The Productivity Paradox: New Models for Organized Bargaining

Since the late 1970s, the American economy has allowed its productive dominance to wither. Iconic U.S.-based corporations ceded market share to (mostly) Asian competitors, often opting not to fight back and instead outsourcing and offshoring their own production to factories in lower-wage places.

Restoring large-scale, competent manufacturing in the United States will require new and comprehensive economic policy. Such policy will need massive government support, enforcement of significant new rules, and, perhaps paradoxically, the reversal of five decades of attacks on workers and their labor unions.

Indeed, the new policy will have to put the worker front and center in the firm. Workers’ expanded role will be born of, and maintain, a higher level of labor-management conflict than the United States has seen since the early 1940s. Although it may be surprising to many readers, this would be a very good thing, so long as that conflict is about effort and reward in the context of higher productivity—which we will refer to as productivity bargaining. While na­tional pattern bargaining (the standard approach of most U.S. unions) today is effective at setting baseline compensation, it misses the bigger picture of productivity and, therefore, costs. Productivity bargaining, by contrast, is focused on local production and plant performance to identify input factors whose underutilization adds avoidable costs and reduces output. Our approach links shop-floor conflict to productivity growth.

The fundamental question is: can a revitalized union movement spur measurable improvements in American industrial productivity and, if so, how? What would workers, unions, companies, and policymakers need to do differently? And at what scale?

Deindustrialization and the Decline of the U.S. Auto Industry

The consensus story of deindustrialization is well known: Asia’s mercantilist juggernauts built wealth through lower-cost manufactured exports, while American corporations relocated their production to lower-wage nations in re­sponse. In industry after industry, U.S.-based corporations reacted to Asian competition by exiting the low end of each market (commodity textiles, steel, small cars, low-end appliances, and computer chips), offshoring and outsourcing to reduce the draw on their own capital, closing their own plants, and seeking concessions from their remaining unionized employees. Given our own backgrounds in the auto industry—one working at General Motors and the other at the United Auto Workers’ (UAW) headquarters—we will focus our analysis of deindustrialization, and possible reindustrialization, on that sector.

In auto manufacturing, the trio of (formerly) Detroit-based automakers had a domestic market share of 84% in 1978; that dropped to 64% by 2000 and to 43% last year.

Deindustrialization in the auto manufacturing sector was a national disaster that resulted from two types of unforced errors. The first was a funda­mental misunderstanding of the economics of manufacturing firms by their management, by neoclassical economists, and by human resources specialists. As U.S.-based automakers lost market share to lower-cost Japanese companies, American managers and economists drew two incorrect conclusions. The first was that the Japanese advantage was rooted in lower vertical integration, which was a misreading of Japan’s keiretsu arrangements, in which automakers held equity in and backstopped investments by their suppliers. The second was that Japanese labor relations were more “cooperative” and less adversarial. While it is true that Japan’s militant, Communist-led unions were broken (with American help) in the 1950s, this too was a fundamental misreading: the “labor peace” that Japan-besotted academics thought they were seeing was, in fact, based on a system in which factories were run by a highly engaged workforce over­seen by surprisingly few managers. We will return to this point.

A second critical error was Wall Street’s punishment of those American manufacturers with the temerity to reinvest earnings in their own growth and process improvement. Led by Drexel Burnham Lam­bert, a generation of “corporate raiders” laid waste to scores of iconic companies, demanding that they instead return earnings to shareholders. In the 1980s, Drexel and subsequently others threatened, and in some cases took over, companies using infamous “greenmail” and “highly confident letter” tactics, and in the process drove the dismantling of targeted companies’ vertical integration, so as to “free up” their capital. In a context of burgeoning investment by Asian industrialists and governments, this was a recipe for disaster. U.S.-based companies were hollowed out just as competitors in Japan, then Korea, and finally China forwent short-term profits in pursuit of market share gains and future technological dominance.

