Reforming capital markets to build broad-based prosperity and reduce economic inequality

In his latest post, PDI Project Lead on Broadening Corporate Governance Participation, Tom Powdrill zeroes in on a timely report from Mainstream — a pro-Burnham Labour faction — titled The Productive State: A Framework for Manchesterism. The report makes a supply-side case for public corporations in energy, water, housing, and transport, but what makes it especially relevant to PDI's work is the governance model it proposes: arm's-length operational independence, workers on boards as a foundational design feature, and democratic accountability running outward to workers and communities rather than upward to ministers. As Tom puts it, corporate governance reform is not merely an accountability mechanism, it is a building block of predistribution.

This discussion paper, the first in PDI's Predistribution AI Lab series, analyzes four scenarios, modeling the cascading effects of income erosion and unemployment on consumption, tax revenue, mortgage markets, corporate debt, equity values, pension systems, and insurance assets. Scenarios are designed to illuminate key transmission channels of potential macro-financial risk through the real economy, markets, and to diversified investment portfolios. Three scenarios are based on higher unemployment numbers as predicted by Anthropic CEO, Dario Amodei, while the “lighter” scenario is built on the historical precedent of declining returns to labor and a “fissured workplace” even as employment has grown with technology. We do not take a view on whether AI will lead to higher unemployment. Rather, we center our attention on the risks of historical and ongoing declining returns to labor that are shared across stakeholders in society, including financial risks to diversified investors’ portfolios. We argue that the advent of AI is an inflection point at which the world is either poised to deepen the current trends toward risk, or at which we can sculpt and refine economic structures to avoid such risks. This first paper primarily focuses on macro-financial analysis. However, safety, blind spots, and cognitive bias risks are also considered, particularly in Part II of the discussion paper series.

This discussion paper, the second in PDI's Predistribution AI Lab series, offers practical models for addressing risks identified in Part I of the series, particularly via broadening equity-linked compensation and corporate governance participation to include workers, communities, and content creators who take risk and create value alongside executives and investors in the production process. The report further contextualizes the current technological transition in the broader backdrop of decades of declining returns to labor versus capital, depreciating cash relative to appreciating asset values, and corporate governance that is more strongly oriented toward shareholders versus other corporate stakeholders. At the core of Part II is a recognition that the economy and productivity have been advanced in recent history through the contributions of workers (formally and informally employed, collectively “human capital”), communities who host infrastructure and natural resources projects (collectively “social capital”), and content creators and others who provide data which has advanced technology (a mix of human and social capital). However, with financial capital being prioritized by corporate governance over human, social, and natural capital, these other stakeholders have not been compensated in a manner that keeps pace with financial capital, resulting in rising economic inequality, misalignment of incentives across stakeholder groups, disenfranchisement, loss of trust in institutions, polarization, and domestic and geopolitical conflict. Broadening ownership and governance of companies can align stakeholder incentives to contribute to technological advancements, leverage important perspectives to safely train and roll-out AI, and sustain the aggregate demand upon which the economy and financial portfolios depend.