We use Adobe Analytics data on online transactions for millions of products in many different categories from 2014 to 2017 to shed light on how online inflation compares to overall inflation, and to gauge the magnitude of new product bias online. The Adobe data contain transaction prices and quantities purchased. We estimate that online inflation was about 1 percentage point lower than in the CPI for the same categories from 2014–2017. In addition, the rising variety of products sold online, implies roughly 2 percentage points lower inflation than in a matched model/CPI-style index.
I call this “the gains from better matching,” as discussed in The Complacent Class.
I have been hearing this question more and more lately, even in China. Overall I think it has gone from an underrated effect to an overrated effect. Tim Worstall offers an introduction to this debate.
Let’s not forget that you do in fact pay for Facebook access, indirectly, when you pay for your cable connection, your iPad, and your smart phone. including the monthly bill, all of which are part of measured gdp. The more value Facebook brings you, the more you would be willing to pay for these goods and services. The same is true for Google and the like. So Facebook and other internet services are part of a bundled package of market value, but that is very different from claiming they are not measured in gdp at all.
There is of course consumer surplus from the internet and Facebook, just as there is from Dunkin’ Donuts. Might that consumer surplus be especially high? Well, we don’t know, but don’t assume it will be. I did some casual googling, and found a number of estimates suggesting that smart phone demand is relatively price elastic, with the iPhone a possible exception to that regularity. That implies consumer surplus isn’t especially high, because many people aren’t willing to buy at the higher price. I thus think Brad DeLong is far too optimistic in his estimates of ratio consumer surplus to market price.
You also could look at the literature on the demand for cable internet services. The results are mixed, but again I don’t see a strong case for a disproportionately high consumer surplus from these services, if anything the contrary.
Now maybe these estimates are wrong, or looking at the wrong margin in some way, but the fact that I hear them mentioned so rarely gives me pause. Cowen’s Third Law.
There is also advertising over the internet. Let’s say Facebook is a profit maximizer. Insofar as Facebook is of value to consumers, the company can get away with putting a lot of ads on the site. These will spur additional market purchases, and so part of the value of the site is again captured in gdp. Obviously some of these ad effects are simply expenditure-switching, and so there is no full capture of value, but still Facebook shows up in gdp statistics in yet another way.
Here are some previous posts on this topic.
Continuing the exchange over the value of the internet, David Henderson writes (and do read the whole thing):
So I can imagine doubling Goolsbee’s and Klenow’s 2% to 4% to reflect the broader version of CS [consumer surplus] that users get directly and then adding another 4% to reflect the lower cost of getting goods and services delivered to us even if we never use the Internet directly. That get’s it to 8%. To me that’s “huge.”
This is useful for helping sort out where David and I disagree. As I see it, the latter factor — the lower prices for goods and services — is already reflected in measured real wage gains, or smaller real wage losses than would otherwise hold, as the case may be. We’re back down to four percent.
In the post, David mentions both the “lower price” gains (which he stresses in the TGS review) from the “lower time cost of shopping” gains. The two are not quite the same. What about the latter? The lower time cost of shopping is already counted in WTP measures of the value of the internet — you’ll pay more for the internet if it saves you more time — and to some extent in gdp statistics and other real income measures. How does the gdp effect work? Let’s say it used to take half an hour to buy a book, now it takes ten seconds. People will buy more books and that gain is already measured. The rest of the saved half hour goes into other methods of production and shopping, which also show up in national income. Some of the saved half hour shows up as “pure leisure” (i.e., sitting on one’s bum, doing nothing) and that one part of the saved time doesn’t show up in gdp statistics though it still does show up in WTP estimates (which again clock in below four percent).
Time use studies also count these “time saved shopping” gains, but in a different way. Goolsbee and Klenow measure the income elasticity of leisure uses of the internet and find it is negative. That means high value of time users find the internet a time sink, on net, rather than a time saver. In any case the magnitude of this value already lies behind their estimate.
Goolsbee and Klenow estimate two percent for consumer surplus, not “2% to 4%,” and they worry they are overestimating the gains because of assumptions they make about substitutability. The other studies I cited are below four percent, unless it’s for computer use more generally. I’m the one who kicks it up to four percent, largely because of Facebook, which de facto postdates some of the papers (though not the FCC study which still gives modest sums), and also because of unmeasured workplace consumption usage. But that’s probably as high as we should go, once we adjust for what is already counted elsewhere.
