The Productivity Puzzle: Mis-measurement, Diffusion, and the Future of Work in the Digital Age

This podcast episode delves into the persistent "productivity puzzle" – the long-term decline in productivity growth across the developed world, despite rapid technological advancements. The first speaker, drawing on his return to academic economics, highlights several potential explanations, starting with significant mis-measurement issues in GDP. These include the uncounted value of digitalized and virtualized work (e.g., self-service tasks), the overstatement of inflation due to the hedonic correction not fully capturing quality improvements and price declines in new goods like telecommunications, and the shift from purchasing capital goods (counted as investment) to intermediate services (netted out of final output). However, these mis-measurement arguments are challenged by scholars like Chad Syverson, who points to the lack of correlation between the slowdown and ITC investment, the GDI-GDP gap (where national income exceeds output, suggesting unmeasured work like open-source software), and the insufficient magnitude of estimated missing GDP.

The discussion then shifts to the nature of technological diffusion, emphasizing that transformative technologies take decades, not years, to fully impact productivity, drawing parallels to the slow adoption of electricity in manufacturing. The concept of "best versus the rest" is introduced, showing that only a small percentage of top firms are experiencing significant productivity gains, while the majority lag. This disparity is linked to the commoditization of digital infrastructure (cloud services offering AI, machine learning) which is now making advanced capabilities accessible to a wider range of firms, potentially leading to a broader diffusion of productivity gains, much like the electricity grid democratized access to power.

The second speaker introduces a more pessimistic view, suggesting that the productivity paradox can be explained by two additional paradoxes. The first is the "Baumol effect," where rapid productivity gains in automatable sectors lead to a proliferation of low-productivity, low-wage, often "high-touch" jobs in non-automatable services (e.g., delivery drivers), dragging down overall average productivity. The second paradox concerns the rise of "zero-sum activities" – jobs that consume economic resources and generate measured GDP but do not necessarily increase human welfare (e.g., cyber defense, financial compliance, divorce lawyers). These activities, driven by competition for relative income and status, may increasingly dominate employment and output, masking underlying rapid technological progress in essential goods and services.

Ultimately, the speakers explore the implications for human welfare and the future of economics. While acknowledging that GDP measurements likely underestimate the real income and quality improvements from new drugs, digital entertainment, and mobile devices (as argued by Marty Feldstein), they question whether these benefits adequately address the fundamental economic stresses faced by low-income individuals. Looking to the end of the century, a future is envisioned where robots perform most work, and the economy is dominated by rents from real property, intellectual property, and subjective brand values, with high incomes concentrated among a very small, skilled elite. This raises profound questions about scarcity, distribution, and the very definition of economic activity in a post-work world.

"the long-term decline in productivity across the developed world"

"we've been digitalising the world we've been virtualizing the world Erik Brynjolfsson and his sidekick dr. o and MIT have got a big program going on trying to find ways to measure missing output because we do the work that bank clerks and travel agents used to do but our work isn't counted in GDP"

"In 1900 contemporary observers well might have remarked that the electric dynamos were just be seen everywhere but in the productivity statistics"

"since 2000 the top 5% have seen their productivity increase by almost 40% in manufacturing and in services it's more than 40 percent and the rest the 95 percent this kind of reminds you of the income distribution data on the growth of income the top 90 percent are barely above where they were more than 15 years ago"

"value is moving from software to data with open source software with the cloud the software is being commoditized and the value is in the data capturing the data that is generated through transactions through interactions across the net and then learning what to do with the data"

"more data dominates better algorithms because by training your algorithms on more data you get better algorithms if you're a better algorithm you have a higher quality more valued service you get more data there's a positive feedback loop here"

"the more rapidly technological progress enables the automation of existing activities the more that high-touch jobs may grow in activities which at least for now cannot be automated or where somebody's got enrollment at excellent all wages are low enough to make automation an economic"

"the more rapidly that the rapid technological progress could eventually automate away all the activities which are truly essential for human welfare while supporting increased intensity of zero competition for relative income and status so that zero-sum activities end up accounting for a relentlessly increasing percentage of employment and measured output over time"