This is the second post in our series discussing The Meritocracy Trap by Daniel Markovits. You can find all of the posts in the series here.
***
Most of the critical attention directed at Daniel Markovits’s The Meritocracy Trap has focused on its claim that well-off parents launder inequality through schooling. While Markovits brings masterfully comprehensive reams of data to bear on the concept of the “meritocratic inheritance,” the most original and provocative part of the book comes later, when Markovits offers his explanation of why educational sorting has come to matter so much: elite schooling leads to top jobs, and “[t]he top jobs pay so well because a raft of new technologies has fundamentally transformed work to make exceptional skills enormously more productive than they were at mid-century and ordinary skills relatively less productive.” This is provocative because it contradicts the pervasive myth that technological change is natural, self-directing, or inevitable. Few reviewers have remarked on this part of the book or reflected on what it suggests: that industrial policy will be vital to building a more equal economy.
Markovits offers a materialist theory of income inequality: the prevailing mix of production technologies has pushed high incomes higher and low incomes lower. It is crucial to distinguish technological materialism from technological determinism. Materialism says that technology has a predominant influence on social and economic structure, while determinism goes one step further and treats technology itself as exogenous to the social system. Markovits is not a determinist. He believes we can explain the course of technological development through law, culture, and institutions like the educational meritocracy: the emergence of highly skilled workers “induced the innovations that then favored their elite skills. A rising supply of meritocrats stimulates its own demand.”
The observation that law and policy brought us a suite of technologies that tend toward inequality naturally raises the question of whether we can redirect the course of technological development toward equality. As a democratic society, we should actively choose technologies that make life less tedious and open up the space for human flourishing. One such consideration is whether a given technology concentrates returns among the few or the many. The notion that polities should guide the course of technological change is known as industrial policy—a tradition practiced but rarely acknowledged in the United States. In his Conclusion, Markovits provides several targeted industrial policy recommendations, and I propose several more below.
Markovits argues that, across industries, information and communication technology has allowed firms to strip discretion from front-line workers and amplify the decisions made by executives and professionals at headquarters. In retail and food service, production has been shifted “up the supply chain” such that workers onsite at McDonald’s are needed only to open packages and press buttons to heat up pre-cooked food. Even in finance, local bankers are no longer valued for their ability to tell the difference between a good loan and a bad one because securitization (itself the product of legal choices and information processing capacity) dampens the significance of each individual loan in a downstream investor’s portfolio.
And across industries, middle management has been the most downsized category since the restructurings of the 1980s. Companies increasingly use “algorithmic management”—encompassing technologies from sensors and bar code scanners to natural language processing (for automatically scanning employees’ emails) and real-time routing software (for directing logistics workers where to drive next)—to coordinate large workforces without direct supervision. Ben Tarnoff describes the new form of control that these technologies make possible as “discipline at a distance.” And as Brishen Rogers observes, algorithmic management often leads to fissuring, or severing work, through subcontracting, from the legal protections of employment.
It may be tempting to read Markovits as raising a version of the now-familiar automation panic, but his argument, as I read it, does not rely on a diagnosis that jobs are now or ever to be replaced by software and robots. The diagnosis is instead about polarization. As economists David Autor and David Dorn have observed, the numbers of both low-wage and high-wage jobs have increased in recent decades, while middle-wage jobs have declined. Especially when we couple this with Rogers’ and other labor scholars’ observation that low-wage jobs are increasingly precarious and poorly protected, polarization has already become a sufficiently dire trend on its own.
On the high end of the polarized income spectrum, Markovits’s materialism works better as an explanation for the skyrocketing incomes of technical and professional workers than for those of executives. Executive compensation (whether formally stock-based or not) reflects demand in both the labor market and the financial markets. A CEO who is 1% better than average may be worth $1M or $100M to investors; the difference is a product of a host of factors, including the company’s size, the quantity of worldwide capital competing for returns, and the various legal and competitive factors that determine the present value of the executive’s future plans. The understanding that executive compensation follows from financial market considerations—or, in other words, that executive compensation is capital rather than labor income—should suggest that returns to skill cannot explain the magnitude of its growth. That is, reducing the very highest incomes will require financial market reforms more than anything else.
Markovits’s account of changes in the labor market over the 20th century fits within a theory known as skill-biased technical change (SBTC). To be clear, the concept of “high-skilled” and “low-skilled” workers is descriptively questionable (I suspect I have fewer skills than most people in the workforce) and morally problematic; the terms actually refer to educational attainment and no more. In any case, Markovits views the fact that innovation has come to favor education and thereby widen inequality as historically contingent. Following the economist Daron Acemoglu, his main explanation is that innovation “chased the new supply of skill that meritocratic education unleashed.” He compares the United States unfavorably to Germany, a country that has spread educational attainment more broadly through vocational training, and made that education valuable by investing in the manufacturing sector, where there are more middle-skill jobs.
Most analysis of technology and the future of work takes the position that polarization is a one-way street. Information technology will continue to deskill workers, and the only question is how to compensate the losers—all roads lead to universal basic income. Markovits resists such deterministic pessimism. He calls for a set of policies to rebalance production toward middle-skill labor: undo prestigious professionals’ monopoly on providing certain services (e.g. empower nurses vis-à-vis doctors); rein in the market for corporate control to empower middle managers vis-à-vis top executives and their capital-owning allies; crack down on subcontracting to promote long-term employment; remove the cap on the payroll tax, which currently incentivizes firms to hire one person for $1 million rather than ten people at $100,000 each; and pay wage subsidies to firms hiring mid-skilled workers.
Industrial policy can go even further, into the heart of the algorithmic systems most responsible for deskilling the work of all but the engineers needed to run them. Restricting the collection of personal data—as many have advocated for fairness and privacy reasons—would also diminish the effectiveness of business models based on stratifying and sorting people. Algorithmic credit rating, for example, would suffer, reviving the need for human loan officers to make interpersonal judgments. Algorithmic systems are probably here to stay for many data classification tasks, but even there, law and industrial policy can promote human-in-the-loop engineering, an approach where humans are responsible for classifying a subset of the data, iteratively improving the algorithm as they go. To date, human-in-the-loop jobs have not been good jobs (think Facebook content moderators), but with the right labor and employment protections, they could be.
Ultimately, the information technology industry may be predisposed to an unequal wage distribution. Software scales unlike any other good or service, so it is not surprising that relatively few firms, with relatively few workers each, can code the world. United States policymakers should therefore recognize that so long as Silicon Valley is the nation’s leading growth industry, the American economy will continue to grow in an unequal direction. For reference, the sectors most responsible for a decline in labor’s share of income in the 21st century are retail, technology-related services, and mining. Taking inequality seriously would mean an industrial policy that promotes labor-intensive, middle-skill work—for example, construction, mechanical maintenance, and healthcare. One argument for a Green New Deal is that the construction of solar and wind generation, transmission, and storage on a massive scale, along with investment in nontraditional green sectors like care work, would do just this.
For LPE scholars and practitioners, who already believe that law’s meaning is socially and politically constructed, it should not take a leap to recognize that technology is similarly indeterminate. I hope this framing will convince some of my friends on the left—justifiably scarred by the history that Markovits recounts—that Silicon Valley corporations do not have a monopoly on the notion of technological progress. Progress can and should be redefined to serve progressive ends.