The auto story of the 1980s makes clear how wrong-headed these imagined solutions were. Until the U.S.-based automakers began to bleed market share in small cars in the 1970s—first to Volkswagen and Renault, and later to Toyota, Nissan, and Honda—they (and especially GM) maintained high levels of vertical integration. These companies invested enough capital (equipment and tools) to ensure that the productivity of their (mostly UAW) workforce would easily offset their higher pay relative to nonunion workers. In 1978, two in every three workers in the motor vehicle and parts industry worked at GM, Ford, or Chrysler; only 336,000 worked at supplier-owned plants, and 195,000 of those worked at UAW-organized supplier facilities. Only 141,000, or 13.7%, of the workers in the sector were not in unions. The so-called Detroit Three automakers built 83.6% of all cars and light trucks sold in the United States.

Two decades later, the Detroit Three’s share was down to 68.2%. Of the 1.276 million men and women working in the auto and auto parts sector (which was 250,000 more than in 1978), 913,000 (or 71.6%) were working in nonunion plants, roughly 200,000 at the U.S. factories of Europe- and Asia-based automakers, and more than 700,000 at nonunion supplier plants, a five-fold increase from two decades prior.

As Stephen Herzenberg, an economist in the U.S. Department of Labor, noted in his seminal 1991 study, sharply declining Detroit Three employment due to outsourcing to nonunion plants had unexpected results. Most nonunion suppliers, pressured by oligopolistic price cutting by their customers, faced substantially higher costs of capital than the largely self‑financed automakers, which could also access highly liquid public capital markets; as a result, those suppliers had sharply lower productivity.¹Stephen Herzenberg, “The North American Auto Industry at the Onset of Continental Free Trade Negotiations,” Department of Labor—Bureau of International Labor Affairs, 1991. For their part, the Detroit Three automakers lost control of production costs and component quality. (That lower productivity was no doubt a large part of why industry employment grew by nearly a quarter, though 1998–2001 were also strong new-vehicle sales years.) Thus, outsourcing most parts-making didn’t result in measurably higher Detroit Three profits; those only came a decade later, courtesy of the dumb luck of a huge swing from cars to pickups and SUVs (which cost about $4,000 per unit more to build but sell for an average of $11,000 more). Ironically, because the UAW contract required the self-shrinking automakers to slough off workers in seniority order, the most experienced—and thus highest-paid and most skilled union workers—survived the layoffs. Had the Detroit Three been able to cast off their most expensive workers, the results of their downsizing would have been far worse.

Moving work from high-wage, high-productivity facilities to lower-wage, lower-productivity facilities was the result of the new Detroit Three strategy encouraged by Wall Street’s Drexel school and abetted by academics. In theory, the Detroit Three, by following this strategy, could have freed up capital to design and build competitive new models, to optimize their facilities to eliminate cost-sapping bottlenecks, and to invest in upgrading their workers’ technical knowledge and capacity to operate plants with less supervision. But, in reality, they did not. If anything, these automakers became less competitive while their market share continued to decline.

Toward a Union Resurgence?

Throughout this long history of decline, the auto companies stepped up their attacks on the remaining union members. They threatened to move work away from the plants where the most militant local unions represented employees and into those where locals were most cooperative or, when appropriate, to outside supplier facilities. This “whipsawing” of locals sapped morale and led to the wholesale electoral defeat of local union officials at the most cooperationist locals, many of whom had to be rescued by being brought onto the UAW’s national staff. By tolerating such whipsawing, the national union lost the respect and confidence of many rank-and-file workers. This set the stage for an insurgency in the national union that rebuked the cooperationist strategy and culminated in a militant strike against all of the Detroit Three two years ago.

