Are we undermeasuring productivity gains from the internet? part I

From my new paper with Ben Southwood on whether the rate of scientific progress is slowing down:

Third, we shouldn’t expect mismeasured GDP simply from the fact that the internet makes many goods and services cheaper. Spotify provides access to a huge range of music, and very cheaply, such that consumers can listen in a year to albums that would have cost them tens of thousands of dollars in the CD or vinyl eras. Yet this won’t lead to mismeasured GDP. For one thing, the gdp deflator already tries to capture these effects. But even if those efforts are imperfect, consider the broader economic interrelations. To the extent consumers save money on music, they have more to spend or invest elsewhere, and those alternative choices will indeed be captured by GDP. Another alternative (which does not seem to hold for music) is that the lower prices will increase the total amount of money spent on recorded music, which would mean a boost in recorded GDP for the music sector alone. Yet another alternative, more plausible, is that many artists give away their music on Spotify and YouTube to boost the demand for their live performances, and the increase in GDP  shows up there. No matter how you slice the cake, cheaper goods and services should not in general lower measured GDP in a way that will generate significant mismeasurement. 

Moving to the more formal studies, the Federal Reserve’s David Byrne, with Fed & IMF colleagues, finds a productivity adjustment worth only a few basis points when attempting to account for the gains from cheaper internet age and internet-enabled products. Work by Erik Brynjolfsson and Joo Hee Oh studies the allocation of time, and finds that people are valuing free Internet services at about $106 billion a year. That’s well under one percent of GDP, and it is not nearly large enough to close the measured productivity gap. A study by Nakamura, Samuels, and Soloveichik measures the value of free media on the internet, and concludes it is a small fraction of GDP, for instance 0.005% of measured nominal GDP growth between 1998 and 2012. 

Economist Chad Syverson probably has done the most to deflate the idea of major unmeasured productivity gains through internet technologies. For instance, countries with much smaller tech sectors than the United States usually have had comparably sized productivity slowdowns. That suggests the problem is quite general, and not belied by unmeasured productivity gains. Furthermore, and perhaps more importantly, the productivity slowdown is quite large in scale, compared to the size of the tech sector. Using a conservative estimate, the productivity slowdown implies a cumulative loss of $2.7 trillion in  GDP since the end of 2004; in other words, output would have been that much higher had the earlier rate of productivity growth been maintained. If unmeasured gains are to make up for that difference, that would have to be very large. For instance, consumer surplus would have to be five times higher in IT-related sectors than elsewhere in the economy, which seems implausibly large.

You can find footnotes and references in the original.  Here is my earlier post on the paper.

Comments

To the extent consumers save money on music, they have more to spend or invest elsewhere, and those alternative choices will indeed be captured by GDP.
---
Or sit on dad's couch and listen to the music.

Yes this is an incredible fallacy. The other stuff consumers spend money on shows up in GDP but the music does not. This sentence doesn't debunk the mis-measurement problem, on the contrary it exemplifies it. Previously I spent $200 on music and got $200 worth of music and GDP showed $200. Now I spend $200 on restaurants and get $200 in restaurant meals plus $200 in music. Measured GDP hasn't gone up but actual GDP has gone up. Sentences like this don't give me much desire to read the rest of the paper. This kind of fallacy is routinely taught in introductory macro.

"Now I spend $200 on restaurants and get $200 in restaurant meals plus $200 in music."

No, you are no longer getting $200 in music.

Interesting - so the music you formerly paid for becomes worthless?

How do you measure the value of music you hear over the radio?

The music that you do not buy this year does not add to the 2019 GDP. The car that you bought three years ago isn't worthless, but it doesn't contribute to this year's GDP, either.

'The music that you do not buy this year does not add to the 2019 GDP. '

True, as far as it goes. How does that work with a solar panel? I buy one this year, and it generates electricity for 20 years. Worse, it supplants the electricity previously sold by a utility, thus actually reducing GDP in all years following its purchase.

As other commenters have suggested, the problem just might be in the concept of GDP as a way to measure anything except what it measures, actually. Much like IQ tests. A person's character is much relevant to society than their IQ, but character is not readily quantifiable.

