Wednesday, August 31, 2016

No, social media is not undervalued in GDP

There's an argument rattling around the Interwebz suggesting that the observed slowdown in productivity growth is, at least in part, a statistical illusion arising from the fact that users don't pay to use Twitter or Facebook.  Or Blogger (to chose a free blogging platform totally at random; I mean, who uses Blogger these days?).

I'm skeptical.

First, let's keep in mind that, when we talk about GDP, we're talking about the market value of goods and services, not their value value.  It's no objection to say that the market value of social media differs from its "real" value -- that's true of literally everything that composes GDP.  It's perfectly valid to argue that GDP is a meaningless construct, at best reifying the current distribution of income and wealth, and at worst merely an exercise in adding apples to oranges because it's easier for our simple brains to think about one number than about two (1).  But it's an entirely different matter to claim that a particular component is being mis-measured.

The usual argument about social media is that, surely, some users would be willing and able to pay to use it.  Ability and willingness to pay for something is, more or less, the definition of market value.  So, to the extent that these users get something they would pay for for free, there is unmeasured economic output being created and consumed.

And that's fine as far as it goes.  However, it's important to keep in mind that a social media property "creates value" for multiple groups of consumers: at minimum, for both users and advertisers.  The total value produced is the sum of the value delivered to all of the consumers.  And that brings us to the question of pricing strategy.

To talk about pricing strategy, I'm going to make up a hypothetical social media property.  My business plan is to combine the decorum and wit of Twitter with Facebook's custom of conducting unauthorized psychological experiments on its users.  Naturally, we will call this transformative platform "TwitBook." (Tagline: "Giving dude-bros and mean girls the social media experience they deserve.")

Now, TwitBook will need to set some prices.  What should we charge users for this unique and innovative experience?  What should we charge advertisers for the opportunity to reach our bizarrely sought-after demographic?  And finally, what should we charge non-participating audience members for the chance to observe the hijinks in real time, much as they might watch a reality TV show or monster truck rally?  OK, I'm kidding about the third one.  Sort of.

To devise this strategy, I'm going to hire the most ruthless greed-heads in sharp suits that I can find.  (Or, as an economist might say, "Assume firms maximize profit, pi, defined as total revenue (TR) less total cost (TC)."  The bit about the sharp suits is, formally, exogenous.)  My greedy minions quickly arrive at two important conclusions.

First, because of the nature of the business, we have unprecedented technical flexibility in price setting.  We can meter literally everything a user does (including interactions with advertisements), and payment processing has never been easier.  This is probably the first industry ever in which implementing a truly optimal price structure is technologically feasible.

Second, there is an inconvenient tension between the prices we charge users and the prices we charge advertisers.  The more we charge our users, the fewer users we will have (because the elasticity of demand is, if not infinite, quite far from zero), and the less we can charge the advertisers.

"Very insightful," I say to my sharp-suited legion of doom, "but I didn't hire you for your penetrating understanding of market mechanisms; I hired you to make me very, very rich.  I don't care whether my money comes from users, advertisers, or space aliens.  Or from spectators, if you can figure out how to make the whole monster truck angle work.  Just find the set of prices that maximizes the total value of the platform."

"You bet, boss," say my compliant lackeys, because that's how my imaginary friends talk.  In my imagination.  Stop looking at me like that.

Soon, they're back with the a pricing structure...and the weird bit is that all of the user-facing prices are zero.  I'm disappointed, because the idea of dude-bros and mean girls literally paying me to run experiments on them made me alarmingly gleeful.  But business is business.

Still, I want an explanation.

"Yes, your highness," say my imaginary business strategists, because now they've started mocking me.  "Our users do value the platform," they explain, as if to a small child, "but not very much, is all."  (If they weren't talking to me as if I were a small child, they'd probably add something clever sounding like, "at the margin."  Jerks.)  "Charging users even the tiniest amount actually reduces the total value of the platform, since the loss of value to the advertisers is measurably higher than the user-revenue we gain."

The upshot of all this is that their advertising revenue is actually a pretty good indicator of the total "value created" by platforms like Facebook and Twitter.  Nobody is leaving "unpriced value" on the table.

But what about consumer surplus?  The fact that the marginal Facebook user places a value of approximately zero on the service doesn't mean that lots of people don't love Facebook.  Surely that must count for something!  But social media is hardly the only industry to yield a large-ish consumer surplus.  Agriculture (a.k.a. "food") comes to mind.  And more generally, the consumer surplus argument is really just another version of the point that "market value is not value value," which is true, but irrelevant to the measurement (or mis-measurement) of GDP.

