The Productivity Paradox
Robin Gaster PhD
Despite all the productivity we see around us, most mainstream economists believe we are in era of low productivity growth, reflecting the reported productivity numbers. That’s a problem for the Great Disruption argument, because if productivity growth is low, then surely talk of disruption is at best overblown. Maybe the robots aren’t coming after all.
Those of who believe the data – “true believers” – offer several explanations for this disconnect: low productivity growth in the service sector which dominates the economy; innovations that aren’t sufficiently important to move the economy; lags between innovations and the subsequent diffusion of innovation; and growing gaps between leading companies and lagging ones in many sectors of the economy, fueled perhaps by a disinclination to invest.
But maybe these true believer theories explain outcomes that don’t really exist. Maybe the productivity data are bad.
On closer examination, there are a lot of problems with the data. The data are poor at capturing both quality changes and new products. They are especially bad at measuring services, where it’s hard to conceptualize outputs, let alone measure them. They struggle mightily with free (Facebook/Google) and near-free (Netflix) services. They completely fail to address nonprofit and government sector productivity, and don’t include nonmarket activity (like environmental gains). They don’t account properly for rapidly growing investments in intangible assets. And they miss out entirely on the black economy and on much of the gig economy. Together, these arguments easily explain the “missing” 1.5 percent of annual growth that separate low and high growth economies.
The heretics can further integrate some of the true believer arguments: it may be that productivity is indeed depressed by lags and the growing innovation gap. But the overall conclusion is inescapable: the productivity paradox is mostly driven by bad measurement. So there is nothing here to suggest that the Great Disruption is unreal, or even that it will be much delayed.
Summary of the argument
Daily, we encounter new apps, platforms, and tools. Daily, we see extraordinary changes in the technology around our lives and only somewhat slower in the economy around us. The primary argument in this book is that innovation – aligned with globalization and shifting corporate culture and structures – will within the next decade fundamentally change our work and hence our lives.
But if that’s the case, why is there no sign of innovation in the productivity data, the first place that economists would look for it? Writing in 1987, Nobel prize-winner Robert Solow famously remarked that “you see the computer revolution everywhere except in the productivity data.” That’s the “productivity paradox:” despite all this apparent innovation, productivity – as measured through Federal data – has been relatively low since 2005.
In fact, since WW2, there are have been four distinct periods of productivity growth: high growth during the post-War boom 1947-73 (2.7%); low growth 1974-1994 after the oil shock and stagflation; high growth again 1995-2004 (2.8%); and finally very low growth since 2005 (1.3%) (see above).
So if innovation is all around us, why don’t we see an explosion of productivity?
We need to start by understanding the basics: do we live in a world that is basically high growth with some low growth periods? Or low growth with some high growth periods?
Low growthers argue that the post-War boom was basically an anomaly, that US growth “normally” runs at 1-1.5% annually. That boom was caused by pent-up demand from after the end of the Depression and the War years plus minimal foreign competition and newly opened export markets. The periods after the oil shock, and more recently after 2004, are thus a return to normalcy. The recent high growth period (1995-2004) resulted from one-time adoption of information and communications technologies (ICT) especially in manufacturing. So low current growth is just reversion to the mean, now that the low-hanging fruit of the computer era have been harvested.
High growthers argue that the period after WWII represents the norm. Low growth after 1973 was caused by from the oil shocks and subsequent macroeconomic stress, notably stagflation in the late 1970s and early 1980s. And slow productivity growth since 2005 results mainly from lags in the adoption of new technologies, perhaps because lagging firms don’t keep up, and many firms fail to invest sufficiently. So they expect growth to accelerate as innovation spreads across the economy.
Both high- and low-growth explanations rest however on the same bedrock assumption: that the productivity data accurately reflect the state of the economy. Explanations vary, but what needs to be explained does not.
Critics – data heretics – argue that the productivity data are, under closer scrutiny, far less sturdy than they appear; that they explain only part of the economy; and they exclude its most dynamic elements.
The Great Disruption: an upcoming period of rapid and extreme disruption in the economy, caused by the interlocking impact of technology (especially automation), globalization, and increasingly monopolistic corporate structure serving executive compensation and shareholder value. The impact will be felt especially in labor markets as existing jobs become obsolete in a wide range of industries and occupations.
If the data heretics are right, then economists who reject the Great Disruption argument because “productivity growth just isn’t there” are relying on a metric whose time has passed. They will still be out there looking for innovation in the productivity data while buying their groceries online and accepting deliveries from autonomous vehicles, leaving them time to take another free online cooking course while using their phone to book flights to Tahiti. The old economy will have morphed into an entirely new one, unnoticed.
True believer arguments
True believers argue that the data are real, that productivity growth has slowed, and that there are good explanations for observed trends. Those explanations differ somewhat, but include some or all of the following arguments:
- Slow productivity growth in fast-expanding sectors. William Baumol argued that as the service sector grows, productivity inevitably slows. For example, outputs in education and health care are not growing as fast as inputs, so we get declining overall productivity as sectors like these expand.
- Secular stagnation – there just isn’t that much innovation, and it’s not that important. Robert Gordon and Tyler Cowan among others see limited innovation today, at least compared to the enormous wave of innovation during what Gordon calls the Second Industrial Revolution of the early 20th There are plenty of nuances and sub-arguments, but the basic claim is simply that today’s innovation is less important, has less impact, and we get lower growth as a result.
- Innovation is diffusing slowly across the economy. Other researchers argue that innovation is powerful and it is coming, but that it just takes a long time to diffuse through the economy. This position seems to be gaining traction. After all, previous waves of innovation also took a long time to fully roll out across the economy.
- Gaps between innovators and lagging firms are growing. Diffusion may be slower partly because the gaps between leading and lagging firms in each sector are growing. Essentially, overall productivity is held down even if leading firms are innovating rapidly, because zombie firms that are just getting by are able to survive. Essentially, while we see innovation around us, we don’t see all the firms that are not innovating as much. The innovation gap itself may be growing because many firms are reluctant to invest, especially in innovative ideas.
- Labor market rigidities and mismatches. Finally, there are explanations rooted in the labor market. Firms claim that it’s hard to hire for the right skills; workers are today clearly more reluctant to move geographically; and there appears to be a growing mismatch between the education system and the economy.
All these explanations are rooted in real economic drivers. Together, they offer some pretty good reasons why the apparent innovation all around us is just not having that much effect on the economy, and hence on growth and productivity.
Without the “productivity paradox,” without the data showing that productivity is growing only slowly, the true believers would have nothing to explain. So if the data are bad, these explanations become largely irrelevant or at least secondary.
Start from a common-sense crosscheck. It just seems unlikely that productivity numbers flattened out exactly when computing became ubiquitous across the economy, smartphones started taking off, and new technologies and business models began to disrupt whole sectors. The chapters on retail, finance, and the gig economy all show something quite different: that we are right in the tube of an enormous wave of disruption. So if official productivity data show only a flat calm, maybe we are measuring the wrong things.
Data heretic arguments
The heretics argue that we see lower productivity gains because we are looking in the wrong direction, at the wrong data. They explain this in different ways:
- The quality problem. Productivity data don’t capture changing quality and innovative products very well. Despite heroic efforts at the Census Bureau, adjustments for changing quality (“hedonic adjustment”) are too narrow and too arbitrary. This is an enormous problem precisely because we are innovation boom: the larger the economic impact of new and transformed products and services, the bigger the gaps not tracked by the productivity data.
- Outputs from the service sector are hard to measure. Factory output is relatively simple to measure. But in the service sector, its hard to even define outputs clearly, let alone measure them accurately. What is the output for a school and how would you measure it? Or for a hospital? Less medical treatment directly reduces GDP – but is that a good thing or bad thing? Health care and education together account for 25% of the US economy, so mismeasurement in these sectors has a big impact, and they are not the only hard to measure and fast- growing sectors. Are more stock-market transactions a good thing?
- GDP measures transactions – but the free or nearly free economy is growing fast. The more time we spend on activities that don’t involve transactions, the less useful standard productivity data can be. Millennials are important canaries in this coalmine: they spend less on goods and services, and fill their time with free (Facebook) or nearly free (Netflix) activities. A British study recently found Millennials used their phones more than 5 hours a day. More widely, an increasing share of our economy is transaction-free (or nearly so) and hence excluded from GDP. Productivity data simply don’t capture much of the digital revolution.
- White-collar time capture. White-collar workers use computers and smartphones to re-appropriate the fruits of their labor. Small increments in productivity are acquired by workers for use in their own interests – surfing the web, Facebook, shopping, porn, and other hobbies. According to one survey, about a fifth of employers think their employees are productive for less than 5 hours a day. These productivity gains are hidden from employers and government data alike.
- Productivity gains in nonprofit and government sectors are excluded. Productivity data exclude sectors where outputs are not valued through a competitive market. These includes government, nonprofits (about 5% of GDP), and the benefits of environmental regulation (which cost about 1% of GNP according to NAM, but whose benefits are likely much larger). Government services have been transformed by technology already – but there is no way to reflect that in the data.
- The gig economy messes up the data. The informal economy is large and growing (especially in the form of second freelance jobs – “side hustles”). The black economy is completely excluded from the data, and only a fraction of the work performed by independent contractors is reported. These informal work arrangements are growing rapidly and probably cover 20% of all US workers.
- Intangibles are increasingly important. Various kinds of intangible investment are not fully captured by national accounts – especially what are called “economic competencies”: investments made by firms in their own capabilities. Overall, intangibles missing from national accounts have been estimated at about 4.6% of GDP.
I think about the productivity paradox like this. Productivity data were initially set up to measure the manufacturing-dominated economy of the pre-War era. They have been adapted since then, but can still measure only part of the economy, and that part not very well. In the digital age, these metrics miss a growing share of modern human activity. Quality changes are difficult to conceptualize let alone measure. Service sector productivity is often not definable or measurable. The rapidly growing free economy is entirely outside measured GNP, and so is the informal sector which is also growing. And the government and nonprofit sectors are excluded from productivity data altogether.
These large and multiple effects can easily account for the “missing” productivity since 2004 – which, may I remind you, is only about 1.5% of GDP annually. Each of these explanations could on their own account for that much unrecorded productivity.