I don't see what the problem with using a GDP is unless someone thinks the GDP should also measure a country's happiness or spiritual growth for the year as well. The GDP was never intended to try to measure that.

The question isn't really about GDP here, it is whether productivity and progress has slowed or stalled. One way to measure productivity is value produced divided by input (usually manhours but could be cost). Economists are using GDP as a measurement of value. On that basis productivity has stalled or slowed in recent times. But if for instance productivity of music business is measured by hours of music consumed by manhours to produce and deliver that product to the consumer, I would wager that productivity has shown as massive increase (in the music industry at least) if manhours are used as the divisor, and even more massively if cost is used as the input.

My example of this kind of thinking at the extreme is if a machine were invented in 2025 that could basically describe all humans wants for free. So in this economy there would only be the input of the cost of making the request for something, and output would be very valuable if compared at 2019 prices. But as no-one would want to buy anything in this economy so value of the products is zero measured at 2025 would be zero. But there is some non zero time to make a request for something, so measured productivity in value per manhour is very low. Now you could argue that inflation adjustment would take care of this by deflating everything so the measured real economy would be growing fast during the introduction of this technology, but I would guess that as things become worthless then they remove them from the inflation basket. Like music today - most music is consumed for free so isn't in the inflation basket.

But isn't that exactly how economists use statistics like National Income or GDP -- as a proxy for happiness? We can't measure your "marginal utility", so we measure your "willingness to pay".

There is quite a bit of evidence from behavior economics and psychology that this assumption is just plain wrong. Yet much public policy is still driven by it simply by lack of other options.

"No, you are no longer getting $200 in music"

If I'm getting the same package of music services that was valued at $200 in your 1995 GDP calculation, and now you want to value that output at zero, then your GDP measure has a problem.

If I'm now getting a package of music services that would have cost $20,000 using the prices you used to measure 1995 output, then your GDP measure has a massive problem.

You can try to save the naive GDP measure with some sort of circular definition of output value, but then it's not really usable for much beyond salvaging your naive GDP measure.

No, there no problem at all. You didn't purchase $200 worth of music this year so the GDP didn't go up and there is nothing naive about it.

We are close to having a significant health improvement by taking a pill that may cost $5 a day. If 100 milliiiin older Americans take that in 2022, the GDP will go up $5 x 365 x 100 million or $180 billion. That's it. No need to calculate increase in happiness, "value", etc.

(my "i" key stuck... that won't happen in 2022!)

"We are close to having a significant health improvement by taking a pill that may cost $5 a day"

Which drug are you referring too?

That's fine Todd, as far as it goes. But it only goes as far as "if you naively define GDP as just the number of nominal dollars spent, then GDP is an easily-implemented measure of the number of nominal dollars spent". Similarly, if I define GDP as "six", then I never have any kind of measurement difficulties, because GDP is always just six.

But neither "six" nor "the number of nominal dollars spent" have any useful policy implications.

"If I'm getting the same package of music services that was valued at $200 in your 1995 GDP calculation, and now you want to value that output at zero, then your GDP measure has a problem."

That's the very definition of productivity -- reducing the cost of producing something.

In 1995, a computer with the power if your iPhone would have cost you $100,000. But we don't value your iPhone at $100,000 (you don't either). We value it at what you paid for it today. The GDP increase from the falling price of that iPhone isn't reflected by adjusting the "value" of an iPhone in the GDP calculation; it's reflected by what you buy with the leftover money you previously would have spent on your computer.

GDP is a money measurement, not a happiness measurement. Your music purchases this year may make you $200 worth of happy, but GDP only goes up by the amount they cost you: whether $200, $10, of $0.

"That's the very definition of productivity -- reducing the cost of producing something"

"In 1995, a computer with the power if your iPhone would have cost you $100,000. But we don't value your iPhone at $100,000 (you don't either). We value it at what you paid for it today."

