The Conundrum of Modern Innovation: Explaining the Productivity Slowdown
Robert Solow, 1986
In 1987, the Nobel Memorial Prize in Economic Sciences was conferred upon Robert Solow, whose academic contributions reshaped the field’s contemporary understanding of long-run economic growth. Solow, then a professor at the Massachusetts Institute of Technology, had engineered in 1956 a macroeconomic model that shifted the paradigm for the understanding of economic growth: over extended periods, the process of capital accumulation, by which an economy builds up its stock of machinery, factories and tools, should not, nor can it, be the foundation of sustainable, maintainable improvements in income per capita and quality of life. Rather, solely innovation and technological progress, through their transformative effects on productivity (whether that of workers, capital, or any other input), can propel aggregate growth to new heights. This foundational conception of growth, which makes paramount the importance of technological headway, remains the bedrock principle and intellectual compass guiding contemporary discussions about growth and productivity.
Yet, in light of this general cognizance of the primacy of technology, over the span of the past two decades, the world, and in particular its most developed economies, has seen the emergence of an engrossing conundrum: in spite of the pronounced (and perhaps, for some, startling ) proliferation of digital tools such as the Internet, advanced software and new forms of computational power, the Earth’s leading economic powers have witnessed a persistent slowdown in measured productivity growth. This paradoxical situation, which has been titled the “Solow Paradox”, constituted a formidable challenge in modern economic research upon its introduction.
“You can see the computer age everywhere but in the productivity statistics.”
To analyse this question, it is imperative to clarify the intended meaning of “productivity”. Two measures dominate the discussion. The first is labour productivity, which relates output to hours worked; the second is total factor productivity (hereinafter “TFP”). TFP reflects the portion of output not explained by labour or capital inputs. It is sometimes known as the Solow residual. In effect, it captures efficiency, knowledge, organizational improvement and technological progress. Yet, TFP is notoriously difficult to measure, since it depends on the accurate measurement of output, capital, and quality adjustments. In the same vein, the rising importance of intangible assets and digital goods complicates the measurement of both inputs and outputs. Firms invest in software, data infrastructures, organizational capital, and platforms that generate value through channels that traditional accounting struggles to track. At the same time, consumers derive extraordinary benefits from digital goods that have no monetary price.
In light of these complexities, two principal explanations have emerged to account for the slowdown in measured productivity: one asserts that the slowdown is real and structural, caused by a decline in the diffusion of new technologies across firms, and the other contends that the slowdown is mainly statistical, the artifact of measurement methods that have not kept pace with a digital and intangible economy.
The first explanation begins with a striking empirical regularity. Indeed, productivity growth at the globe’s frontier firms (the top five to ten percent of firms of each of their respective industries measured in terms of productivity, or TFP) has remained robust. Although the most advanced and innovative firms continue to improve at a rapid pace, the vast majority of firms appear unable to replicate these advances. Further to this, evidence collected across advanced economies suggests that frontier firms maintain a strong productivity performance, while non-frontier firms exhibit considerable stagnation. As such, the gap between the two groups has only widened over time. This divergence thus suggests that the source of the problem lies in diffusion, as opposed to innovation in and of itself (new technologies exist, but they do not spread widely throughout the productive structure). In different terms, the difficulty lies not in generating ideas, but rather in absorbing them. Furthermore, additional evidence deepens this perspective by quantifying the degree of dispersion within industries. In effect, when one examines firm-level datasets, a persistent increase in productivity dispersion becomes visible: leading firms in many sectors exhibit rising efficiency, rapid digital adoption and effective use of data-driven processes, all the while lagging firms show minimal progress. Importantly, this gap is neither transitory nor transient. Indeed, it has steadily expanded for nearly two decades and appears to have affected a wide range of industries. This widening dispersion indicates that technological diffusion is increasingly uneven. Historically, general-purpose technologies have eventually permeated the entire economy, raising productivity across the board. Contrastingly, in the digital era, many firms face significant barriers to adopting new technologies. These barriers include organizational inertia, limited managerial capability, insufficient worker training, high fixed costs and complex integration challenges. In the same vein, this explains why innovation at the frontier coexists with stagnation elsewhere.
Additionally, understanding why diffusion has slowed requires a broader historical and sociological reflection. Indeed, earlier waves of technological progress and innovation, such as electrification, refrigeration, sanitation and internal combustion, fundamentally reshaped production and daily life. Put differently, they diffused across firms and households in a profound and transformative manner. Contemporary digital technologies, despite their extraordinary computational power, often produce more incremental improvements, and many of them require complementary investments in skills, organizational routines and firm structures. Still, without these complements, their productive potential remains latent. Moreover, some modern innovations enhance communication, entertainment, or personal convenience rather than directly boosting industrial output. In light of this, it can be argued that the nature of innovation has shifted: it remains spectacular at the frontier but translates less readily into broad-based productivity gains. In fact, adoption is more challenging, diffusion is more uneven and organizational transformation is more demanding.