A student of social affairs who is interested in the total productivity of the nation, including those efforts which, like housewives’ services, do not appear on the market, can therefore use our measures only with some qualifications.
– Simon Kuznets, inventor of GDPThe productivity numbers are a bad metric for a modern economy. Nominal GDP is presented with false precision, given the gaps described above. “Real GDP” is even more of a guess, taking nominal GNP and then deflating by the cost of a set basket of goods whose accuracy and relevance is questionable. Large and growing swaths of economy are captured poorly or not at all. And the denominator for calculating labor productivity is off as well – the large and fast-growing gig/informal sector, accounting for a significant percentage of labor activity, is badly tracked and mostly excluded from the employment count.
There is also an asymmetry here.
True believers fundamentally reject heretic arguments as a solution in search of a problem: if the data are good, arguments based on attacking the data are inherently unnecessary.
Heretics however can accept many of the explanations provided by true believers. They just have to claim that the data are wrong, not that all the arguments presented by the realists are wrong. It may be both that there is innovation all around us and that it is having a profound effect on the economy, and that Baumol’s disease exists, that there are growing gaps between top and trailing firms in a given sector, and that that there are longer-than-expected lags in new technologies fully impacting the economy.
For myself, I would argue that the data are inadequate for the many reasons listed above, that we are indeed experiencing a wave of innovation, and that some of the issues explored by the realists also throw important light on future paths for the economy. The expanding gap between top and trailing firms within sectors, the impact of growing concentration, and diffusion lags for new technologies (perhaps driven by insufficient investment) all seem plausible. But the grim world of secular stagnation does not.
One final point. If the heretics are right and GDP is actually accelerating on the back of automation and globalization, that may bring us even faster to a world where jobs become harder and harder to find? Labor productivity may go through the roof, but that doesn’t mean that the economy is serving us well. As Simon Kuznets – the originator of GDP statistics – warned at the time,
“the welfare of a nation can scarcely be inferred from a measure of national income. If the GDP is up, why is America down? Distinctions must be kept in mind between quantity and quality of growth, between costs and returns, and between the short and long run. Goals for more growth should specify more growth of what and for what?
Focusing on GDP and reported productivity growth means that we miss much of what is dynamic about the economy. So we are less prepared as the wheel of innovation accelerates, threatening to throw so many of us out of the working economy altogether.
True believers: the productivity data are real
Those who accept the official productivity numbers have to explain why all the innovation that we see around us somehow doesn’t matter. They offer a number of solutions.
In 1966, William Baumol observed that wages in industries with no productivity growth still tended to rise in line with wage increases in more productive industries. He famously noted that it took the same number of musicians the same amount of time to play a Beethoven string quartet as it had in the 18th century, but that musicians still had to be paid the going rate rather than what they were paid by Beethoven.
Baumol argued that in some sectors – especially services – technological change doesn’t generate much increased productivity, and that the slow growth in these sectors weighs down average productivity growth across the entire economy. Baumol’s Disease has since been used to explain low productivity gains in healthcare, education, government and other services. These sectors now dominate the US economy. If productivity is flat or even falling in these sectors, overall US productivity growth will be low too.
Baumol’s Disease seems like a good explanation especially for sectors where prices (costs) have exploded, like higher education and healthcare. In constant dollars, the cost of acquiring a four-year undergraduate degree at an in-state public university has tripled over the last 40 years. Yet the value that college graduates get from this expense is growing much more slowly (if at all), so higher education productivity (output value divided by input costs) (and excluding university research) seems to have declined.
There are actually two arguments here, one about innovation and the second about patterns of wage behavior. The second in easily verified: there does not seem to be much correlation between sectoral patterns of wage growth and recorded productivity growth. As the cost of labor inputs rises, measured productivity will fall unless more output is generated – and Baumol argues this may be hard to do in the service sector (which is 85% plus of our economy).
The first claim – that there is little innovation in these sectors and hence not much output growth – is much more contentious. Has there really been so little innovation in health care, in entertainment, in transportation?
New business models and business processes matter at least as much as a new digital widget. Walmart has transformed the retail sector, for example, without manufacturing anything. The cost of an online bank transaction is $0.20, while a teller transaction costs $4.25. Is that really nothing?
Detailed industry level research also shows that at least for the high growth period up to 2004, labor productivity in service industries grew as fast or even faster than productivity overall. 
Baumol’s argument becomes even more important as the economy tilts further toward lower productivity sectors: among the top 5 fastest growing occupations in the US are services for the elderly and persons with disabilities, offices of physical, occupational and speech therapists and audiologists, and home health care services – all service providers with apparently low productivity growth.
The way out of Baumol’s Disease is to look harder at what is being produced: after all, those musicians playing Beethoven now generate a performance that can be recorded with high fidelity. 100 years ago it could be recorded on low fidelity vinyl, extending the reach of that performance to thousands of new listeners. Soon after, that recording became available on the radio, reaching millions. 60 years ago the sound improved radically as high-fidelity recordings became were available. Fifteen years ago music became personal and portable first with the Sony Walkman and cassettes, and then more broadly with CDs. Today that recording is streamed instantly to every corner of the globe. And the cost has declined to near-zero.
Perhaps those musicians are a lot more productive after all.
Innovation slowdown and secular stagnation
Another argument, presented best by Robert Gordon, simply claims that productivity growth is down because innovation is down: the innovations of today – impressive as they seem – are chickenfeed compared to the great innovations of the 19th and 20th centuries. Maybe electrification of the economy was fundamentally much more important – and had a much greater impact on growth – than digitization and the rise of the Internet.
Gordon correlates the higher economic growth and productivity gains in the middle of the 20th century with what he calls the Second Industrial Revolution – electricity and modern transportation in particular. Gordon’s remarkable analysis shows in particular how some hugely important changes followed – notably urbanization and the mass entry of women into the workforce – and argues that these cannot changes are one-offs that cannot be replicated. There are no more women to add (more or less), and we are now already highly urbanized.
Even if innovation were to continue into the future at the rate of the two decades before 2007, the U.S. faces six headwinds that are in the process of dragging long-term growth to half or less of the 1.9 percent annual rate experienced between 1860 and 2007. These include demography, education, inequality, globalization, energy/ environment, and the overhang of consumer and government debt. A provocative “exercise in subtraction” suggests that future growth in consumption per capita for the bottom 99 percent of the income distribution could fall below 0.5 percent per year for an extended period of decades.
– Robert Gordon (2012)But Gordon then argues that the cluster of innovations associated with digital technologies have had less impact and have been unable to overcome what he sees as an increasingly negative economic environment, characterized by six headwinds: “demography, education, inequality, globalization, energy/environment, and the overhang of consumer and government debt.” That’s much more controversial.
Tyler Cowen offers a similar argument. “Secular stagnation” now dominates the economy because the low-hanging fruit of technological change that impacts productivity have all by now been picked. He further argues that the productivity slowdown after 2005 is so big that mismeasurement is simply not large enough to provide a compelling story. Cowan says that
The tech economy just isn’t big enough to account for the productivity gap. That gap has caused measured G.D.P. to be about 15 percent lower than it would have been otherwise, yet digital technology industries were only about 7.7 percent of G.D.P. in 2004. Even if the free component of the Internet has become more important since 2004, it’s hard to imagine that it is so much better now that it accounts for such a big proportion of G.D.P.
Cowen also observed that the productivity slowdown appears to be affecting many countries, including those with smaller high-tech sectors.
However, by focusing only on “digital technology industries,”: Cowen misses much of the real impact of new technology. All sectors are tech sectors today – they are tech-using sectors, even if not tech-producing sectors. The productivity story outside high tech sectors is what really matters. If the job of bank teller is eliminated, or Wendy’s implements its threat to replace order-takers with machines, there may be demand effects within the IT-producers, but the real and direct impact will be outside, in the tech-using sectors.
Is Gordon right about the unique power and reach of the Second Industrial Revolution? Is Fernald correct to argued that the gains from IT were effectively a one-time benefit that did not and will not translate into ongoing productivity growth. Are the low hanging fruits gone for good as Cowan believes? I think not, but these are powerful arguments.
Timing and innovation lags
When IT investments in the 1980s didn’t show up in the productivity numbers, some economists argued that this would simply take time. And indeed, productivity numbers did improve after 1995. But maybe what we are seeing now is simply lags that we should expect before the impact of another wave of innovation fully deploys across the economy and hence reflected in the productivity numbers.
Paul David’s well-known article on the dynamo and the computer illuminates the decades needed before electricity was fully deployed across the US economy, and the long delays – and industrial reorganization – needed before manufacturing processes fully integrated the new technologies of the Second Industrial revolution. David claims ICT technologies are similar, and hence that we can expect more growth as they are more fully deployed.
There is, obviously, plenty of evidence that innovations take time to fully diffuse across the economy. Figure 3 (from Goldman Sachs) illustrates adoption lags for 10 major innovations. And the “lags explanation” has been accepted both by Eric Brynjolfsson – a proponent of a radically transformed digital future – and Chad Syverson, generally a true believer who is much more cautious on anticipated transformation. They argue that “The most impressive capabilities of AI, particularly those based on machine learning, have not yet diffused widely. More importantly, like other general-purpose technologies, their full effects won’t be realized until waves of complementary innovations are developed and implemented.”
Obviously, there is some truth to this. As William Gibson famously said, “The future is already here – it’s just not evenly distributed.” What a select few have today will be commonplace tomorrow, to be followed by yet more innovations not yet even in prototype.
This second economy… is vast, silent, connected, unseen, and autonomous (meaning that human beings may design it but are not directly involved in running it). It is remotely executing and global, always on, and endlessly configurable. It is concurrent—a great computer expression—which means that everything happens in parallel. It is self-configuring, meaning it constantly reconfigures itself on the fly, and increasingly it is also self-organizing, self-architecting, and self-healing.
– W. Brian Arthur (2011)
So maybe if we are patient, measured productivity will at some point reflect the full integration of new technologies into the economy. However, critics of the lag argument like Robert Atkinson have pointed out that adopting ICT is much easier than electrifying: These technologies are designed for use by nonexperts and require much less physical adjustment than electricity ever did. And Brian Arthur’s concept of the hidden “second economy” – communication entirely between machines (see box) describes a wide range of extraordinary electronic innovation (e.g. electronic ticketing for airlines) that was already in place more than half a decade ago.