So I'm inclined to doubt that the growth of social media leads us to systematically underestimate growth, productivity, or the growth of productivity.  In fact, there's a stronger argument for the reverse.

Time is the universally binding budget constraint: there aren't any more hours in the day than there were in 1972 (or in 1066, or in 800 BCE).  Consequently, people's consumption of social media comes at the expense of other activities, notably at the expense of watching broadcast television.

Now, in terms of pricing, TV looks a lot like social media: it's advertising supported and free-to-consume for viewers.  However, there is a major technical difference:  charging users to view broadcast television has been infeasible until very recently.  So, while we know that the observed pricing structure for social media is optimal (or, at least, unconstrained by the feasibility of monitoring and billing for usage), we do not know this for broadcast TV.  That is, the argument that advertiser-supported media is undervalued in the economic statistics applies with much more force to broadcast TV than it does to social media.  And the fact that people turn out to be willing to pay for commercial-supported cable channels suggests that broadcast TV has been undervalued for decades.

So the real story of bias in GDP measurement is the decline of an undervalued product (broadcast TV) in favor of  a properly valued one (social media), which means that economic and productivity growth are both being systematically overstated.

Now, I'm not persuaded that this is, quantitatively, a very important phenomenon.  But it does leave me skeptical of the notion that free stuff on the Internet is somehow concealing a productivity bonanza from the national income and product accounts.

(1) What to make of the fact that I score intellectual laziness as "worse" than disingenuously naturalizing the social status quo is left as an exercise to the reader.


  1. "But social media is hardly the only industry to yield a large-ish consumer surplus. Agriculture (a.k.a. "food") comes to mind. And more generally, the consumer surplus argument is really just another version of the point that "market value is not value value," which is true, but irrelevant to the measurement (or mis-measurement) of GDP."

    OK, now change the argument to what it really is. Well, at least what it really is when I make it.

    What we're really interested in is value value. GDP is only a proxy for this. We generally assume that value value is 200% of market value, ie consumer surplus is 100%.


    Now we've got something that comes along where value value is 20x, 50 x, market value. And I don't think that's out of order. In GDP stats WhatsApp appears as the engineers wages and nothing else. But 1 billion people use it.

    GDP may well be being properly recorded. But we must remember that it is only a proxy for what we're actually interested in, value value. That latter soaring while everyone complains about stagnation.

    1. Hi Tim! Thanks for taking the time to weigh in here. (Also, if you'd like to add a link or two pointing to places where you or others have made the argument in greater length than a blog comment permits, please do.)

      Unfortunately, I have to say that I'm unconvinced. As far as I can tell, your argument boils down to: "The water-diamond paradox matters more today than it used to because Facebook." I'm having a hard time accepting the proposition that price-value delta Facebook is a lot higher than that for water. I'm also not seeing any accounting for the fact that consumption of social media displaces consumption of other media, and that the price-value delta for the declining "old media" is, if anything, larger than the price-value delta for the expanding new media.

      Where we kind-of agree is on the proposition that GDP is not the same as social welfare. Even here, though, we seem to part company. Your view, as I understand it, is that in the past GDP was an acceptable proxy for social welfare, and that its usefulness as a proxy is declining, because Facebook. My view is that GDP has always been 100% worthless as a proxy for social welfare, so its validity is exactly what it always was.

      But that doesn't mean that GDP isn't something worth measuring or worrying about. GDP is a pretty good measure of the aggregate market value of the output of domestic labor. If that happens to be the concept of interest at the moment, GDP is as fine a measure as ever.

      And that brings me to your assertion that "[Value value is] soaring while everyone complains about stagnation." This seems to come out of nowhere. Perhaps you've identified your preferred measure of social welfare in other writing. But without that context, I've got no idea what you're even claiming.

      As an aside, I'll add that taking consumer surplus as a measure of "value value" entails committing to a rather peculiar social welfare function. Let's recall what consumer surplus means. If he had no alternatives, Donald Trump (to pick an ostensibly wealthy individual at random) might be willing and able to pay $10,000 for a sandwich. (Yum!) But because he has the ability to transact in a competitive market, he only has to pay $2. That yields a consumer surplus of $9,998. Yay? Don't get me wrong: I think it's a fine thing that the Donald can buy lunch as cheaply as the rest of us. But consumer surplus depends way too much on the distribution of income and available pricing technology to function as a social welfare barometer.