Although this structural explanation accounts for much of the evidence, a second explanation begins from the recognition that modern economies are difficult to measure. Importantly, the measure of GDP was designed for an industrial economy built around physical goods, all of which have been assigned a given value (which is positive and non-negligible). As such, when consumers use free digital services, such as online maps, search engines, translation tools or learning platforms, the official statistics assign these goods a monetary value of zero. Nonetheless, the welfare they extract from the aforementioned services is immense, given that these goods permit them to enhance information access, reduce transaction costs and improve daily life in ways that traditional measures do not capture. In light of this, some researchers propose that productivity is not truly slowing; instead, they assert that the gains produced by digital technologies are largely invisible to the metrics used in official national accounts. In other terms, productivity data grossly understate and underestimate the value created by digital innovation because they have omitted the significant consumer surplus generated by zero-price goods. Further to this, a second difficulty in measurement arises from intangible capital: modern firms rely heavily on assets such as software, brand equity, data architectures and organizational systems. These assets are difficult to quantify because their value unfolds gradually and often interacts with other forms of capital in a complex and nonlinear manner. For instance, a company may invest heavily in a data platform that enhances decision-making, yet the improvement may only materialize after new processes are implemented. Thus, traditional accounting frameworks struggle to capture such investments; even when intangible capital is included, adjusting for quality improvements remains problematic. Notably, when the power or functionality of a device increases while its price remains constant, the quality improvement is often underrepresented in price indices. In the same vein, these measurement challenges imply that TFP may be understated in economies where intangible capital and digital goods dominate. Nevertheless, a careful examination of the evidence reveals that mismeasurement, despite its supposed importance, cannot be used to fully explain the observed phenomenon of productivity slowdown, as the timing between the two does not align in a manner that could corroborate such a relationship. Indeed, digital goods and intangible investments expanded significantly before the slowdown began. In this context, if mismeasurement were the primary cause, one would expect a more gradual decline rather than a sharp break. Furthermore, productivity growth slowed even in sectors where measurement is relatively straightforward (manufacturing, for instance, provides a clear example). In other words, measuring physical output in manufacturing is comparatively easy, yet productivity growth there has declined as well. In light of this, it becomes difficult to maintain that the slowdown is an illusion generated by statistical shortcomings. As such, the weight of evidence indicates that mismeasurement accounts for a fraction of the decline, but the significant majority of the observed slowdown is genuine and legitimate.
All in all, these two explanations illuminate distinct facets of the modern economy. The structural explanation clarifies why rapid innovation at the frontier has not translated into broad-based improvements and accounts for the widening distribution of productivity outcomes across firms, highlighting the importance of organizational, managerial and institutional factors (for instance, an appalling lack of worker training or a hesitant managerial leadership). In the same vein, it adequately situated the modern economy within a longer historical trajectory where the most transformative technologies required substantial complementary investments. The measurement explanation clarifies the disconnect between visible technological progress and flat productivity indices, and posits that digital consumer surplus is real yet invisible, and that intangible capital remains difficult to quantify. However, in spite of their differences, the two perspectives cannot be interpreted as mutually exclusive or contradictory. Indeed, they offer instead complementary and reciprocal views of a complex reality. The digital economy generates significant welfare gains that GDP does not capture, yet these gains do not directly translate into output growth or firm-level productivity. Meanwhile, diffusion barriers constrain the extent to which technological innovations reshape the broader economy. In simpler terms, the interplay of these elements produces a world where innovation feels abundant, but measured output grows slowly.
Sources:
Andrews, Dan, Chiara Criscuolo, and Peter N. Gal. “The Best versus the Rest: The Global Productivity Slowdown”. OECD Productivity Working Paper No. 5. Paris: OECD Publishing, 2016. https://www.oecd.org/en/publications/the-best-versus-the-rest_63629cc9-en.html.
Brynjolfsson, Erik, Avinash Collis, and Felix Eggers. “Using Massive Online Choice Experiments to Measure Changes in Well-Being.” Proceedings of the National Academy of Sciences 116, no. 15 (2019): 7250–55. https://doi.org/10.1073/pnas.1815663116.
Byrne, David M., John G. Fernald, and Marshall B. Reinsdorf. “Does the U.S. Have a Productivity Slowdown or a Measurement Problem?” Brookings Papers on Economic Activity, 2016. https://www.brookings.edu/articles/does-the-united-states-have-a-productivity-slowdown-or-a-measurement-problem/.
Corrado, Carol, Charles Hulten, and Daniel Sichel. “Intangible Capital and U.S. Economic Growth.” Review of Income and Wealth, Series 55, Number 3, September 2009 . https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1475-4991.2009.00343.x.
Fernald, John. “Productivity and Potential Output Before, During, and After the Great Recession.” In NBER Macroeconomics Annual 2014, edited by Jonathan A. Parker and Michael Woodford, 1–51. Chicago: University of Chicago Press, 2015. https://www.journals.uchicago.edu/doi/full/10.1086/680580.
Foster, Lucia, Cheryl Grim, John Haltiwanger, and Zoltan Wolf. “Innovation, Productivity Dispersion, and Productivity Growth”. NBER Working Paper No. 24420. Cambridge, MA: National Bureau of Economic Research, 2018. https://www.nber.org/papers/w24420.
Gordon, Robert J. “The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil War”. Paperback edition. Princeton: Princeton University Press, 2017. https://doi.org/10.1515/9781400888955.
Solow, Robert M. “A Contribution to the Theory of Economic Growth.” The Quarterly Journal of Economics 70, no. 1 (February 1956): 65–94. https://www.jstor.org/stable/1884513.
Syverson, Chad. “Challenges to Mismeasurement Explanations for the U.S. Productivity Slowdown.” Journal of Economic Perspectives 31, no. 2 (2017): 165–86. https://www.aeaweb.org/articles?id=10.1257/jep.31.2.165.
Triplett, Jack E. “The Solow Productivity Paradox: What Do Computers Do to Productivity?” The Brookings Institution, March 1, 1999.https://www.brookings.edu/articles/the-solow-productivity-paradox-what-do-computers-do-to-productivity/.