The 2023 insurgency also reflected members’ reaction to unforgivable mistakes that the UAW leadership made between 2007 and 2019, partly in response to the concessions mandated by the government in return for the $90 billion in loans that bailed out GM and paid for Fiat’s takeover of Chrysler in 2009. By 2014, GM had $30 billion in the bank, but none of it went to reimburse the workers who had “taken one for the team” by giving up $11 billion in pay and benefits. The Obama administration’s Presidential Task Force on the Auto Industry that guided the restructuring of GM and Fiat-Chrysler did not mandate any pay resets, nor (incomprehensibly) did UAW leaders demand them. Auto executives earned more and more, not just because their companies became more profitable but because their hand-picked boards approved dividend increases and stock buybacks that inflated the share prices to which CEO compensation was closely tied.

When the union agreed to an “in-progression” lower tier in 2007, it rightly insisted that lower-tier workers make up no more than 25% of any automaker’s hourly workforce. But first at Stellantis, and later at GM, and finally even at Ford, the UAW ignored that limit and let the automakers exceed the ceiling. That was followed by allowing GM and Stellantis to replace about 10% of permanent workers with temps. As a result, at those companies, more than two-thirds of the non-trades UAW workforce is now composed of temps and lower-tier workers (at Ford, it’s about 40%). In the crisis of 2007–11, the UAW’s looking the other way on contract violations made some sense; but once solid profitability returned in 2014, this diffidence became unconscionable.

Workers took notice. The UAW’s 2023 “Stand-Up Strike” was, by all appearances, the most class-conscious one in recent memory. The UAW’s initial demands seemed outlandishly large—a 40% wage increase, elimination of “tiers” (in which workers who were hired at about half the top rate didn’t get to it for at least eight years), and even a reduction in the work week to 32 hours with no reduction in pay. While the last of these fell by the wayside, the strike won raises that surprised even UAW members: pay gains of 25–50%, a shortening of the pay progression from eight to three years, and even a promise that most of these gains would also be implemented in still-to-be-built elec­tric battery and vehicle factories. Unlike the three previous negotiated agreements, this time UAW members voted by large margins to ratify the new contracts.

The UAW strike is emblematic. American workers are acting feistier than they have in decades. In 2023, 30 significant work stoppages occurred, involving 465,000 workers and resulting in 16.7 million idle workdays (the most since 2000).²Drew DeSilver, “2023 Saw Some of the Biggest, Hardest-Fought Labor Disputes in Recent Decades,” Pew Research Center, January 4, 2024; cited in Arthur MacEwan “What is the State of Organized Labor?,” Dollars & Sense (May/June 2024). In 2024, there were 31.³https://www.bls.gov/wsp/  The UAW’s strike followed several other large, successful strikes, ranging from UPS drivers to Hollywood screenwriters to Amazon warehouse workers and beyond.⁴DeSilver, “2023 Saw Some of the Biggest, Hardest-Fought Labor Disputes in Recent Decades.”

That said, unions remain small and weak. Despite increased union efforts to organize, year-over-year gains in union ranks remain small. It is unclear if recent union victories can be scaled up and, more importantly, whether those victories can precipitate a new labor-capital-state contract that results in higher productivity and lower costs.

Lessons for Reindustrializers: Engaging the Workforce to Improve Productivity

This, then, is the crux of the issue: can an America First policy worthy of the name change the way that U.S. companies do business? Will such policies cause companies to invest more of their own capital, rather than returning it to their shareholders? Even assuming modest decoupling from China and America’s ability to catalyze domestic productive capacity addi­tions, there remains the question of cost. The Asian mercantilist model has come to be rooted in scale and technology, but it began with—and continues to enjoy (or in the case of Japan, enjoyed until about 2000)—advantages from lower pay and, in China’s case, lower material costs resulting from weaker regulation.