These lags may also be affected by policy. Atkinson lays much of the blame for low productivity growth on low investment especially in ICT, arguing that “quality adjusted nonresidential equipment and software investment peaked in 2001, fell, stabilized until 2007, and then fell again.” Possible explanations for low investment since the early 2000’s include competition effects and an increase in short termism among corporate leaders. (these issues are discussed in an upcoming article on corporate structure and culture).
The low investment argument can be viewed as an add-on to technology lag. Perhaps companies will eventually invest. And productivity will grow faster. But listed firms have for the past ten years focused on returning funds to stockholders, and growing industrial concentration has led to growing piles of cash at large companies. So maybe investment is low because on balance companies prefer other uses for their money.
Yet in the end the lag theory is not that convincing. More investment might accelerate innovation, and perhaps lags are the best available explanation for observed productivity numbers. But that just makes lag theory the best theory available for explaining bad data. And it’s not really clear that diffusion is so slow. E-ticketing for airline passengers went from 30% of tickets in 2005 to effectively 100% in 2008. That’s just one example, but fast adoption is a characteristic of digital technologies (see chapter on Amazon and retail). The lag theory doesn’t make the productivity data any more believable.
Bigger gaps between leading firms and the rest
Some firms are just much more productive than others, and if the gap is growing, that could help explain flatter overall productivity. Dan Andrews and his colleagues at OECD argue that while leading firms continue to innovate, and the productivity of the most productive firms continues to accelerate, those innovations are not being adopted as quickly (or sometimes at all) by lagging firms. On this reading, adoption lags are reflected accurately in the productivity data, and outweigh the productivity gains of the leading firms.
This is intuitively a compelling argument. We are all aware that some firms are more efficient than others. And this could mesh well with the lag argument described above.
If Andrews et.al. are right, productivity will accelerate again if lagging firms are forced to improve or forced out. If they can survive and even prosper despite poor productivity, then lagging firms may continue to drag down overall productivity. If on the other hand firms have to adjust or are forced out of business, overall productivity will begin to accelerate.
This argument ties into other work. Ryan Decker and his colleagues note that “the pace of business dynamism in the U.S. has fallen over recent decades and that this downward trend accelerated after 2000.” Declining dynamism affects productivity both because fewer startups means less innovation, and because more failing firms somehow cling to life. The impact of zombie firms on Japan’s stagnant economy is well known, and this effect is felt in other countries as well. In Italy the share of the capital stock sunk in “zombie firms” rose from 7 to 19 percent between 2007 and 2013.
Another related argument comes from industrial concentration. Industries are consolidating, and concentration is growing in the US. That reduces competition and hence incentives to adopt innovations. More recently, arguments have been made that the explosion of passive investment through huge funds run by firms like Vanguard and BlackRock reduce inter-firm competition as these funds invest in all large firms across a given sector (e.g. airlines), so incentivize management to grow profits at the expense of competition (and hence presumably innovation).
Finally, the digital world may be especially prone to expanding the gap between leaders and laggards. Given the enormous resources concentrated in the leading firms, it is hard to see how competitors can even begin to the close the gap. Google, Facebook, Amazon, Apple and a few others (in selected sectors) dominate, and their ownership of scale and network effects means that they can innovate at a speed that competitors cannot match. It may also be that these firms are especially good at forcing lagging competitors to the wall (or acquiring them). Andrews and his colleagues argue that the growing gap especially in ICT may reflect increased difficulty in lagging firms catching up – especially in a networked, digital economy.
Clearly, the expanding productivity gap argument has important elements of truth. It may be that we are biased our own experience with the extraordinary pace of innovation near the productivity frontier, and cannot see the thousands of lagging firms lurking in the shadows.
Other true believer explanations
Given the gap between leading and lagging firms and industries maybe part of the problem is that labor is not shifting into higher productivity industries and firms. High productivity digital firms have limited needs for labor: Google looks nothing like its big industrial predecessors like General Motors (GM). Google employed only 74,000 staff in 2017. Even after years of downsizing, GM employed almost three times as many – 209,000. And sectoral effects matter: about one third of the productivity gap in the UK is explained by weak productivity growth in the oil and gas and financial service industries. So maybe it’s just that the fast-growing firms and sectors are also technologically advanced and need fewer people than lagging firms and sectors.
The growing demands of the knowledge economy may also generate increasing mis-matches with the human capital of workers – a core source of productivity growth. About half of all recent college graduates are working in jobs that don’t require a college degree. And college education may actually offer limited skills development, serving instead just as a signaling mechanisms for potential employers. Technical training (apprenticeship) is less valued in the US and apprenticeship programs are far less effective than in some North European countries. This matches the claims from some high-tech companies that there is a STEM worker shortage.
But this argument too needs to explain why things suddenly got worse starting in 2005. Perhaps the demands of the knowledge economy really are different. However, it seems unlikely that the mismatch would suddenly have become so much worse.
Heretics: the productivity numbers don’t add up
Official productivity numbers measure output per unit of input. Labor productivity measures output per unit of labor used to produce it. The heretics argue that these productivity estimates are too low because they simply don’t capture well what is being produced.
Accounting for changes in quality
It’s blazingly obvious that the quality of goods and services continues to improve, sometimes massively so. No one would prefer to be the richest man in the world circa 1800 rather than a middle class American today. No one today would buy even a top of the line car, sound system, or phone from 1980 if they had the choice of a current model instead. Yet in nominal terms, none of these items actually costs any more than the 1980s versions.
Quality changes are from a productivity perspective conceptually identical to changes in quantity. Using the same number of workers to make a product that is twice as good is, from a productivity perspective, exactly as productive as making twice as many of the original products. But it’s much easier to count the number of widgets manufactured than it is to measure the improved quality of each widget.
Government economists make “quality adjustments” to GNP. For some products these adjustments (known technically as “hedonic adjustments”) try to capture physical changes in the product – the size of the screen for a TV, or the thread count of a suit. (See a complete list of hedonic adjustments here). The list includes housing, clothing, and some audio-visual equipment. No cars, no computers, no smartphones – and no healthcare, telecoms, education, or any other services. And we import almost all of our clothing and audio-visual equipment – so that’s not actually produced in the US.
Quality is also adjusted by switching out old for new products. As one BEA analyst pointed out to me, while smartphones are still phones they are also a lot of other things and it makes sense to create a new consumer expenditure category to contain them. Yet many observers – including Alan Greenspan – have pointed out that the basket is a lagging indicator of reality. Nordhaus shows how long it took before lighting was properly included, and cell phones were excluded from the basked for many years. Clearly, the faster consumer behavior changes, the more difficult it is for BEA to keep up.
Adjustment for quality is hard to do at all, let alone well and consistently. Even in the case of a single easy-to-identify manufactured product like a television, it’s hard to see a basis on which a consensus weighting could be adopted: take a $900 12-inch color television from 1964 with three channels, and a $900, 54-inch flatscreen LED with surroundsound and 300 channels in 2018. How much has the quality improved? Twice? Five times? 50 times? Arguments can be made for all three, and it’s difficult to see any objective basis for deciding which to pick. The methodology used by BEA decomposes a TV into its components – e.g. screen size – and assigns weights to each of these components. For example, if the price of a TV doubles, but the screen size more than doubles, adjusted for quality the price of the TV has declined even though in cash terms the price has increased (perhaps by a lot). But that ignores a key reality: what matters is what you can watch, and that has changed in ways that no simple quality adjustment can capture.
The quality and novelty problem is worse for services. Take music for example.
Before the Internet, free music was available on the radio, supported by advertising. The number of channels (stations) was very limited, and listeners had no control at all over the music being played on a radio station. Today, services like Spotify offer essentially unlimited channels, and users have enormous control within a given channel. Essentially, the music you want is available on demand. Spotify is freemium service, mostly free and supported by advertising, just like radio was (with a paid subscription option to eliminate ads and get some other benefits).
Even if the amount of music delivered was identical (and of course Spotify delivers far more), and the quality was identical (but of course Spotify offers users much more choice and control), Spotify is massively more productive: it employs 1,600 people to deliver music to every Internet-accessible consumer in the world. In 2013, even after a decade of decline, US radio stations employed about 100,000 people. So taken as a whole, radio offers one tenth the productivity of Spotify and has to charge a lot more for much less targeted advertising. That’s why the free music radio audience is declining rapidly.
From a national accounts perspective, Spotify has driven down measured GNP. Advertising revenues for free radio stations have fallen sharply, and Spotify has acquired only a fraction of that amount. So the Spotify story is one of massive growth in usage by consumers, a far superior product, radical improvements in the company’s labor productivity, and better targeted ads (which hence need to be deployed less often) – all accompanied by a decline in GNP, and a substantial decline in radio-related employment.
Or take photography. Hal Varian points out the digital revolution here has increased quality, increased ease of use – but has reduced measured economic output – and hence productivity. In 2000, the world took (or at least processed) 80 million photos. In 2015, it took 1.6 trillion digital photos, which required no processing. This 2,000-fold increase actually reduced US GDP, as film and processing sales fell close to zero (and Kodak went bust) and photos stopped being a product and instead became something shared – more than 3 billion photos are now shared daily. So smartphones became the new cameras, enormous new nonfinancial transactions came into existence, but GNP declined as people stopped buying both cameras and film.
Or take telecoms. In Britain, data usage grew by around 900% between 2010 and 2015, for example, but reported real Gross Value Added (GVA) for the British telecoms industry fell by 4%. Telecoms companies made up for the decline in traditional voice revenues with new data revenues but were able to only capture a fraction of the utility generated by digital services, and the fall in voice revenues (in Britain) more than offset any growth in data revenues. Hence GVA declined.
In a nutshell, technological innovation is increasingly taking the form of new and improved products and services rather than just making the old wheat-and-steel economy more efficient, and productivity statistics are just not designed to reflect these innovations.
– Jan Mokyr
Productivity accounting just handles new products poorly. It is designed to measure relative changes in inputs and outputs in the course of production, but that’s possible only if the thing being produced remains constant. New products cannot be included, because those products were not included in the baseline against which growth is being measured. As Jan Mokyr points out, “Product innovation has been with us for a long time, but … its relative importance has been particularly pronounced in the past 20 years. And if that’s the case, productivity statistics systematically under-measure the rate of technological progress and its implications for economic welfare.”