Addendum: The Billion Price Index matches the CPI pretty closely, so there is not much gain from invoking the “CPI doesn’t measure internet bargains” argument, though there may still be a small effect there. And Matt Yglesias offers interesting comment.
David Henderson raised this question again, as has Bryan Caplan in the past. Both seem to suggest that the consumer surplus from the internet is quite high or perhaps even “huge,” although I am not sure what number they have in mind. I am disappointed that they are not engaging with the academic literature on this topic.
1. An 86-page 2010 FCC study concludes that “a representative household would be willing to pay about $59 per month for a less reliable Internet service with fast speed (“Basic”), about $85 for a reliable Internet service with fast speed and the priority feature (“Premium”), and about $98 for a reliable Internet service with fast speed plus all other activities (“Premium Plus”). An improvement to very fast speed adds about $3 per month to these estimates.”
2. A study from Japan found that: “The estimated WTP for availability of e-mail and web browsing delivered over personal computers are 2,709 Yen and 2,914 Yen, on a monthly basis, respectively, while average broadband access service costs approximately 4,000 Yen in Japan.” By the way, right now the exchange rate is about 80 Yen to a dollar.
3. The Austan Goolsbee paper, based on 2005 data, does a time study to find that the consumer surplus of the internet is about two percent of income.
4. This paper finds a four percent consumer surplus from the personal computer more generally, not just the internet.
5. Robin Hanson serves up an excellent debunking of some exaggerated consumer surplus claims.
6. Many of the benefits of internet cruising are captured in gdp figures, such as using it to find a job or the money you spend on smart phones. By the way, here is a good paper on consumer surplus in the book market, though it offers no overall CS estimate from the internet.
You can take issue with these papers for ignoring personal internet use at work, the inframarginal benefits to infovores, or other issues, such as whether the existence of the internet increases workloads in what are supposedly leisure hours. But there is the place to start and the numbers are not outrageously high, not close to it.
Or put all that aside and think through the problem intuitively, in terms of time use decisions. Your marginal hour of non-internet leisure time is worth more than spending another hour of time on the internet. In other words, at the margin your consumer surplus from the internet is about the same as your consumer surplus from going to the movies or taking a walk. That’s nice, but suddenly the consumer surplus from the internet doesn’t seem like such a big deal any more. It’s probably not going to add up to millions. If the internet were as awesome for consumer fun as some people claim, it would have pushed out more of our other uses of leisure time.
What about the inframarginal units of internet use? Might the consumer surplus there be huge? If you think of books, movies, newspapers, and CDs as some of the relatively close substitutes for some uses of the internet, we know from cultural economics that the demand curves for those enjoyments are usually smooth, normal, and continuous, more or less. They don’t have enormous, hidden inframarginal benefits.
Penicillin probably does have such an enormous inframarginal benefit; the initial doses can be of great value but past some margin the value falls to zero or negative. The internet doesn’t seem to be like penicillin.
You can even make an argument that the inframarginal valuations of internet use are especially low, relative to the marginal values. Have you ever heard that the internet is “addictive”? That doing some makes you want to do more? That the internet has a virtually unending supply of interesting content? Personally I find that I could read more working papers, without much decline in their interestingness, except that the exigencies of my daily life interfere (at some margin). Those are all signs that the marginal valuation of the internet does not fall off so drastically as one moves down the demand curve. If you’re not using the internet more, it’s not because the internet is getting much worse with additional use units, it’s because it is digging into increasingly important parts of your non-internet life. That brings us back to the inframarginal unit having a value not so far away from the marginal units.
It is likely that the consumer surplus of the internet is in the range of two to four percent of gdp. On one hand, that’s “a lot” but on the other hand it’s not enough to close the “stagnation gap” in wages since 1973. It’s not close. It also may be quite small compared to the consumer surplus from the major innovations from earlier in the 20th century, such as antibiotics, without which I probably would be dead.
There is a new and very interesting study out from McKinsey on this topic. If you get past the press release, however, and give it a closer read it is consistent with stagnation hypotheses, contrary to some claims. The study shows a few things:
1. Most of the economic benefits of the internet are in fact captured in current economic statistics, which I’ve already argued do not look so fabulous. The point is not to blame the internet, as without it things would have been worse. The point is that the internet gains, in absolute terms, haven’t been large enough to produce a rosy picture overall.
2. The direct and indirect economic effects of the internet account for 3.8% of U.S. gdp, as currently measured (p.15). That’s less than many people think and that value is already incorporated in the current gdp measure. The good news — and it is good news — is that there is lots of room for future growth from internet impact. We’ve yet to really organize our economy around the internet, as we someday will, and then the gains will be enormous. In the meantime we are waiting.