The approach outlined in the second sentence ensures that GDP is not a useful measure of productivity - the amount of resources necessary to produce a consistent bundle of outputs. You seem to think you can just relabel "productivity" as some sort of convenience sample of leftover money, but that's not really a measure of anything, beyond, as you say, nominal dollars. But "the amount of nominal dollars is about the same" is simply not a useful contribution to a debate about the productivity impacts of the internet.

(We don't value goods that went unpurchased in 1995 at their notional 1995 cost because there's no revealed preference justification for penciling them in at that figure. But any serious measure of productivity needs to have an approach to valuing them beyond "before no one bought them, now they cost six bucks, so that's not very much productivity")

So then GDP is not the correct thing to measure? I have more “things” (stuff, “Free” music, experiences) for the same amount of money spent.

This. The value of the music to the consumer is roughly the same, while the cost is much lower. The productivity gains from lower prices doesn't show up in GDP calculations.

Access to the Internet is pretty important to me, and while it is difficult to say how much I would have been willing to pay, I would for instance rather halve the size of my apartment and drop holiday travels abroad than lose digital services. That's 25% of my (the?) economy right there.

I don't think this argument makes sense in the context. I understand what you're saying, but I think in a crude sense you're discussing consumer surplus.

Regarding things you'd give up before losing access to digital services, that type of extreme comparison holds for many things. I'd rather give up those things than lose access to water, but no one is claiming water access is underpriced in GDP. Same with roads, etc.

GDP is a measurement of production by counting things and assigning a value to them. The only systematic way to do that today is to count what it costs to acquire these things. We cannot take on sector of the economy and treat it differently.

There's nothing wrong with GDP, there's something wrong with what we try to do with GDP.

"There's nothing wrong with GDP, there's something wrong with what we try to do with GDP."

Yes! GDP doesn't equate to happiness, yet much public policy (and personal decision) is derived from the assumption that it does.

Most things that actually matter to our well-being (love, companionship, family, children, church, self-improvement) don't show up in GDP.

"To the extent consumers save money on music, they have more to spend or invest elsewhere, and those alternative choices will indeed be captured by GDP"

This is necessarily bounded by the pre-internet share of spending on music. I now consume a bundle of music which, prior to the internet, would have cost more than 10% of my income. Valuing this is tricky, because there's no revealed preference case for counting it at it's 1995 cost, but looking at the reduction in my spending on music over the period is absolutely irrelevant to the calculation.

What's tricky about it? You have lots of things that used to be very expensive (cars, travel, canned food, computers, generic medicines). Are those tricky to measure?

Music is an example of a product which has experienced unbelievable productivity gains, dropping the price by 99.9% in a couple of decades. The only reason it seems tricky is that most products don't do that.

"What's tricky about it? You have lots of things that used to be very expensive (cars, travel, canned food, computers, generic medicines). Are those tricky to measure?"

I want to say... yes? Like, if your approach to measuring changes in productivity over time is to look at a bunch of things that we consume that got way, way cheaper to make, and therefore cheaper to sell, and also better than the old versions and to just shrug and say "if productivity really went up there would be more dollars. Ima count the dollars", then you've set yourself an easy but profoundly unhelpful job.

And I think maybe you agree that the thing you say you're measuring is unhelpful? But to me that suggests an additional step to get you something that can be used to talk about whether the internet boosted productivity and which is, as I say, actually pretty difficult.

Yeah, with the commentators above, there is no measurement problem. Some people don't like GDP. Alternatives have been written up. Use them or make new versions. Full stop.

But the internet is not only music, or facebook. It is a means of communication.

The question that begs is whether communication in overwhelming amounts with no barriers or cost improves productivity.

I would suggest not. For many reasons.

1. It allows an illusion of control. If cross enterprise communication is cheap and possible we would see more of it. But more meetings do not equate to more productivity. Likely the opposite.

2. The centralization impulse was always limited by what was possible, but now massive amounts of data can be fed to central locations for decision makers and managers to do their magic. Again, what large centralized system has ever been characterized by innovation and productivity gains?