Measuring service outputs
An even larger-scale example is K-12 education. Let’s assume initially that per-student outputs don’t change, and that student quality remains constant. If we reduce the number of students per teacher, that effectively reduces productivity: it takes more teachers to complete one student-year of education, and we have no means of adjusting output to reflect what should be better quality teaching. Similarly, if we add administrators, or accommodate learning-disabled students in small classrooms, that reduces measured productivity.
To show an increase in productivity – aside from firing administrators and failing our disabled students – we would have to show that the student being “produced” by the factory school is better quality. But the metrics we now use to measure student quality are rubbish, so we have no real idea whether “better” students are being produced. There is certainly no effort at all to include such metrics in GDP calculations.
Critically, this is a story about how difficult it is to measure productivity in a service economy. There is simply no good way to measure the quality of education from a national accounts perspective. Indeed the question itself seems faintly ridiculous. But if that’s the case, then how are we to measure productivity in an economy that is 85% services?
The examples above are just the tip of the iceberg. As Timothy Taylor points out,
When thinking about cost of a “unit” of health care services,” or a “unit” of banking service, or “unit” of legal services, it’s quite hard to think about what the “unit” should be. In health care, for example, a day in a hospital, or a specific procedure like a colonoscopy, are quite different in their qualities now than they were a decade or two ago.
So quality is very hard to measure. But its worse than that. that’s not all. From a GNP perspective, more colonoscopies are a good thing! More coloscopies equals higher GDP. But of course that’s not true for those of us on the receiving end. A world without colonoscopies at all would be much better (assuming that we found better and less invasive ways to monitor our colons). Doing more colonoscopies raises GDP, but in what sense is that producing health? A recent study estimated that 20% of all procedures are medically unnecessary. One fifth! But completely eliminating them would reduce reported productivity in health care by about the same amount.
Economists like using GNP data (and hence productivity data) because it gives the illusion that growth can be quantified, that it matters whether official growth is 1.3% or 2.5%. and because they take it seriously, it does matter: US federal and state budgets are heavily influenced by predictions of future GDP growth. But if we cannot even define what is being produced for much of the economy, or what unit to measure, these small differences are lost in a mist of fuzziness.
Clearly, there are major problems in measuring those parts even the economy where there are at least reported transactions to measure. But the biggest changes are probably happening elsewhere.
The free and Nearly-Free economy
Productivity and official measures of growth are designed to reflect transactions. Not utility, not the pleasure or benefit generated, only transactions. The digital/smartphone revolution is to a considerable extent all about the transition from the formal paid economy to unpaid individuals. Right across the service sector, digital capabilities are enabling individuals to perform for themselves things that used to require the direct or indirect purchase of services from someone or some company (this is sometimes called the self-service economy).
Some mainstream economists (e.g. Byrne et.al.) simply assert that the huge benefits of the digital economy – its transformative power – is irrelevant to the economy because its applications are not part of the transaction economy, or only a small part of the value of these applications is captured by the productivity data: “many of the tremendous consumer benefits from smartphones, Google searches, and Facebook are, conceptually, non-market: Consumers are more productive in using their nonmarket time to produce services they value.” The statement is true – but the conclusion drawn is not.
Varian’s point about photos is just the tip of the iceberg. We know that smart phones have totally transformed our lives already, but it’s worth itemizing a few of these capabilities to see just how deeply these changes have taken hold, and also how poorly they been captured and official data. Ahmad and Schreyer write that
Perhaps the best example is the use of internet search engines or travel websites to book flights and holidays, previously the preserve of a dedicated travel agent. But there are many other examples that merit consideration under this broad umbrella where market production blurs with non-market activity: self-check in at airports, self-service at supermarkets, cash withdrawal machines and on-line banking to name but a few.
We can all see that digital tools let us do what we previously paid for – or were not even able to do. These new digital capabilities make us more productive but are not captured as “productivity” by traditional ways of measuring. The self-service economy is not one tracked in GDP.
Some economists argue that these capabilities are like any other non-transactional activities like cooking, cleaning, or parenting, and should be treated in the same way. But this seems mistaken. Digital is different because it lets individuals replace existing nondigital services – travel agents or bank tellers with self-service capabilities. Unlike home cleaning or parenting, these services used to be part of the transactional economy.
Digital services eliminate (“disintermediate”) middle men, replacing their work with new consumer capabilities. To book a flight online, the work I do is unpaid, so the net impact is to replace paid work with my free work which is not included in the national accounts.
This adds up: If we assume that booking a flight takes an average of 20 minutes on the phone or in person, and that the booker was historically paid perhaps $15/hr, the shift from paid to unpaid labor reduces GNP by $5 flight booked. There are about 90 million airline passengers ticketed annually in the US, so the switch to self service reduced reported by GNP by $360 million annually. Of course, airlines may find the new system more profitable; it may lead to additional passengers; airlines and web sites invest in additional booking technology; there may be other downstream impacts. But just looking at the booking process, it’s pretty clear that GNP is reduced – quite substantially – by the innovations surrounding online booking.
Not all digital examples are so straightforward. Airbnb rentals for example are captured by national data only through Airbnb revenues (a fraction of the total transaction) and as renters report their new income (to an unknown degree). Airbnb expands the transactional economy by providing a platform through which hitherto private assets become part of the marketplace although, the hotel industry claims that it also replaces hotel reservations with a cheaper alternative. The overall impact is not clear. Drawing previously dormant assets into the transaction economy increases GNP, but replacing hotel rentals reduces it. In the short time since Airbnb took off, hotel rentals haven’t suffered, so the net effect has probably been positive.
Services like Facebook and Google offer deals that involves no payment by individuals – another challenge. Google and Facebook monetize their connection to consumers by selling advertising that is more carefully targeted and hence (they argue) more effective than traditional media. This is valuable to advertisers who also of course find that Google and Facebook have unmatched reach in their capability to put relevant ads in front of consumers. To a considerable extent, they are the digital market for advertising.
Advertising sales are of course included in GNP. But does advertising revenue capture the full value of these enormous new services? Of course not.
How can we determine the value of these services given that the user pays nothing directly? One way is to estimate the amount of time saved by these services. A University of Michigan study found that a Google search saved on average 15 minutes – and Google (as of January 2018) processes 3.5 billion searches a day.  That’s impressive but itself only captures a tiny fraction of the time savings facilitated by digital tools.
Some economists have used advertising revenues as a proxy for measuring the GNP impact of this activity, but that seems largely to miss the point: these services deliver value far beyond the cost of advertising, a surplus captured by the users. While advertising revenues are included, this additional value is entirely excluded from GNP accounts, and increases the divergence between the measured transactional economy and the day-to-day life of individuals.
In a recent report for the UK government, Charles Bean offered two ways to assess the mismeasurement of digital activity: “We can use average wages to estimate the value of the time people spend online using free digital products, or we can adjust telecommunication services output to account for the rapid growth in Internet traffic.” Bean argues that this component would add between 1/3 and 2/3 of a percentage point to UK growth annually. However, time-use surveys suggest that the usage of digital services is rapidly increasing, and is for example rapidly replacing TV for young adults (TV usage for Millennials is down 48% since 2008). It is not clear how well Bean’s methodology accommodates these rapid changes.
Unpaid effort and the rise of new media
GNP excludes activities that are provided for free – GNP is a measure of transactions, not activities. So for example, free services provided by volunteers through traditional nonprofits like the Red Cross are excluded. But so are important activities, like Wikipedia or local services such as the listserv that I maintain for 400 families. These unpaid volunteer services are just as valuable as equivalent paid services. The difference is that they are not part of the transaction economy.
New kinds of unpaid work are also emerging, for example work on open source software such as Linux, and the vast explosion of user-generated content available on YouTube, Wikipedia, and of course Facebook. These tools provide a new way to distribute “free” work. Some is entirely altruistic (or nearly so – maybe Wikipedia editors get a quick self-esteem boost).
The time has come to reopen the 1950s debate about how we should define the economy, and ensure that GDP or its replacement counts the vital work that goes on in the home, and in the community, as well as the marketplace.
– Diane Coyle
Of course, the elephant in the unpaid room is the massive range of work around the home provided (usually by women) for no pay. This includes taking care of kids, the elderly, and the sick; parenting; cooking and cleaning; and the hundreds of other “jobs” which are all excluded from GNP. The OECD found in a 2011 study that so-called “home production” would add between 20% and 50% to the GDP of its member countries. Diane Coyle argues that in reality, the way we measure GNP is inherently sexist precisely because it values paid work over the unpaid work that is necessary for paid work to be possible.
Women and unpaid work
Because GDP measures only paid work, what used to be known as “women’s work” – caring, parenting, home-making – is excluded. This is no longer work for women only in the advanced economies though they still bear more than their share of this burden, but this exclusion has been recognized as illustrating the weakness of GDP as a measure of social progress.  While Simon Kuznets – the original progenitor of the measure – warned against its use for such a measure, the failure to include non-transactional work really illustrates profound weakness.
This has a long history. In 1972, Tobin and Nordhaus estimated that unpaid work was equivalent to approximately 40% of GDP. BEA more recently estimated that it would add 26% to GDP (in 2011).
So far as change in the amount of unpaid home work is concerned, there are arguments on both sides. There have been important labor-saving devices – for example the washing machine – although sometimes those devices make us more efficient and hence able to spend the same amount of time while washing our clothes much more often. It turns out that we just wash clothes much more often. Yet there is also some evidence that parenting is more labor intensive than it used to be, and the growth of heavily involved parenting would suggest that the overall amount of home working has increased. Parents for example spend about an hour a day more on their kids than they did in 1965.
Recently, Benjamin Bridgeman estimated that household production has fallen fairly sharply, from about 37% of GDP in 1965 (not fart off the Tobin and Nordhaus estimate) to 23% in 2016, which he attributes mainly to the further entry of women into the workplace, as employed women spend much less time on household tasks. This decline suggests that there are other uncounted factors that also tend to reduce productivity growth. And if we consider aggregate economic growth (i.e. GNP), it’s clear that just shifting women from household to paid production may increase the reported numbers, but not overall growth including areas beyond the transaction economy such as household production.
