Our Understanding of Inequality on the Cusp of AI Disruption.
Our Understanding of Inequality on the Cusp of AI Disruption.
Bill Bowen’s posting last week emphasized the dire threat we all face from systematic attacks on the core social institutions of scientific inquiry in the U.S. As Bill wrote,
“… science is not an infallible collection of facts or a priesthood of experts. Rather, it is a method—a disciplined, collective process for identifying and correcting errors. Its credibility stems from transparency, replicability, and open critique, not from authority. Science thrives on being challenged, revised, and improved.”
Using scientific research to inform and improve public policy is an inherently dynamic process. Scientific understanding of problems will change as the scientific process itself evolves with new data, new methods, and new perspectives. Consequently, public policies to address problems should evolve over time as our scientific understanding of those problems evolves.
It is difficult, however, for public policy to respond to the evolution of scientific research findings. During the Covid-19 pandemic, for example, advice from the U.S. Centers for Disease Control (CDC) attempted to keep up with the latest data about how the new virus could be transmitted. Yet the constantly changing advice, especially during the early months of the pandemic during the Spring and Summer of 2020, created fertile opportunities for politicians to weaponize for partisan purposes the otherwise professional process of issuing updated public health advisories based on evolving scientific research about the new deadly pandemic.
Despite the toxic politics, efforts to end the pandemic benefited from having at least one clear measure for all to track: the rate of new infections. We know how to measure it. Despite the myriad social and political disruptions fostered by those who leveraged the pandemic for political gain, the first Trump Administration’s Operation Warp Speed project, begun officially in May of 2020, was able to use science to develop vaccines rapidly to gain control over that measure in record time.
Yet in hindsight the challenges of responding to the pandemic may look simple compared to the upcoming challenges we will likely face as we respond to dislocations caused by artificial intelligence (AI). The critical measures to monitor are more varied, our methods for measuring them are more contested, and the social, economic, technological, and political dynamics involved in these measures are more complex.
One critical measure that will undoubtedly be involved is the concept of economic inequality. Any effort to understand the impacts of AI on economic inequality need to be informed by the strengths and weaknesses of our current understanding of this phenomenon.
Income inequality is one of the most studied and debated components of economic inequality. High rates of income inequality provide strong evidence of social dysfunction whereas lower rates suggest more social justice. The following summary of current social science research on income inequality describes the research framework that will be the starting point for understanding the impacts that AI may have on this important phenomenon.
The traditional measure of income inequality is known as the Gini Coefficient, which uses broad macroeconomic data on a national basis to estimate overall inequality at any one point in time. Recent advances in the availability of much more disaggregated data over many years, including income earned by both labor and capital, has resulted in much better social scientific methods for measuring this important indicator.
Thomas Piketty’s landmark 2014 book, Capital in the Twenty-First Century, is perhaps the most famous example of this new wave of social science research. Piketty’s work has framed the academic and policy debates on this topic since it was published.
Piketty uses detailed, individual tax returns and inheritance records across several centuries to assess changes in economic inequality. He examines details about the accumulation of financial capital, the distribution of its ownership, and the distribution of the income it generates.
As with most economic historians, Piketty devotes considerable attention to examining a core element of Karl Marx’s critique of capitalism. Marx argued that the process of capitalist economic development would inevitably create extreme inequality because the financial return to capital would always be greater than the overall rate of economic growth.
Consequently, Marx argued that the share of wealth that would be available to labor in the form of wages would continue to shrink, thereby constantly reducing wages and reducing all workers to perpetual poverty. All sources of social authority within Marx’s view of industrial social relations would yield in response to the overwhelming tyranny of the new, elite, capitalistic class.
The only path to create different industrial social relations, Marx argued, would require a social revolution to transfer ownership of all capital, and thus all returns to capital, to the state, which would be entirely in the hands of a party devoted to the workers.
Piketty’s historical analysis of data from the 18th and 19th centuries confirmed Marx’s argument that the returns to capital were consistently larger than the overall rate of economic growth in those years. Yet Piketty’s analysis revealed that relationship was interrupted during the 20th century by the unexpected destruction of capital that occurred during World War One, the Great Depression that ensued afterward, and the cataclysm of World War Two. This era destroyed several generations of accumulated capital.
When peace was restored, a new era emerged during which labor’s share of income could grow even while the returns to capital were restored. The postwar era gave rise to a large middle class, an upwardly mobile working class, and a much more powerful state capable of major redistributions of capital to investments in infrastructure and social welfare. To many observers, it looked like capitalism had avoided Marx’s predictions without the need for social revolution, albeit at the cost of horrific wartime traumas.