3. Software has become more complex and less useful. I have been amazed over the last decade how many companies I deal with who have implemented major software initiatives that have had the net result of making jobs that took 15 minutes into jobs that take 90 minutes. These are people who generate income and business, and they are burdened with generating the data and communication the idiots who run the company feel they can demand since it is so easy. I remember in the mid 80's watching someone update a database on an IBM Pc with a floppy disk. Likely a Dbase application. Crank crank crank, press a few keys, press enter, crank crank crank. Most enterprise software that I interface with is as bad as that. Productivity through software? Are you kidding.

4. The easily automated tasks were don in the 90's; warehouse management, word processing and many others.

5. The illusion of ease of data production produces demands for that production with no productivity rationale.

6. Communication technology makes it easier to move production somewhere else than automate it. A few mexicans or vietnamese doing the mundane tasks are cheaper and more flexible than complex and extremely expensive automation processes.

7. Software is extremely complex and ridiculously expensive to produce, meaning that existing software dictates how things are done, even if it is less optimal than what existed before.

I'd give it another two decades before we see productivity improvements. Walmart did their magic a decade and a half ago.

"The question that begs is whether communication in overwhelming amounts with no barriers or cost improves productivity.

I would suggest not. For many reasons."

I think you are wrong, and obviously wrong. You just haven't thought about it deeply. Just look at your everyday life versus the same situation 20 to 30 years ago.

1) When's the last time you were truly lost? I used to routinely get lost driving somewhere that was unfamiliar to me. No one was surprised when it took an extra 20-30 minutes to get to a party or shop or some such, because you were looking for the right street, then the right address and maybe stopping to find a pay phone to call.

Now, I've got a GPS and 3 different map programs on my phone. And if I need to call, I don't stop the car.

2) How do you find a business you need? Remember when we actually used the yellow pages. And the only thing you knew about a business was from a 1" Ad? Maybe you could call the Better Business Bureau and see if they were guilty of fraud. Now you hit the internet, look at the reviews and have a much better idea of how good they are and if they provide the exact service you need.

"3. Software has become more complex and less useful. "

TurboTax has taken my yearly tax preparation from a 2 day weekend task down to a Saturday morning task. Furthermore, I used to have to drive to the post office to mail the envelope on time. Now I E-File.

"4. The easily automated tasks were don in the 90's; warehouse management, word processing and many others."

Except for all those that weren't. Automation is an ongoing task. In 30 years, fast food restaurants will be mostly automated. It's quite possible that driving will be automated in 20 years. Manufacturing automation is on a spectrum, but it's continuously growing.

"A few mexicans or vietnamese doing the mundane tasks are cheaper and more flexible than complex and extremely expensive automation processes."

But a few Mexicans or Vietnamese or Chinese working in an automated factory will inevitably displace their manual counterparts. When automation is cheaper than labor it dominates. Automation is always getting cheaper for a given output. Labor got cheaper for a given output when we opened up the third world, but that's a once and done.

"7. Software is extremely complex and ridiculously expensive to produce, "

True, but it's unbelievably cheap to reproduce and sell. Software might cost $1 million to produce and that only ensures $100 in savings. But if you can sell a million copies of it for $10 each, it's all worth while.

If you swap out Spotify with any sort of file sharing service (aka bittorrent), does this mean that the effect of the Internet on productivity gains is negative? Or can we now finally ignore the RIAA/MPAA claims of massive economic harm?

For that matter, what is the cost in lost productivity by having to watch all those FBI warnings over the past couple of decades?

'Economist Chad Syverson probably has done the most to deflate the idea of major unmeasured productivity gains through internet technologies.'

One has to wonder if he compared the costs of communications and data transfer for globalized companies in 1985 compared to 2015. Or the fact that globalized companies pretty much rely on 'Internet technologies' to remain globalized companies at this point. And what is interesting, this part would seem to relate to globablization - 'Using a conservative estimate, the productivity slowdown implies a cumulative loss of $2.7 trillion in GDP since the end of 2004.' Makes one wonder just how large Foxconn's or Samsumg's productivity gains have been. Or maybe China and South Korea are outliers when it comes to increased productivity compared to 1985.

"If you swap out Spotify with any sort of file sharing service (aka bittorrent), can we now finally ignore the RIAA/MPAA claims of massive economic harm?"