How white-collar workers appropriate productivity gains
“Production” is a word with deep connotations. It brings to mind images of grimy men in old-fashioned clothes toiling in noisy factories, with steam belching from mysterious machines. A world of hard physical labor.
Of course, today’s manufacturing plants look nothing like that. They are gleaming temples filled with robots, where white coats and workers – where they are necessary – perform precise work on pristine objects. But our mental model of production is still tied to the factory. In understanding and thinking about productivity, it is still all about the production line: input raw material, add technology and labor, stir with management hocus-pocus, and out come boxes of products.
That is of course not even true of modern manufacturing. But it is far less true in the service sector that now dominates our economy. Of course, some parts of the service sector find it easy to measure outputs: McDonalds knows precisely how many customers a team has served. And more generally, where the service sector is providing actual services directly to customers, there are ways to measure labor inputs against outputs.
But that mental model breaks down in lots of places where measuring outputs is hard or impossible. And because measuring outputs is so hard, the door is open for workers to find ways to manage their workload, and in effect to appropriate the incremental fruits of productivity gains.
As Noah Smith puts it, “We get away with checking Facebook, Twitter and Snapchat on the job because we’re getting more done in less time.” Basically, if all those new digital tools and communication capabilities are in fact improving office productivity, but we see no growth I either output or employment (fewer workers completing the same amount of work), then where does that improved productivity go?
Maybe in the short term white-collar workers themselves appropriate that productivity (to use an excellent old Marxist term). They do their work more efficiently, but do they put up a hand and say, “more work over here please?” No they do not. And do managers see that more is being produced and reduce the number of workers accordingly. No they do not. Not only I it hard for managers to see that this incremental productivity is growing and that workers are easing off in consequence, there is plenty of evidence that managers are more interested in empire building than in efficiency. More workers = bigger empire.
Before the digital revolution, if efficiency increased white collar workers appropriated it in other ways – coming in work a bit later, leaving a bit earlier, spending more time chit-chatting at the water cooler, or taking those longer lunch breaks.
The modern office worker uses two miraculous devices to suck up that extra time: the computer, and the smartphone. In places where the computer is available and unmonitored, it’s the preferred medium because it’s better for movies and videos, it’s faster, and its screen can quickly be converted into a dull spreadsheet or presentation at the click of a mouse button. But where keystrokes are monitored by employers, workers now have their own personal pocket supercomputer-cum-video player – the smartphone. Videos, shopping, Facebook. Porn, music, news (fake or otherwise) … That downturn in productivity in 2005 coincides pretty much exactly with the full-scale entry of personal entertainment into the workplace.
For obvious reasons, it’s hard to find direct evidence to support this theory. Workers don’t readily admit that they are slacking off, and management equally has little incentive to report its own failures. Still, the American Time Use Survey does ask respondents how much time they spend at work not working. For those workers who do not claim that they work 100% of the time, the figure is 50 minutes per day – about 10 percent of a working day.
Other data suggests that the amount of time appropriated by workers is much higher. In a survey of 750 employees from Salary.com, the percentage reporting that they waste time at work is up from 69% in 2013 to 89% in 2014. 4% of workers reported that they wasted at least half their working time. And according to a (bigger) survey of managers and HR executives for CareerBuilder, a quarter of workers spent at least an hour a day on personal phone calls and texts, and half took personal calls or texted during working hours (see box).
How to waste time at work: Results from CareerBuilder survey.
Percent of respondents reporting that they:
1) Cell phone/texting – 50%
2) Gossip – 42%
3) The Internet – 39%
4) Social media – 38%
5) Snack breaks or smoke breaks – 27%
6) Noisy co-workers – 24%
7) Meetings – 23%
8) Email – 23%
9) Co-workers dropping by – 23%
10) Co-workers putting calls on speaker phone – 10%
Source: Harris poll for CareerBuilderThe digital revolution offers far more opportunity because everyone has the technology. Smartphones have become part of normal office culture for workers and mangers alike. And if everyone does it, the amount of time stolen through smartphones grows slowly and silently during periods of slow and steady productivity increases. A recent study found that participants used their phone for 5.05 hours daily and checked it 84 times on average.the study also found that people actually checked their phones more than twice as often as they thought they did.
But when the economy turns down and recessions bite, the workplace convulses. Lots of workers are let go. But there is no actual decline in white collar output! Somehow, fewer workers can meet the company’s needs.
So all of those productivity gains of previous years are snatched back by management during recessions. As workers buckle back down they put those phones away for a while. The lack of output decline and the sudden improvement in productivity is the clue here, showing that workers could have produced at that level prior to the recession (and related firings), but had not needed to do so.
Is more of the economy producing non-market outputs? Government, nonprofits, and other sectors with non-market outputs
Government spending includes defense, education, health care, judiciary and the justice system, infrastructure like roads and highways, libraries, and the arts – as well as the bureaucracies needed to deliver these goods and services. State, local, and Federal governments account for about 36% of US GNP. Because some of those expenditures are through contractors, BEA estimates the economy outside non-farm private businesses at 22.2% of the economy. Yet the Bureau of Economic Analysis (the source of productivity data) estimates that productivity growth in this sector totally only 15% between 1973 and 2006 – less than one fifth the productivity growth of the private sector.
Its hard to even begin to estimate productivity for these services. More is generally seen as better, but is it a productivity improvement to add pages to the form required for a building permit? Is it a productivity gain to require one more approval for that permit? Outcomes in the market economy can at least be measured in market terms – their aggregate market value. Because government services are not sold on the market, we can’t value them in the same way.
Today, economists solve the problem in relation to GDP by using cost data. But this doesn’t help at all when measuring productivity gains. Today, I can conduct almost all my business with Montgomery County online. This saves me time, and may save the county staffing costs. But unless workers are fired no productivity change is recorded. Increased outputs – in quality or quantity – are simply ignored.
The government economy is likely becoming much more productive, but those gains are excluded from productivity estimates at the national level. If productivity gains in the government sector only matched the gains in the private sector, overall productivity would have grown y% annually over the past x years.
Nonprofit output is subject to similar difficulties. We can try to measure social programs – like the number of meals on Wheels delivered, beds found for the homeless, or tutors working with underprivileged kids. But while workers may be paid to do these things (though often they are volunteer), aggregate data has no vision into how much is being provided – only into the costs of inputs. If a nonprofit somehow miraculously found a way to double the number of meals on Wheels served, that would not show up in the productivity data because there is no transaction to record. All we have are the input costs – the labor and the meals.
Finally, there is the huge are of environmental costs and benefits. We have imposed very substantial costs on producers to clean our air and water and reduce accidents at work. Robert Gordon notes that “we have understated the growth of productivity from 1970 to the turn of the 21st century when we had major improvements in air and water quality mandated by legislation. We have incorporated part of this clean-up into productivity statistics in a very subtle way by accounting for emissions control devices on auto engines. But most of the improvements in the environment are missing from GDP.” This is a big deal! In comparison to the post-war period, we spend a lot of energy and resources on environmental improvement, but as Gordon says, this is excluded from the productivity data. Estimates from the National Association of Manufacturers based on research by Robert Hahn and John Hird, and using additional data from OMB, found that the cost of environmental regulation was on the order of $250-300 billion annually – about 1.75% of GDP.
Gordon’s point underscores the reality that GNP is designed to capture the transaction economy. It does that – though with increasing difficulty. The problem is that in a modern sophisticated economy more and more outputs are not transactional.
The impact of ICT on productivity has been an important component in the overall productivity debate. There was a significant lag between the introduction of computers in the late 1980s and the subsequent upturn in productivity almost a decade later in 1995 during this period economists queries the “missing” productivity.
The second high-growth period for productivity between 1995 and 2004 is by some explained in large part by the rapid adoption of computers and the flow of business software across the entire economy during this period. Windows because ubiquitous, as did Microsoft Office, and white-collar work quickly became more productive (in part because some former jobs – like typist and secretary – became obsolete.
Since then, the regular improvements in semiconductor technology predicted by Moore’s Law have slowed, leading some economists to question the impact – and especially future impact – of declining semiconductor prices on productivity (see Technology Chapter for extended discussion of Moore’s Law). 
This point is discussed in the Technology chapter. However, it is worth noting that the productivity gains from ICT are on some accounts systematically under-estimated. Goldman Sachs argues that this mis-estimate alone accounts for about 0.9 percentage points of GNP growth – or around ¾ of the “missing” GNP. 
Syverson and others have argued that the size of the “missing” productivity is too large to be accounted for by mismeasurement. They note that the same apparent slowdown has occurred in all advanced countries, regardless of the size of their ICT sector. Similar points have been made about different patterns of ICT across US states.
But this misses the main point of ICT. Mismeasurement doesn’t come from the production side of ICT, which is essentially irrelevant. iPhones are not made in New York or North Dakota, they are used there. So it’s not the ICT production sector that matters.
The black economy
Finally, there is some evidence that the size of the black (off-the-books) economy is growing. By definition, activities that are hidden from the state are not counted by the state, so are not included in official data.
For the black economy to help explain the productivity paradox, though, it heeds to be growing faster than the economy as a whole. Is it?
A recent federal reserve bank survey found that in 2013 about 40% of respondents had some type of informal paid activity; by 2015, a subsequent survey found that these respondents were now majority, and now accounted for 52% of men and 60% of women. These are very large numbers. Moreover, these workers often don’t report income: an IRS study in the 1970s found that 47% of independent contractors reported none of their contracting income; Cebula and Feige’s estimate that overall, 18-23% of income is not reported in the US.
Tracking the black and informal economy is by definition difficult because participants don’t want to be tracked. Still, there is ample evidence that this sector is large and growing quite rapidly.
For a more detailed analysis, see the Gig Economy chapter.
Historically, national accounts treated expenditure on intangible inputs as an intermediate expense and thus excluded these investments from GDP. That began to change with the capitalization of software, which added significantly to GDP. But software is only a small part of intangibles – perhaps 15% of all intangibles. Carol Corrado shows that total investments in intangibles grew steadily in the last decades of the 20th century, and amounted to more than 12% of tangible investment by 2000. This share must have grown – perhaps substantially – since the turn of the Millennium.