Yet Piketty argues that since 1980 capitalism’s inherent dynamics have reasserted themselves. Marx had it right all along. The postwar era was not the beginning of a new type of capitalism, Piketty concludes. Rather, it was a temporary anomaly brought on by the unexpected destruction of so much capital from the violent cataclysms of wartime.
Since 1980 the rate of economic growth has begun to slow down once again, yet the returns to capital remain high, and are growing even higher. The owners of capital have quickly reasserted their social authority to limit, and then restructure, the political power of the postwar welfare state. Inequality began to grow once again, even within the European Union. The European national traditions of using state policies to redistribute income were indeed more extensive than the U.S. tradition, Piketty acknowledges. Yet even European governments have been unable to stop the inevitable resurgence of inequality that has taken hold of capitalist economic development in the 21st century.
Piketty’s social scientific assessment of the causes of income inequality has been widely accepted among many economists and policy makers. Until recently, his work has framed the debate about what can, and cannot, be done to reduce income inequality. Yet the scientific research process continues to challenge Piketty’s framework because that’s what the scientific process does.
For example, the Swedish economics professor Daniel Waldenström and two American economic policy advisors, Gerald Auten and David Splinter, have recently published new challenges to Piketty’s findings based on their own critiques of Piketty’s methods and new data Piketty did not have.[i]
These newer analyses conclude that income inequality today is much less than Piketty’s data suggest, largely because both European and U.S. tax policies and social spending programs compensate for capitalism’s inherent trend toward income inequality much more than Piketty’s analysis concluded. The 20th century welfare state may be less powerful in the 21st century, but its effects remain very effective at blunting capitalism’s inherent push toward income inequality.
In addition, today’s rates of return to capital may indeed be higher than the return to labor. Yet the recent research reveals that workers today have widespread ownership of capital through pension plans and other tax-preferred methods of accumulating capital. Consequently, the effective income received by labor is thereby increased. Furthermore, hundreds of millions of employed people of all income levels own their homes today, which creates another previously uncalculated source of wealth that needs to be factored into a broader understanding of income that is received by labor.
This brief (and woefully incomplete) summary of our current social scientific understanding of the complex dynamics of income inequality is the evolving framework for our understanding of this aspect of economic inequality at the beginning the AI’s dissemination throughout the global economy. Consequently, it is our starting point for designing research programs to test various hypotheses about AI’s impacts on this important aspect of our economic performance. After all, a core benefit of the scientific method is that excellent research always leads to more refined questions about next steps. Some critical questions to investigate include the following. Many more will follow.
First, if AI creates historically important increases in overall economic growth (i.e. GDP), how will it affect the relative relationship between returns to capital and returns to labor? There’s good reason to suspect that the result will be a dramatic decrease in the returns to labor from wages and salaries with corresponding increases in returns to capital. If correct, how will these returns vary in different places within the global economy?
Second, if AI reduces the returns to labor as much as some people suggest because of widespread elimination of millions of jobs, how will the loss of consumer income impact prices, and thereby the profits that turn into high returns to capital?
Third, if today’s tax policies and social spending programs in Europe and the U.S. are more effective than previously thought at blunting capitalism’s inherent push toward income inequality, can today’s tax and spending policies be scaled up enough to compensate for AI’s potential to cause ever higher rates of income inequality? At what scale do current policies become fiscally unsustainable, programmatically ineffective, or both?
Fourth, what level of job destruction caused by the dissemination of AI, and what scale of financial support from tax and social spending programs, combine to create a new social framework in which income inequality is no longer an effective measure of the broader concept of economic inequality?
Fifth, will the widespread dissemination of AI in the global economy make the entire social science framework that defines economic security primarily through the lens of income earned by capital and labor obsolete? If so, what new framework will be needed to measure economic security for individuals and households in a global economy no longer dominated by human labor?
These questions may be useful, or maybe not. But they represent the kinds of continual questioning that is the heart of the scientific method. There seems to be little doubt that we are at the cusp of an era of widespread disruption that will be driven by the dissemination of artificial intelligence. It is almost unthinkable that we are also on the cusp of widespread attacks on the core social institutions of scientific inquiry.
Our institutions that conduct social science research that is independent, replicable, and transparent need to be defended vigorously. These institutions subject all their research products to rigorous, independent, peer review to ensure that errors are corrected, and that additional research builds on solid previous scholarship. They are essential if we desire to create effective public policies to manage the complex social, economic, and political challenges that will evolve as new forms of AI permeate our lives.
Bob Gleeson
[i] Daniel Waldenstrom, Richer and More Equal: A New History of Wealth in the West (2024); Gerald Auten and David Splinter, “Income Inequality in the United States: Using Tax Data to Measure Long-term Trends”, September 29, 2023, available at www.davidsplinter.com.