The fact that musicians to provide loss leaders on YouTube or Spotify for free doesn't mean that they don't incur costs. It just means they're hoping to recoup those costs later on. They're making an investment, secure that they have a property right in the songs they produce in the future.

Sharing songs that are not available from the author for free (ala BitTorrent) violates that property right. Yes, there is real economic harm when you make it impossible for someone to enforce control over the product they produce, thus driving the price to zero. That's true whether it's authors, musicians, or drug manufacturers: if they aren't secure in the property rights of their future products, they will not invest in making those products.

In short, sharing songs on BitTorrent makes musicians making and giving away free music less likely.

I suspect many musicians would rather stay home and collect royalties from the sale of records, but who knows. I remember the day when my monthly federal express bill was large, very large. Now, not so much. Look at federal express's revenues over the past 20 years. What saved federal express? Online commerce. The internet made working remotely not only possible but maybe more productive, and cut down on traffic congestion and the resulting global warming to boot. On the other hand, working remotely and being alone may have contributed to the increase in depression and suicides, offsetting the productivity increase. Not to mention time wasted surfing the internet and obsessing on social media. Cowen's approach to the lackluster gains in productivity is similar to his approach to rising inequality: we are doing it all wrong, the doing being the measuring of it. Sure, Cowen may well be right, but the approach has a feeling of mild desperation in the defense of big business. Of course, global productivity and GDP have soared while America's have lagged, globalization producing a net gain. Or not. The gains have come at a high cost, as the carbon emissions in China and other developing countries with less regulation have also soared, offsetting much of the productivity and GDP gains if not risking the planet. On the other hand, the rising threat from China may induce America to significantly increase defense spending, offsetting some of the economic losses in America as the result of globalization. Indeed, a good hot war with China would require enormous public and private investment in infrastructure and other productive capital, triggering a long period of productivity and GDP growth.

"consumer surplus would have to be five times higher in IT-related sectors than elsewhere in the economy, which seems implausibly large."

If IT-related products have close to zero, or indeed zero, marginal cost, then why is this implausible? Price reflects marginal cost (zero) and, thus, the marginal consumer may indeed gain only negligible utility from consuming the marginal unit. *Inframarginal* consumers, however, may still gain much utility from consuming inframarginal units, which will be valued at zero for GDP purposes. That could be a lot of consumer surplus.

Think of a two-product economy producing apples and oranges. Initially, half the land is dedicated to producing 50 apples, the other half 50 oranges, and apples and oranges are both priced at 1. Nominal GDP=100. Now, suppose a technology shock allows one to produce unlimited apples from one tree. The price of an apple drops to zero, or close to it, reflecting its marginal cost. Apple production skyrockets to x >> 50. Almost all of the land gets re-allocated to producing oranges, doubling orange output to 100, in addition to the x apples now valued at 0. NGDP is still 100. The pre-shock GDP basket of 50 apples and 50 oranges in post-shock prices is worth 50, so the GDP deflator is 1/2. Real GDP has doubled, reflecting the doubling of orange production, but does not include the apple production. Worse, one might be tempted to dismiss the apple production because the post-shock apple industry is such a small fraction (zero) of the post-shock economy. There's a lot of consumer surplus because the first 50 apples are worth so much more than the x-th apple even though all are priced at 0.

One might avoid this problem by using the post-shock GDP basket, 100 oranges + x apples, instead of pre-shock basket to compute the GDP deflator. Using today's GDP basket might lead to pretty big computed deflation if one considers that, to replicate Spotify in the past one would have had to buy every CD in the world, to replicate Google one would have had to buy (and store) every publication in the world and hired a research staff to use and maintain it, etc. Of course, that's also unrealistic because people wouldn't have consumed post-shock quantities if they had to pay pre-shock prices. (I think Bryan Caplan wrote a blog post about this a while ago.)

Here is Caplan's post: https://www.econlib.org/paasche-prosperity/

You say it better than I could. Yesterday I used my phone to video a performance by one of my kids. Fifty years ago that video would have cost thousands of dollars to make and employed probably about 8 people for a week to set it up, do the actual filming, process the film and prepare it for me to view. How much of that progress is included in the productivity measurements? Almost none, because my time was free and the phone and the computer was paid for. Because according to today's prices the value was zero for the video I made. If all the videos made in the world were suddenly valued at 1955 prices, imagine the huge productivity improvement we would measure?