Corrado estimated that total investment in intangibles was about $1 trillion in 2003 – about the same as tangible investment – which suggests that about $850 billion was not captured by the national accounts (in 2003).  More recent analysis finds that intangible investments of about 4.6% of GDP are missing from US national accounts, and that this missing investment has been growing.
The measurement of GDP is immensely complicated – this paper has only touched on some of the more important arguments, and I claim no expertise at all in this subject. The chart on the following page from Diane Coyle lists a considerable number of issues that have not even been mentioned in this paper, as well as some that have been touched upon. Like many constructs in economics. If one takes a microscope to the apparently firm and well-built data structure encapsulated in GDP, it dissolves into a pile of assumptions that crumble at the first contact with critical scrutiny.
I want to conclude by reviewing the nature of the conflict between true believers and heretics. True believers base their arguments and theories firmly in the official data. They may acknowledge some little local difficulties, but in general they reject the heretics’ claims as a solution in search of a problem. The data say that innovation and growth are down, and the problem is to explain that not to challenge it.
The heretics can be more eclectic. They just have to claim that the data are wrong, not that all the arguments presented by the realists are wrong. It may be both that there is far more innovation all around us, and that it is having a profound effect on the economy, and that Baumol’s disease exists, that there are growing gaps between top and trailing firms in a given sector, and that that there are longer-than-expected lags in the introduction of new technologies.
For myself, I would argue that the data are wrong for the many reasons listed above, that we are experiencing a wave of innovation, and that some of the characteristics explored by the realists also throw important light on probably future paths. The growing gap within sectors, the impact of growing concentration, and implementation perhaps driven by insufficient investment all seem plausible. But the grim world of secular stagnation does not.
 David Byrne, Stephen Oliner, and Daniel Sichel, “Is the Information Technology Revolution Over?,” SSRN Scholarly Paper (Rochester, NY: Social Science Research Network, March 27, 2013), https://papers.ssrn.com/abstract=2240961.
 Dan Andrews, Chiara Criscoulo, and Peter N. Gal, “The Global Productivity Slowdown, Technology Divergence, and Public Policy: A Firm Level Perspective” (Brookings Institution, September 16, 2016), https://www.brookings.edu/wp-content/uploads/2016/08/andrews-et-al.pdf.
 Robert D. Atkinson, “To Solve Our Growth Problem, We Must First Solve Our Productivity Problem,” Christian Science Monitor, July 21, 2016, http://www.csmonitor.com/Technology/Breakthroughs-Voices/2016/0721/To-solve-our-growth-problem-we-must-first-solve-our-productivity-problem.
 Sally Andrews et al., “Beyond Self-Report: Tools to Compare Estimated and Real-World Smartphone Use,” PLOS ONE 10, no. 10 (October 28, 2015): e0139004, https://doi.org/10.1371/journal.pone.0139004.
 “CareerBuilder Study Reveals Top Ten Productivity Killers at Work – CareerBuilder,” accessed January 21, 2018, http://www.careerbuilder.com/share/aboutus/pressreleasesdetail.aspx?sd=6/12/2014&id=pr827&ed=12/31/2014.
 Brice S. McKeever and Sarah L. Pettijohn, “The Nonprofit Sector in Brief 2014,” Washington, DC: Urban Institute, 2014.
 W. Mark Crain and Nicole V. Crain, “The Cost of Federal Regulation to the US Economy, Manufacturing and Small Business” (National Association of Manufacturers, 2014).
 See the Gig Economy chapter for a more extensive discussion
 Kuznets, Simon, National Income, 1929-1932 : Letter from the Acting Secretary of Commerce Transmitting in Response to Senate Resolution No. 220 (72nd Cong.) a Report on National Income, 1929-32, Congressional Documents (U.S. Government Printing Office, 1934), https://fraser.stlouisfed.org/scribd/?title_id=971&filepath=/files/docs/publications/natincome_1934/19340104_nationalinc.pdf.
 William J Baumol and William G Bowen, Performing Arts – the Economic Dilemma: A Study of Problems Common to Theater, Opera, Music and Dance (London: Gregg Revivals, 1993).
 “Tuition and Fees and Room and Board over Time, 1975-76 to 2015-16, Selected Years,” Trends in Higher Education (College Board, ND), https://trends.collegeboard.org/college-pricing/figures-tables/tuition-and-fees-and-room-and-board-over-time-1975-76-2015-16-selected-years.
 Daniel Castro, Robert D. Atkinson, and Stephen J. Ezell, “Embracing the Self-Service Economy” (ITIF, April 2010), http://www.itif.org/files/2010-self-service-economy.pdf.
 Jack E. Triplett and Barry P. Bosworth, “Productivity Measurement Issues in Services Industries: ‘Baumol’s Disease’ Has Been Cured,” Economic Policy Review, no. Sep (2003): 23–33.
 Bureau of Labor Statistics, Occupational Employment Survey
 Robert Gordon, The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War, The Princeton Economic History of the Western World (Princeton NJ: Princeton University Press, 2016).
 Robert J. Gordon, “Is U.S. Economic Growth Over? Faltering Innovation Confronts the Six Headwinds,” Working Paper (National Bureau of Economic Research, August 2012), https://doi.org/10.3386/w18315.
 Tyler Cowen, The Great Stagnation: How America Ate All the Low-Hanging Fruit of Modern History, Got Sick, and Will (Eventually) Feel Better (New York: Dutton, 2011)
 Tyler Cowen, “Silicon Valley Has Not Saved Us from a Productivity Slowdown,” The New York Times, March 4, 2016, http://www.nytimes.com/2016/03/06/upshot/silicon-valley-has-not-saved-us-from-a-productivity-slowdown.html.
 John Fernald and Bing Wang, “The Recent Rise and Fall of Rapid Productivity Growth,” Federal Reserve Bank of San Francisco Economic Letters 2015, no. 4 (February 9, 2015), http://www.frbsf.org/economic-research/publications/economic-letter/2015/february/economic-growth-information-technology-factor-productivity/.
 Paul A. David, “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox,” The American Economic Review 80, no. 2 (1990): 355–61.
 Erik Brynjolfsson, Daniel Rock, and Chad Syverson, “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics,” Working Paper (National Bureau of Economic Research, November 2017), https://doi.org/10.3386/w24001.
 Robert D. Atkinson, “Think like an Enterprise: Why Nations Need Comprehensive Productivity Strategies” (ITIF, May 2016), http://www2.itif.org/2016-think-like-an-enterprise.pdf?_ga=2.22246587.1699919044.1514421354-308457684.1514421354.
 W.Brian Arthur, “The Second Economy,” McKinsey Quarterly, October 2011, http://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/the-second-economy.
 Robert D. Atkinson, Daniel Castro, and Stephen J. Ezell, “The Digital Road to Recovery: A Stimulus Plan to Create Jobs, Boost Productivity and Revitalize America” (ITIF, January 2009), http://www.itif.org/files/roadtorecovery.pdf.
 Germán Gutiérrez and Thomas Philippon, “Investment-Less Growth: An Empirical Investigation,” Working Paper (National Bureau of Economic Research, December 2016), http://www.nber.org/papers/w22897.
 “Fact Sheet: Electronic Ticketing (ET),” IATA, September 2009, http://www.iata.org/pressroom/facts_figures/fact_sheets/et.htm. Quoted in Castro, op.cit.
 Dan Andrews et al., “Frontier Firms, Technology Diffusion and Public Policy: Micro Evidence from OECD Countries” (OECD Publishing, 2015), http://www.oecd.org/eco/growth/Frontier-Firms-Technology-Diffusion-and-Public-Policy-Micro-Evidence-from-OECD-Countries.pdf.
 Andrews, Criscoulo, and Gal, “The Global Productivity Slowdown, Technology Divergence, and Public Policy: A Firm Level Perspective.” Presentation at the Brookings Institution, 2016
 Ryan Decker et al., “The Secular Decline in Business Dynamism in the US,” Manuscript, University of Maryland, 2013, http://econweb.umd.edu/~haltiwan/DHJM_6_2_2014.pdf.
 Catherine L. Mann and Dan Andrews, “How Zombie Companies Stop Productivity Growth,” Bloomberg.Com, May 5, 2017, https://www.bloomberg.com/view/articles/2017-05-05/how-zombie-companies-stop-productivity-growth.
 “America’s Uncompetitive Markets Harm Its Economy,” The Economist, July 27, 2017, https://www.economist.com/news/finance-and-economics/21725552-new-research-suggests-too-little-competition-deters-investment-americas.
 Andrews, Criscoulo, and Gal, “The Global Productivity Slowdown, Technology Divergence, and Public Policy: A Firm Level Perspective.”
 Jonathan Haskel, Peter Goodridge, and Gavin Wallis, “Accounting for the UK Productivity Puzzle: A Decomposition and Predictions,” Working Paper (Imperial College, London, Imperial College Business School, February 12, 2015), https://econpapers.repec.org/paper/impwpaper/21167.htm.
 Jaisen R. Abel, Richard Dietz, and Yagin Su, “Are Recent College Graduates Finding Good Jobs?,” Current Issues in Economics and Finance (Federal Reserve Bank of New York, 2014), https://www.newyorkfed.org/research/current_issues/ci20-1.html.
 Bryan Caplan, “The World Might Be Better Off Without College for Everyone,” The Atlantic, February 2018, https://www.theatlantic.com/magazine/archive/2018/01/whats-college-good-for/546590/.
 “Halfon: Half of Higher Education Students Should Be Apprentices,” Times Higher Education (THE), November 8, 2017, https://www.timeshighereducation.com/news/halfon-half-higher-education-students-should-be-apprentices.
 Nordhaus, William D. “Do Real Output and Real Wage Measures Capture Reality? The
History of Light Suggests Not.” In The Economics of New Goods, ed. Timothy F. Bresnahan and Robert J. Gordon, 29-66. Studies in Income and Wealth, vol. 58. 1997 Chicago: The University of Chicago Press.
 Erik Hurst, “Video Killed the Radio Star,” Chicago Booth Review, September 1, 2016, http://review.chicagobooth.edu/economics/2016/article/video-killed-radio-star?adbsc=social_20160905_65598496&adbid=772816206399025153&adbpl=tw&adbpr=1702155980.
