Tyler Cowen's The Great Stagnation (2011) argued that apparent innovation since the 1970s masked a deeper reality: the transformative, broadly-shared productivity gains of prior eras—electricity, automobiles, antibiotics, mass education—had been picked. What remained was incremental improvement delivering marginal benefits to narrow populations. Median household income stagnated. Productivity growth decelerated. The internet changed how people consumed entertainment and communicated but did not change the median wage. Cowen later identified the stagnation's end around 2020 with the arrival of mRNA vaccines, large language models, and GLP-1 obesity drugs—a cluster of genuinely transformative innovations breaking the forty-year pattern. The question for the 2020s is whether societies can capture the growth these innovations make possible, or whether institutional bottlenecks will convert technological breakthrough into another round of disappointing aggregate statistics.
The stagnation thesis was controversial because it challenged Silicon Valley's self-conception. The technology industry believed it was transforming civilization. Cowen's data showed it was mostly transforming leisure—faster phones, better streaming, more addictive social media. Genuine improvements in quality of life, but not the kind that moved median incomes or productivity statistics. The internet gave everyone access to information; it did not give them higher wages or better healthcare. The great innovations of the postwar era—antibiotics, jet aviation, interstate highways, mass higher education—had produced broadly-shared prosperity because they addressed fundamental constraints on production and distribution. The internet innovations addressed constraints on entertainment and communication, which mattered but were second-order.
Cowen's periodization places the stagnation's end at 2020, identified not by a single event but by a cluster: the mRNA vaccines (demonstrating a new platform technology for medicine), GPT-3 and its successors (demonstrating that machines could handle natural language at human-competitive levels), and semaglutide/GLP-1 drugs (addressing obesity at pharmacological scale). Each represented not incremental improvement but a phase transition—a technology crossing a threshold that made previously impossible things routine. Cowen: 'Add all that up, and we're many times past the great stagnation being over. It was not a trickle. It's been a flood.' The flood's economic effects lag the technological breakthrough by the institutional bottleneck interval, but the direction is clear: stagnation has ended, and the question is how much of the resumed growth societies can capture.
The structural shift from stagnation to growth carries its own dangers. Cowen notes that under many AI scenarios, the more unhappy people are, the better the economy is doing—because unhappiness correlates with the speed of change, and speed correlates with the magnitude of transformation being captured. A society experiencing rapid, broadly-distributed growth necessarily experiences rapid, broadly-distributed disruption. The professionals whose skills are being repriced are unhappy. The students whose educational investments are depreciating are unhappy. The parents uncertain about what to tell their children are unhappy. None of this unhappiness contradicts the claim that the economy is improving; it is the price paid for improvement that comes too fast for comfort. The policy challenge is not preventing the unhappiness—that would mean preventing the growth—but ensuring the gains are distributed broadly enough that the unhappiness does not metastasize into political backlash.
Cowen developed the stagnation thesis in the late 2000s as he observed the gap between innovation rhetoric and median wage reality. The Great Stagnation was published as a short Kindle ebook in 2011, became a surprise bestseller, and shaped elite economic conversation for the decade that followed. The periodization of the stagnation's end comes from Cowen's 2023-2025 reassessments, where he identified mRNA vaccines as the turning point and large language models as confirmation that the technological frontier had resumed advancing at transformative rather than incremental rates. The framework is distinctively Cowenian: empirically grounded, willing to revise prior claims when evidence shifts, and focused on the question of what actually moves median welfare rather than what generates headlines.
Apparent innovation masked real stagnation. The internet age produced genuine quality-of-life improvements but failed the median-wage test—incremental rather than transformative at the level that matters for broadly-shared prosperity.
Stagnation ended around 2020 with a cluster of breakthroughs. mRNA vaccines, large language models, and GLP-1 drugs represent phase transitions, not improvements—technologies that make the previously impossible routine.
Growth resumption precedes institutional capture. The technological potential for multi-percent annual growth exists; whether societies capture half a point or two points depends on institutional adaptation speed.
Unhappiness tracks transformation speed. Rapid growth produces rapid disruption; the professionals being repriced are rationally unhappy even when the economy is rationally improving—the pain is real, the gains are also real.