I took Professor Cowen's draft seriously and thoroughly read it. From a cursory glance of the "new" version against the mark up that I sent on to him, I see little change if any. The grammatical problems that I found in the original are still in this version. The authors should decide between 'labor' and 'labour' (just to cite one example). Maybe economists enjoy beginning sentences with 'And'

The next time an offer to take a look at one of his draft papers is extended, I shall decline.

Much of these effects are simply to improve on our standard of living. And how does one measure that improvement in dollar terms?

The value of the cellphone for instance, where we can now so easily now call and receive calls about anywhere. And texting. We instantly keep in touch with family, friends and business. Improved standard of living and huge values there.

None of those arguments actually work. None of those actually capture the value of additional free music.

- gdp deflator: even if done perfectly (impossible), only measures paid music
- spending elsewhere: measures something additional
- concerts: measures something additional

A long list of half-baked arguments about how your conclusion might be true... is called rationalizing.

Shouldn't the arguments around imputed rents have some analog to the digital economy? As I understand it, we include imputed rents into GDP because otherwise GDP would decline with increasing home ownership. That's the entirety of the reasoning, right? We don't want GDP to have that property.

Perhaps we should imput purchases since we don't want GDP to fall from increased subscription rates?

"Work by Erik Brynjolfsson and Joo Hee Oh studies the allocation of time, and finds that people are valuing free Internet services at about $106 billion a year. "

That seems incredibly low. Over 329M people that's about $32 a year. I'd be interested to see how they came to that number.

Not having to have a map, all the worlds knowledge at your fingertips, not having to go to the store to buy stuff, the list goes on.... I would say by having free internet I save well over $100 alone just parking, taxi and gas let alone the time I spend having going there and browsing around.

If I had to pay for google maps alone I'd probably spend more than $30 a year for the service but that's me...

"That seems incredibly low. Over 329M people that's about $32 a year. I'd be interested to see how they came to that number."

+1, I agree.

$32 is roughly 4 meals at a fast food restaurant. If I had to choose between skipping 4 meals or skipping the internet for the year, it's clear the internet would win out.

Isn't the idea that you're spending that money either way, and the value is found in other variables? E.g., you navigate to a hard-to-find restaurant, the extra time you spent driving, the extra gas, the extra snacks or bottle of water to hydrate on the road, etc, capture the new value created.

Compared to canned food, vaccinations, the refrigerator, the PC... the actual economic value of free Internet services may not be that high. After all, how much productivity is lost by responding posting on economics blogs instead of actually producing stuff. :-)

Could we be overmeasuring? Is all productivity growth equal? I don't feel like 280 characters is much of an innovation

While GDP may measure service, I think what hurts this economic measure the most is that it can't measure the quality of service. The internet provides people an opportunity to have a service at their door within five minutes. With services such as Uber, Lyft, GrubHub, and many others it's amazing how much production the internet has led too. The impact of these services is immeasurable because the services can come be of good quality one day and bad quality the next. People will take a ride from an Uber driver with a five-star rating rather than one with a two-star rating. The internet has provided us with a rating system for the quality of these online services and GDP fails to measure that.

We can do many different things on the internet now and all of that is without anyone's help. When we don't necessarily pay a middle-man for certain things, GDP goes down. While overall productivity may be going up, GDP doesn't follow suit in that case. There's many more examples of this but those are two that came to mind. When it comes to the internet, a lot of the time, productivity goes up while GDP goes down.

We are good at measuring GDP, but less good at measuring POTENTIAL GDP and the gap between Actual and Potential GDP.

We have been expanding the boundaries of potential GDP with new technologies, reducing the costs of obtaining, producing and servicing the population with goods and services.

GDP is measured in dollars. If the goods are expensive, and you only purchase one, that counts as much as if you purchase 2 of that item at half the price you previously paid.

Comments for this post are closed