 Hal Varian, “Intelligent Technology,” Finance and Development 53, no. 3 (September 2016), http://ritholtz.com/2016/09/intelligent-technology/.
 Larry Downes, “The 15 Most Important Slides in Mary Meeker’s Internet Trends Report,” Washington Post, June 2, 2016, https://www.washingtonpost.com/news/innovations/wp/2016/06/02/the-15-most-important-slides-in-mary-meekers-internet-trends-report/.
 Mo Abdirahman et al., “A Comparison of Approaches to Deflating Telecoms Services Output,” IET/ONS Working Paper, 2017.
 Allison Nathan, “Interview with Jan Mokyr,” Top of Mind (blog), October 5, 2015, http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
 Timothy Taylor, “How Well Does GDP Measure the Digital Economy?,” CONVERSABLE ECONOMIST (blog), July 19, 2016, http://conversableeconomist.blogspot.com/2016/07/how-well-does-gdp-measure-digital.html.
 Heather Lyu et al., “Overtreatment in the United States,” PLOS ONE 12, no. 9 (September 6, 2017): e0181970, https://doi.org/10.1371/journal.pone.0181970.
 David M. Byrne, John G. Fernald, and Marshall B. Reinsdorf, “Does the United States Have a Productivity Slowdown or a Measurement Problem?,” Brookings Papers on Economic Activity, March 4, 2016, http://www.brookings.edu/about/projects/bpea/papers/2016/byrne-et-al-productivity-measurement.
 Nadim Ahmad and Paul Schreyer, “Are GDP and Productivity Measures Up to the Challenges of the Digital Economy?” International Productivity Monitor, Spring 2016.
 Yan Chen, Grace YoungJoo Jeon, and Yong-Mi Kim, “A Day without a Search Engine: An Experimental Study of Online and Offline Searches,” Experimental Economics 17, no. 4 (December 1, 2014): 512–36, https://doi.org/10.1007/s10683-013-9381-9.
 Charles Bean, “Measuring the Value of Free,” Project Syndicate, May 3, 2016, https://www.project-syndicate.org/commentary/measuring-gdp-in-digital-economy-by-charles-bean-2016-05. Bean is former Deputy Director of the Bank of England.
 Veerle Miranda, “Cooking, Caring and Volunteering: Unpaid Work Around the World,” OECD Social, Employment and Migration Working Papers, September 20, 2011, https://doi.org/10.1787/5kghrjm8s142-en.
 Diane Coyle, “The Way We Measure Economies Is Inherently Sexist,” World Economic Forum (blog), April 13, 2016, https://www.weforum.org/agenda/2016/04/why-economic-policy-overlooks-women/.
 Dotti Sani, Giulia M, and Judith Treas, “Educational Gradients in Parents’ Child‐Care Time Across Countries, 1965–2012,” Journal of Marriage and Family 78, no. 4 (August 1, 2016): 1083–96, https://doi.org/10.1111/jomf.12305.
 Benjamin Bridgeman et al., “Accounting for Household Production in the National Accounts, 1965–2010,” Survey of Current Business 95, no. 2 (2012), https://www.bea.gov/scb/pdf/2016/2%20February/0216_accounting_for_household_production_in_the_national_accounts.pdf.
 Noah Smith, “Goofing Off at Work Masks Rising Productivity,” Bloomberg.Com, May 10, 2017, https://www.bloomberg.com/view/articles/2017-05-10/goofing-off-at-work-masks-rising-productivity.
 “CareerBuilder Study Reveals Top Ten Productivity Killers at Work – CareerBuilder.”
 Andrews et al., “Beyond Self-Report.”
 Dean Baker, “The Productivity to Paycheck Gap: What the Data Show,” Washington, DC: Center for Economic and Policy Research, 2007.
 Allison Nathan, “Interview with Robert Gordon,” Top of Mind (blog), October 5, 2015, http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
 Tim Cross, “After Moore’s Law,” The Economist Technology Quarterly, March 2016, http://www.economist.com/technology-quarterly/2016-03-12/after-moores-law.
 Jan Hatzius, “Picking Apart the Productivity Paradox” (Goldman-Sachs, October 5, 2015), http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
 Chad Syverson, “Challenges to Mismeasurement Explanations for the U.S. Productivity Slowdown,” Working Paper (National Bureau of Economic Research, February 2016), http://www.nber.org/papers/w21974.
 Roberto Cardarelli and Lucine Lusinyan, “U.S. Total Factor Productivity Slowdown : Evidence from the U.S. States,” IMF Working Papers, May 28, 2015, https://www.imf.org/en/Publications/WP/Issues/2016/12/31/U-S-Total-Factor-Productivity-Slowdown-Evidence-from-the-U-S-42967.
 Anat Bracha, Mary A. Burke, and Arman Khachiyan, “Changing Patterns in Informal Work Participation in the United States 2013–2015 – Boston Fed,” Federal Reserve Bank of Boston: Current Policy Perspectives 15 (October 2015), https://www.bostonfed.org/economic/current-policy-perspectives/2015/cpp1510.htm.
 Edgar L. Feige and Richard Cebula, “America’s Unreported Economy: Measuring the Size, Growth and Determinants of Income Tax Evasion in the U.S,” MPRA Paper (University Library of Munich, Germany, 2011), https://ideas.repec.org/p/pra/mprapa/34781.html.
 Carol Corrado, “Measuring Capital and Technology: An Expanded Framework,” in Measuring Capital in the New Economy, ed. Carol Corrado, John Haltiwanger, and Dan Sichel (University of Chicago Press, 2005), 11.
 Carol Corrado et al., “Intangible Investment in the EU and US before and since the Great Recession and Its Contribution to Productivity Growth,” EIB Working Papers (European Investment Bank (EIB), 2016), https://ideas.repec.org/p/zbw/eibwps/201608.html.
 Diane Coyle, “Do-It-Yourself Digital: The Production Boundary and the Productivity Puzzle,” 2017.
Abdirahman, Mo, Diane Coyle, Richard Heys, and Will Stewart. “A Comparison of Approaches to Deflating Telecoms Services Output.” IET/ONS Working Paper, 2017.
Abel, Jaisen R., Richard Dietz, and Yagin Su. “Are Recent College Graduates Finding Good Jobs?” Current Issues in Economics and Finance. Federal Reserve Bank of New York, 2014. https://www.newyorkfed.org/research/current_issues/ci20-1.html.
Ahmad, Nadim, and Paul Schreyer. “Are GDP and Productivity Measures Up to the Challenges of the Digital Economy?” International Productivity Monitor, Spring 2016.
“America’s Uncompetitive Markets Harm Its Economy.” The Economist, July 27, 2017. https://www.economist.com/news/finance-and-economics/21725552-new-research-suggests-too-little-competition-deters-investment-americas.
Andrews, Dan, Chiara Criscoulo, and Peter N. Gal. “The Global Productivity Slowdown, Technology Divergence, and Public Policy: A Firm Level Perspective.” Brookings Institution, September 16, 2016. https://www.brookings.edu/wp-content/uploads/2016/08/andrews-et-al.pdf.
Andrews, Dan, Chiara Criscuolo, Peter Gal, and others. “Frontier Firms, Technology Diffusion and Public Policy: Micro Evidence from OECD Countries.” OECD Publishing, 2015. http://www.oecd.org/eco/growth/Frontier-Firms-Technology-Diffusion-and-Public-Policy-Micro-Evidence-from-OECD-Countries.pdf.
Andrews, Sally, David A. Ellis, Heather Shaw, and Lukasz Piwek. “Beyond Self-Report: Tools to Compare Estimated and Real-World Smartphone Use.” PLOS ONE 10, no. 10 (October 28, 2015): e0139004. https://doi.org/10.1371/journal.pone.0139004.
Arthur, W.Brian. “The Second Economy.” McKinsey Quarterly, October 2011. http://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/the-second-economy.
Atkinson, Robert D. “Think like an Enterprise: Why Nations Need Comprehensive Productivity Strategies.” ITIF, May 2016. http://www2.itif.org/2016-think-like-an-enterprise.pdf?_ga=2.22246587.1699919044.1514421354-308457684.1514421354.
———. “To Solve Our Growth Problem, We Must First Solve Our Productivity Problem.” Christian Science Monitor, July 21, 2016. http://www.csmonitor.com/Technology/Breakthroughs-Voices/2016/0721/To-solve-our-growth-problem-we-must-first-solve-our-productivity-problem.
Atkinson, Robert D., Daniel Castro, and Stephen J. Ezell. “The Digital Road to Recovery: A Stimulus Plan to Create Jobs, Boost Productivity and Revitalize America.” ITIF, January 2009. http://www.itif.org/files/roadtorecovery.pdf.
Baker, Dean. “The Productivity to Paycheck Gap: What the Data Show.” Washington, DC: Center for Economic and Policy Research, 2007.
Baumol, William J, and William G Bowen. Performing Arts – the Economic Dilemma: A Study of Problems Common to Theater, Opera, Music and Dance. London: Gregg Revivals, 1993.
Bean, Charles. “Measuring the Value of Free.” Project Syndicate, May 3, 2016. https://www.project-syndicate.org/commentary/measuring-gdp-in-digital-economy-by-charles-bean-2016-05.
Bracha, Anat, Mary A. Burke, and Arman Khachiyan. “Changing Patterns in Informal Work Participation in the United States 2013–2015 – Boston Fed.” Federal Reserve Bank of Boston: Current Policy Perspectives 15 (October 2015). https://www.bostonfed.org/economic/current-policy-perspectives/2015/cpp1510.htm.
Bridgeman, Benjamin, A. Dugan, M. Lal, S. Villones, and M. Osborne. “Accounting for Household Productionin the National Accounts, 1965–2010.” Survey of Current Business 95, no. 2 (2012). https://www.bea.gov/scb/pdf/2016/2%20February/0216_accounting_for_household_production_in_the_national_accounts.pdf.
Brynjolfsson, Erik, Daniel Rock, and Chad Syverson. “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics.” Working Paper. National Bureau of Economic Research, November 2017. https://doi.org/10.3386/w24001.