  2. Actually the thesis is "the water-diamond argument works (far) more for the last 20 years than for the 40 years before it.

    Let's rewind it a little. There is a huge political claim going around based on which quality of life is worse now than "x" years ago (x= usually a few decades). That claims gets a lot of supporter.

    We want to disprove that claim, because it feels objectively wrong. First thing we do is looking at PPP pro capite gdp, then disposable income i guess right? but that's pretty stagnant.

    So what's missing? perhaps the fact that the consumer surplus around us right now, if you have the right attitudine, is far greater than 20-30 years ago.

    And it's not strange that the people who claim that 30y ago life had a better quality are mostly the same that call smartphones "stupid toys", that aren't very good at using google search (or don't feel like they need to use it often), have a set of value anchored in the past (traditional family, male provider role etc etc).

    Gdp wasn't worthless as an indicator of social welfare when social welfare was improving mostly by the better availability of stuff in people houses, cars, and similar things. It was actually a good proxy.

    What i claim (and i think Tim also claims) is that with the same real income as 20 years ago, if you have the right attitude and value scale (you value knowledge a lot for example), life is currently extremely more enjoyable than 20 years ago.

    That's a strong claim i know. But it's a point that current GDP measure is missing much more than it was missing (if any) when comparing the 80s to the 60s.

    We keep hearing that the current generation could be the first one to be "worst off" than the previous one for example. And that's insanely wrong if you analyse reality with the lens of digital consumer surplus unless you have a personal value function anchored to past/traditional values so much, and valuing knowledge so little, that you can correctly claim that life was better off (or similar) 20-30y ago.

    And then we have the tipping point; i am among those who think that people that value knowledge very little are INFERIORS. Are "worse people". Are WRONG, in a moral sense. And so, basically, they are whining because they are "so bad morally" to not give value to knowledge enough to appreciate what the digital revolution brought to us.

    Ofc much of this is non-economical in any sense, but that's the claim in its fullest form

    1. Hi Luciom! Thanks for your thoughtful comment. And, in general terms, I actually agree with most of what you say.

      I should clarify that my post isn't really intended to be part of a political argument about whether "America is in decline" or something along those lines. It's meant to be part of a more-or-less technical conversation within the econo-blogosphere about the observed slowdown in the growth of productivity (in the sense of "gross output per hour of labor"). Within that conversation, it's taken for granted that we're talking about GDP because we mean GDP and have a pretty solid handle on what that means (and doesn't mean). I did point out, quite specifically, that GDP is a market-value measure, not a "value value" measure, but I may not have done enough to signpost the assumed context.

      Let me shift gears a little and focus on the broader "America: better off or worse than yesterday?" question. I think there's a very strong case to be made for "Better! And not by a little!" But I think its proponents undermine their argument by making it in quasi-economic terms.

      For example, I see the term "consumer surplus" used to mean "the difference between market price and 'value value.'" But consumer surplus is a technical term, and that is simply not what it means. Consumer surplus is the difference between the price a consumer pays and the price they would have been willing *and able* to pay. That "and able" bit is crucial. If you give free Internet access to somebody who wasn't able to pay for it in the first place, you create exactly zero consumer surplus. On the other hand, if you give $100 / month mobile phone service to the bond trader who was previously paying $1,000 / month for his '80's era car phone (remember car phones?), you create a big chunk of consumer surplus. "Consumer surplus" really isn't the droid you're looking for.

      I also don't see anything to be gained by taking on the job of arguing that GDP was a *good* proxy for social welfare at some point in the past, or in arguing about the direction of the "trend" in its "goodness." (Example: thanks to a variety of regulations, Americans are experience less ambient lead poisoning, and thus less brain damage, than in the early 1970's. Does that mean GDP is a *worse* indicator of welfare today -- since it omits the value of not-getting-lead-poisoning -- or *better*, because something that should be deducted from GDP -- the value of all that lead poisoning -- has become smaller? Answer: who cares?) GDP isn't a bad measure of welfare because of technical measurement issues; it's wrong *in concept.*

      I think you're on the right track in the last couple of paragraphs: if we're going to make an argument that social welfare is improving, we need to actually spell out what we mean by social welfare -- that is, how we believe different things ought to be valued. I might or might not agree with the valuation you propose, but *that's* the right way to have a conversation about the subject.

      (Surprise! A political argument that turns out to be a quarrel about values! Who could have sen that coming?)