3. What about the unpriced consumer surplus gains from the internet? The study considers that too:
In general, this surplus is generated from the exceptional value users place on Internet services such as e-mail, social networks, search facilities, and online reservation services, among many others. This value far outweighs the costs, both actual costs such as access and subscription fees and annoyances such as spam, excessive advertising, and the need to disclose personal data for some services. In the United States, for example, research conducted with the Interactive Advertising Board found that consumers placed a value of almost €61 billion on the services they got from the Internet, while they would pay about €15 billion to get rid of the annoyances, suggesting a net consumer surplus of about €46 billion.
A nice gain, but in an economy with a $14 trillion gdp it’s not that big a deal. It’s also less than the previous Goolsbee and Klenow estimate of about two percent of gdp, for the consumer surplus from internet use. That is not nearly enough to refute the view that median income growth is much slower in recent times, and that’s without adjusting for the problematic status of expenditures on health care and K-12 education. Arnold Kling and Bryan Caplan stress this point — the unpriced benefits of internet use — but I don’t see them offering a better number than this rather paltry calculation, now confirmed as low, at least relative to the claims of internet boosters, by two differing sources.
In fairness, I should note that I cannot trace the study cited above. It may be wrong. The Goolsbee and Klenow paper may be wrong, and it is somewhat out of date, from 2006. I think a consumer surplus estimate of three or four percent is entirely plausible for 2011. We’re still left with relative stagnation. After all, penicillin had some consumer surplus too.
Here is some media coverage of a recent Facebook study of its economic impact in terms of revenue and jobs. Facebook claims it added $227 billion to the global economy, but they approached the question the wrong way.
The correct method is to treat jobs as a cost of Facebook, not a benefit, admittedly that is not how politics works nor is it how corporate PR works. We should measure the benefits directly by consumer time use studies, much as Austan Goolsbee and Peter J. Klenow did in their paper on the internet (pdf).
My question today is this: what is the most accurate one line back-of-the-envelope estimate you can come up with for the gross benefits of Facebook, not bothering to subtract for the costs of running the site? Here is one (hypothetical and illustrative) example, for America only:
100 million regular users, one hour a day, time valued at $10 an hour, and multiply for $365 billion a year.
You will notice this method implicitly captures the value and disvalue of the ads on Facebook. The better and more useful the ads are, the more time people will spend on the site.
I don’t devote time to Facebook (I can thank MR for that), so surely you can do better than I in building a plausible one-line estimate. Please leave your answer in the comments.
You will find it here, with contributions from Viral Acharya, Scott Sumner, Hal Varian, and Paul Seabright. From elsewhere, Noah Smith cautions economists not to invoke technology too often. Brad DeLong chimes in. From a few years ago, Austan Goolsbee measures the consumer surplus from the internet; his numbers do not refute the standard view that median income growth has become much much slower.
Professors Ghose, Smith and Telang chose a random sample of books in print and studied how often used copies were available on Amazon. In their sample, they found, on average, more than 22 competitive offers to sell used books, with a striking 241 competitive offers for used best sellers. The prices of the secondhand books were substantially cheaper than the new, but of course the quality of the used books (in terms of wear and tear) varied considerably.
According to the researchers’ calculations, Amazon earns, on average, $5.29 for a new book and about $2.94 on a used book. If each used sale displaced one new sale, this would be a less profitable proposition for Amazon.
But Mr. Bezos is not foolish. Used books, the economists found, are not strong substitutes for new books. An increase of 10 percent in new book prices would raise used sales by less than 1 percent. In economics jargon, the cross-price elasticity of demand is small.
One plausible explanation of this finding is that there are two distinct types of buyers: some purchase only new books, while others are quite happy to buy used books. As a result, the used market does not have a big impact in terms of lost sales in the new market.
Moreover, the presence of lower-priced books on the Amazon Web site, Mr. Bezos has noted, may lead customers to "visit our site more frequently, which in turn leads to higher sales of new books." The data appear to support Mr. Bezos on this point.
Applying the authors’ estimate of the displaced sales effect to Amazon’s sales, it appears that only about 16 percent of the used book sales directly cannibalized new book sales, suggesting that Amazon’s used-book market added $63.2 million to its profits.
Furthermore, consumers greatly benefit from this market: the study’s authors estimate that consumers gain about $67.6 million. Adding in Amazon’s profits and subtracting out the $45.3 million of losses to authors and publishers leaves a net gain of $85.5 million.