Byrne, David M., John G. Fernald, and Marshall B. Reinsdorf. “Does the United States Have a Productivity Slowdown or a Measurement Problem?” Brookings Papers on Economic Activity, March 4, 2016. http://www.brookings.edu/about/projects/bpea/papers/2016/byrne-et-al-productivity-measurement.
Byrne, David, Stephen Oliner, and Daniel Sichel. “Is the Information Technology Revolution Over?” SSRN Scholarly Paper. Rochester, NY: Social Science Research Network, March 27, 2013. https://papers.ssrn.com/abstract=2240961.
Caplan, Bryan. “The World Might Be Better Off Without College for Everyone.” The Atlantic, February 2018. https://www.theatlantic.com/magazine/archive/2018/01/whats-college-good-for/546590/.
Cardarelli, Roberto, and Lucine Lusinyan. “U.S. Total Factor Productivity Slowdown : Evidence from the U.S. States.” IMF Working Papers, May 28, 2015. https://www.imf.org/en/Publications/WP/Issues/2016/12/31/U-S-Total-Factor-Productivity-Slowdown-Evidence-from-the-U-S-42967.
“CareerBuilder Study Reveals Top Ten Productivity Killers at Work – CareerBuilder.” Accessed January 21, 2018. http://www.careerbuilder.com/share/aboutus/pressreleasesdetail.aspx?sd=6/12/2014&id=pr827&ed=12/31/2014.
Castro, Daniel, Robert D. Atkinson, and Stephen J. Ezell. “Embracing the Self-Service Economy.” ITIF, April 2010. http://www.itif.org/files/2010-self-service-economy.pdf.
Chen, Yan, Grace YoungJoo Jeon, and Yong-Mi Kim. “A Day without a Search Engine: An Experimental Study of Online and Offline Searches.” Experimental Economics 17, no. 4 (December 1, 2014): 512–36. https://doi.org/10.1007/s10683-013-9381-9.
Corrado, Carol. “Measuring Capital and Technology: An Expanded Framework.” In Measuring Capital in the New Economy, edited by Carol Corrado, John Haltiwanger, and Dan Sichel, 11. University of Chicago Press, 2005.
Corrado, Carol, Jonathan Haskel, Cecilia Jona-Lasinio, and Massimiliano Iommi. “Intangible Investment in the EU and US before and since the Great Recession and Its Contribution to Productivity Growth.” EIB Working Papers. European Investment Bank (EIB), 2016. https://ideas.repec.org/p/zbw/eibwps/201608.html.
Cowen, Tyler. “Silicon Valley Has Not Saved Us From a Productivity Slowdown.” The New York Times, March 4, 2016. http://www.nytimes.com/2016/03/06/upshot/silicon-valley-has-not-saved-us-from-a-productivity-slowdown.html.
———. The Great Stagnation: How America Ate All the Low-Hanging Fruit of Modern History, Got Sick, and Will( Eventually) Feel Better. New York: Dutton, 2011. https://www.amazon.com/Great-Stagnation-America-Low-Hanging-Eventually/dp/0525952713/ref=tmm_hrd_swatch_0?_encoding=UTF8&qid=&sr=.
Coyle, Diane. “Do-It-Yourself Digital: The Production Boundary and the Productivity Puzzle,” 2017.
———. “The Way We Measure Economies Is Inherently Sexist.” World Economic Forum (blog), April 13, 2016. https://www.weforum.org/agenda/2016/04/why-economic-policy-overlooks-women/.
Crain, W. Mark, and Nicole V. Crain. “The Cost of Federal Regulation to the US Economy, Manufacturing and Small Business.” National Association of Manufacturers, 2014.
Cross, Tim. “After Moore’s Law.” The Economist Technology Quarterly, March 2016. http://www.economist.com/technology-quarterly/2016-03-12/after-moores-law.
David, Paul A. “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox.” The American Economic Review 80, no. 2 (1990): 355–61.
Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. “The Secular Decline in Business Dynamism in the US.” Manuscript, University of Maryland, 2013. http://econweb.umd.edu/~haltiwan/DHJM_6_2_2014.pdf.
Downes, Larry. “The 15 Most Important Slides in Mary Meeker’s Internet Trends Report.” Washington Post, June 2, 2016. https://www.washingtonpost.com/news/innovations/wp/2016/06/02/the-15-most-important-slides-in-mary-meekers-internet-trends-report/.
Feige, Edgar L., and Richard Cebula. “America’s Unreported Economy: Measuring the Size, Growth and Determinants of Income Tax Evasion in the U.S.” MPRA Paper. University Library of Munich, Germany, 2011. https://ideas.repec.org/p/pra/mprapa/34781.html.
Fernald, John, and Bing Wang. “The Recent Rise and Fall of Rapid Productivity Growth.” Federal Reserve Bank of San Francisco Economic Letters 2015, no. 4 (February 9, 2015). http://www.frbsf.org/economic-research/publications/economic-letter/2015/february/economic-growth-information-technology-factor-productivity/.
Gordon, Robert. The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War. The Princeton Economic History of the Western World. Princeton NJ: Princeton University Press, 2016. https://www.amazon.com/dp/B0131KW67U/ref=dp-kindle-redirect?_encoding=UTF8&btkr=1.
Gordon, Robert J. “Is U.S. Economic Growth Over? Faltering Innovation Confronts the Six Headwinds.” Working Paper. National Bureau of Economic Research, August 2012. https://doi.org/10.3386/w18315.
Gutiérrez, Germán, and Thomas Philippon. “Investment-Less Growth: An Empirical Investigation.” Working Paper. National Bureau of Economic Research, December 2016. http://www.nber.org/papers/w22897.
“Halfon: Half of Higher Education Students Should Be Apprentices.” Times Higher Education (THE), November 8, 2017. https://www.timeshighereducation.com/news/halfon-half-higher-education-students-should-be-apprentices.
Haskel, Jonathan, Peter Goodridge, and Gavin Wallis. “Accounting for the UK Productivity Puzzle: A Decomposition and Predictions.” Working Paper. Imperial College, London, Imperial College Business School, February 12, 2015. https://econpapers.repec.org/paper/impwpaper/21167.htm.
Hatzius, Jan. “Picking Apart the Productivity Paradox.” Goldman-Sachs, October 5, 2015. http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
Hurst, Erik. “Video Killed the Radio Star.” Chicago Booth Review, September 1, 2016. http://review.chicagobooth.edu/economics/2016/article/video-killed-radio-star?adbsc=social_20160905_65598496&adbid=772816206399025153&adbpl=tw&adbpr=1702155980.
Kuznets, Simon. National Income, 1929-1932 : Letter from the Acting Secretary of Commerce Transmitting in Response to Senate Resolution No. 220 (72nd Cong.) a Report on National Income, 1929-32,. Congressional Documents. U.S. Government Printing Office, 1934. https://fraser.stlouisfed.org/scribd/?title_id=971&filepath=/files/docs/publications/natincome_1934/19340104_nationalinc.pdf.
Lyu, Heather, Tim Xu, Daniel Brotman, Brandan Mayer-Blackwell, Michol Cooper, Michael Daniel, Elizabeth C. Wick, Vikas Saini, Shannon Brownlee, and Martin A. Makary. “Overtreatment in the United States.” PLOS ONE 12, no. 9 (September 6, 2017): e0181970. https://doi.org/10.1371/journal.pone.0181970.
Mann, Catherine L., and Dan Andrews. “How Zombie Companies Stop Productivity Growth.” Bloomberg.Com, May 5, 2017. https://www.bloomberg.com/view/articles/2017-05-05/how-zombie-companies-stop-productivity-growth.
McKeever, Brice S., and Sarah L. Pettijohn. “The Nonprofit Sector in Brief 2014.” Washington, DC: Urban Institute, 2014.
Miranda, Veerle. “Cooking, Caring and Volunteering: Unpaid Work Around the World.” OECD Social, Employment and Migration Working Papers, September 20, 2011. https://doi.org/10.1787/5kghrjm8s142-en.
Nathan, Allison. “Interview with Jan Mokyr.” Top of Mind (blog), October 5, 2015. http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
———. “Interview with Robert Gordon.” Top of Mind (blog), October 5, 2015. http://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf.
Sani, Dotti, Giulia M, and Judith Treas. “Educational Gradients in Parents’ Child‐Care Time Across Countries, 1965–2012.” Journal of Marriage and Family 78, no. 4 (August 1, 2016): 1083–96. https://doi.org/10.1111/jomf.12305.
Smith, Noah. “Goofing Off at Work Masks Rising Productivity.” Bloomberg.Com, May 10, 2017. https://www.bloomberg.com/view/articles/2017-05-10/goofing-off-at-work-masks-rising-productivity.
Syverson, Chad. “Challenges to Mismeasurement Explanations for the U.S. Productivity Slowdown.” Working Paper. National Bureau of Economic Research, February 2016. http://www.nber.org/papers/w21974.
Taylor, Timothy. “How Well Does GDP Measure the Digital Economy?” CONVERSABLE ECONOMIST (blog), July 19, 2016. http://conversableeconomist.blogspot.com/2016/07/how-well-does-gdp-measure-digital.html.
Triplett, Jack E., and Barry P. Bosworth. “Productivity Measurement Issues in Services Industries: ‘Baumol’s Disease’ Has Been Cured.” Economic Policy Review, no. Sep (2003): 23–33.
“Tuition and Fees and Room and Board over Time, 1975-76 to 2015-16, Selected Years.” Trends in Higher Education. College Board, ND. https://trends.collegeboard.org/college-pricing/figures-tables/tuition-and-fees-and-room-and-board-over-time-1975-76-2015-16-selected-years.
Varian, Hal. “Intelligent Technology.” Finance and Development 53, no. 3 (September 2016). http://ritholtz.com/2016/09/intelligent-technology/.
Table of Contents
- The Productivity Paradox
- Major sections
- Summary of the argument
- True believers: the productivity data are real
- Heretics: the productivity numbers don’t add up
- Accounting for changes in quality
- Measuring service outputs
- The free and Nearly-Free economy
- How white-collar workers appropriate productivity gains
- Is more of the economy producing non-market outputs? Government, nonprofits, and other sectors with non-market outputs
- ICT effects
- The black economy
- Additional points
- Table of Contents