lost EST sales have a statically significant positive
increase in piracy, okay? Another way to see this is you can just plot the lost EST sales
and lost VOD sales as a function of the conditional predicted VOD--DMA--DVD sales and piracy.
And again, what you find is, you know, a nice--a nice strong correlation here, not much correlation
here and not much correlation for DVD sales, but a nice strong correlation for the change
for DMAs that have higher levels of lost EST sales with respect to increases in piracy.
Does that make sense? All right, cool. Bob had asked me to finish in about 40 minutes
and I think we've actually done that. So the punch line here I think is, it sure does start
to look like it's possible to compete with free. Now this isn't a huge shock for marketers,
right? Competing with free is really just a special case of price competition. I've
got to figure out a way to differentiate the paid product from the free product. And what
this--what this suggests is that I can differentiate that in a bunch of different ways. I can differentiate
that based on usability, based on convenience, what have you, but showing it on television,
showing it on HBO, and apparently, potentially, showing it on things like Hulu might increase
the demand for the paid pieces of those content. The other one that I like is that consumers
sure seem to be much more tied to their mode of consumption than they are to the legality
of their consumption. So I think the conventional wisdom in the industry was you've got a set
of pirates and you've got a set of legitimate purchasers. The pirates, you're never going
to be able to get back and the legitimate purchasers are going to choose between your
physical channels and your digital channels. If I had a digital channel, it's going to
suck away people from the physical channel without sucking away many people from piracy.
And we see in each of these three cases exactly the opposite, that adding a digital channel,
you know, at a reasonable price and a reasonably convenient format reduces the demand for piracy
and doesn't have a big impact on the demand for the physical good. The way--and I--I'd
be interested in interacting on this, but the way I think about this is I think consumers
decide, "I want to consume digitally." And once I decide I want to consume digitally,
I'm going to find it, at least some of these consumers are going to find it in a legitimate
channel if I can, and if I can't, I'm going to steal it, or whatever you're preferred
word is for piracy, okay? The other piece here is I think there are a bunch of interesting
opportunities to try to understand the impact of these digital channels through these--these
sorts of natural experiments. And so we're out here talking to different companies. We're
talking to Facebook earlier today, talking to Yahoo tomorrow to try to see if we can
identify some of these experiments or changes in digital distribution and use them as ways
to better understand how consumers respond, okay? All right. So I think almost at 40 minutes.
I would love to take questions and interact on this because I know you guys have seen
some interesting experiments along these lines. Sure.
>> [INDISTINCT] dug up that although, Google might [INDISTINCT].
>> SMITH: Yeah. We talk--so the--I'm sorry, yeah. The question was, "Have we looked at
revenue impacts?" Is--as I, you know, one interesting question off of that is, "Is it
true that the digital channel is lower margin than the physical channel?" I don't know whether
that's--I don't know whether that's true or not. And I'm not sure the studios have a good--have
a good idea of whether that's--whether that's true or not. You could imagine a world where,
you know, I cut out a whole bunch of production phases and I cut out a whole bunch of middle
men and I can go directly and get higher margins. I could also imagine a world where the competition
from piracy leads to lower margins. I don't know. With studio X, we were interested in
a different question for them. And--yeah? >> [INDISTINCT] related to that, it's actually,
if you're moving to a digital method as you--your point here, then actually--and then half of
that is, let's say to piracy, "How are you at digital content?" through from a analog
and converting that to a lower [INDISTINCT]? >> SMITH: That's the fear, right?
>> I mean, my point about this is it seemed more just, you know, because of today, just
the way the technology is, you know, you either have to choose one of the other [INDISTINCT]
or you don't and that changes it if you have more Internet connected TVs [INDISTINCT]?
>> SMITH: Yes, this is the challenge. And again, I think the conventional wisdom in
the industry was if we ignore these digital distribution channels we'll keep people longer
in these higher margin physical channels. That is, you know, adding a digital distribution
channel will cause people who otherwise would have bought physical to go over to digital
distribution. And we don't--you know, while that's possible, we don't see that in the
data. It sure looks like the people who want to consume digitally are already gone and
they're getting it in another way. So, you know, managerially, you might worry about,
"Am I--am I leading consumers to a lower margin channel?" I don't know whether it's lower
margin or not but I think practically it sure looks like the consumers are going to figure
that out on their own and go there directly. So I'm not sure how.
>> HAL: So how rapid is the response time is being seen after the broadcast and the
uptick in the channel distribution? >> SMITH: That--good question. So Hal's question,
I'll try to remember. If I forget to repeat the question, somebody raise their hand. But
Hal's question is, "How rapid is the increase in DVD sales?" I know this best. For studio
X, they gave us weekly DVD sales, so I can't see it with a whole lot of granularity. I
know it best for the previous study we did where we were looking at it on an hour by
hour basis. And in fact, if you sit with your browser open--if you're geeky enough to sit
with your browser open, and I know Hal is, while a movie shows on television, you can
almost see an uptick in sales rank almost immediately after the movie stops showing
on the East Coast. Or it's really--it's a beautiful thing.
>> HAL: Okay. So, the interesting thing then would be we just designed that experiments
where you release something and we tease something else a few days later. Well, let's say, synchronize
the releases so that you have your [INDISTINCT] without another?
>> SMITH: Yes. >> HAL: So, then you [INDISTINCT] very rapid,
maybe with just a couple of days, staggering releases would give you a lot of information
about it. >> SMITH: Yeah, that's what--that's what we're
hoping to find settings where we can do cool things like that because, again, I think that's
how the industry is going to move forward is understanding these subtle changes to how
consumers respond to their digital distribution channels. Yeah, it's a very good idea. There's
one question back here. >> [INDISTINCT] in terms of availability [INDISTINCT]
when it's available on the piracy [INDISTINCT]. >> SMITH: We have--we're looking at that question.
It's probably the best way to put it. We think we've got a cool experiment. And so, the--I'm
sorry, the question is, "Have you looked at the impact of prices?" That is, "How sensitive
are people, you know, is $3--is an iTunes episode at $3, is there a significant amount
of elasticity versus an iTunes episode sold at $2?" The anecdotal evidence I've read suggests
that the change in iTunes music pricing from 99 cents to $1.19 actually showed a fair amount
of elasticity and we're hoping to look at that question. The other related question
is actually dealt with in the first paper I presented; the NBC paper. So what we tried
to do--again, as an economist, if I see somebody buying something on iTunes that they could
get for free my perverse response to that is there must be some non-monetary cost associated
with free or else this person would have--wouldn't have bought it, they would have gotten it
for free. So the question is, "Is that non-monetary cost a marginal cost?" You know, "Is it that
each download I make just has less usability or less convenience and that's why I choose
to get it on iTunes?" Or is there a fixed cost, right? Is it that, "I don't know how
to use BitTorrent and there's this big fixed cost associated with that and that's what's
keeping me away." And in the NBC paper, all the results we got sure pointed towards a
fixed cost component that, you know, people were avoiding BitTorrent because they didn't
know how to use it. Once they couldn't get it on iTunes anymore, they invested that fixed
cost and then they were harder to get back. >> [INDISTINCT] what's the value of my [INDISTINCT]
>> SMITH: Yeah. So--no, it's a great point. So, the question is, you know, as I understand
it, what's the--is there also a moral component associated with this or is it sort of my consciousness?
I'm aware of this. The other piece we explored was, "Is there a moral component?" And originally,
we thought a moral component would primarily have a marginal cost effect that, you know,
sort of each time Bob--I shouldn't use real names--each time--each time this gentleman
sitting here pirates something, you know, he gets a slightly more a bad feeling about
himself. We presented this to our faculty, who includes some very prominent criminologists
who said, "You know what, the literature says that actually there's a fixed cost to crime
that the first crime you commit is the hardest one and after that it becomes much easier."
So we had to model the moral cost as both a fixed in and a marginal component. But everything
lined up with sort of fixed cost. So, you know, maybe it's when you sell your soul to
piracy it becomes much easier to pirate over that and I see some people nodding.
>> Well, I know it's a lot, you might as well hang for a cow than hang for cheese.
>> SMITH: Okay. That'd be a good title for a paper, actually. Sure.
>> Have you looked at any impacts of [INDISTINCT] because you mentioned being in [INDISTINCT]
in some ways are [INDISTINCT]? >> SMITH: Right. It's an interesting question.
I haven't looked at that, right? But you can stare at iTunes and you can wonder whether
the DRM that Apple was able to control created this wonderful ecosystem for them where the
more iTunes you bought the more you wanted to buy iPods, and the more iPods you bought
the more you wanted to buy from iTunes, and churn them into whatever market share they
have, you know, 90% market share, which makes them a very tough downstream negotiating partner
for all these--all these record companies. You know, that they--like Wal-Mart does in
DVDs, Apple can be a very tough downstream negotiating partner and you wonder how of
much that is due to just Steve Jobs being brilliant and how much of that is due to the
DRM scheme that was put in place that was proprietary. It's an interesting question.
Yeah? >> It's interesting you have access to video
in DMA level, did you see any difference in the behavior between different DMAs?
>> SMITH: Yeah, there's a pretty dramatic difference. The question--thank you, Bob.
The question was, "Do you see any difference in behavior across DMAs?" And the answer is,
yes. There's, you know, obviously, there's a correlation. You know, DMAs that have high
iTunes sales also have high VOD sales, but there's enough variation and there's a decent
amount of variation across these DMAs that we were able to identify. And I should have
pointed out that the model has--we cluster the standard errors at the movie level, so
we're controlling for potential correlation across the different DMAs. But there's enough
variation there that you can pull it out. It's kind of cool. I don't know how to take
questions from these nice people up on the screen. Okay, cool. Any other questions? Any
other personal anecdotes? The nice thing is the camera isn't facing on you. Do we have
any questions from our audience in the cloud? Can you guys still hear me in the cloud? Okay,
they're all--they're putting thumbs up. Okay, no questions then. All right. Well thank you
very much. I think we ended a little bit early. I appreciate it.