One obvious way for American manufacturers to reduce production costs would be to invest heavily in research and development and fixed industrial capital. But that alone probably won’t be enough. Prior waves of willy-nilly substitution of capital inputs for labor inputs generally ended badly. Of course, even at constant output, the reduction in labor input showed up as higher labor productivity (LP). But total factor productivity (TFP)—that is, accounting for all of the factor inputs—typically flatlined when this approach was followed. In the auto sector, the biggest wave of capital-for-labor substitution occurred between 1987 and 2015. But during that period, while labor productivity grew by 132.5%, total factor productivity grew by just 3.4%, suggesting that productivity growth attributable to other factors (mainly capital and materials) was strongly negative. This explains the anemic automaker and supplier profitability of this era: the automakers paid more for materials (includ­ing parts) than when they’d made them in-house, partly because suppliers’ prices had to reflect their higher cost of capital versus what the automakers enjoyed.

And beginning in 2011, when the automakers started to bring on new hires at the much lower wages permitted in their revised contracts with the UAW, even labor productivity flatlined, thanks to a less experienced workforce and the erosion of shop-floor solidarity and learning that resulted from having (in today’s terms) $15- and $30-an-hour workers doing the same jobs, with the latter fighting for less mandatory overtime and the former trying to get every work-hour they could.

Despite the fact that dumping union labor and outsourcing parts-making turned out to be an expensive, ineffective approach, most American manufacturers keep doing it. The recent strike wave in the United States holds promise for a more engaged labor movement.

A useful labor resurgence must overcome not just worker cynicism and individualism but also corporate management’s habit of opposing unions. It must defeat the bankrupt idea—accepted as fact by opinion-leaders in both major U.S. political parties—that class conflict is bad for economic growth.

The HR literature advances a number of plausible hypotheses linking unionization to productivity. Chief among them is that union membership lowers employee turnover because it usually provides workers with better wages, benefits, and greater protection against arbitrary discipline and discharge.⁵See, among many others, the ILR Review symposium on Richard Freeman and James Medoff, What Do Unions Do? (New York: Basic Books, 1984), in ILR Review 38, no. 2 (January 1985). But this kind of union advantage is just the tip of the iceberg of what’s possible if and when union members engage.

One of the first economists to rigorously and quantitatively study the issue of macro productivity in the early postwar period was John W. Kendrick, who worked as an economist from 1946 to 1953 at the Office of Business Economics, the predecessor to the U.S. Bureau of Economic Analysis; later, he was economics professor at George Washington University.⁶John W. Kendrick, “U.S. Productivity in Perspective,” Business Economics 26, no. 4 (October 1991). In 1980, he argued that in the postwar period well-managed companies should share about 60% of total factor pro­ductivity gains with labor. Sharing more than that could endanger the profits required to reinvest in the business; sharing less would increase inequality with no justification, leading to costly labor strife. Kendrick also noted that, by his definition of productivity, “since the late nine­teenth century, the secular rate of growth in real gross product per labor hour in the U.S. domestic business economy gradually accelerated from about 0.5% a year to a maximum average annual rate of 3.5% in the subperiod 1948–66. Since then, it declined to about 1% during the period 1973–78…”⁷John W. Kendrick, “Survey of the Factors Contributing to the Decline in U.S. Productivity Growth,” Boston Federal Reserve, 1980. It is worth noting that the 1948–66 period was one of high unionization, with frequent local and national strikes. Paradoxically, such labor militancy can be harnessed to the benefit of both companies and workers.

Many manufacturing managers, especially those in the trenches on the shop floor, and especially those in industry before 1980, know that conflict is not inconsistent with productivity growth and cost reduction. In the 1980s, Detroit Three executives began to classify each of their remaining factories as being either cooperative or “bad boy” plants. As the companies diverted more and more of their earnings to shareholders, and their market share continued to decline, they responded by closing more and more of their plants. Bad boy plants, of course, were the ones more often closed: if there was too much production capacity, why not shed it where the workforce was most militant? But the data coming in from bad boy plants slated for closure startled the executives: they sometimes exhibited the lowest costs and highest quality among the automakers’ facilities. Moreover, when lower-level managers were con­sulted about this apparent anomaly, they revealed that these plants had been high performers all along, not just after their workers learned they were about to lose their jobs.

Why would that be? Think about what one might call the production economy—the millions of jobs in which a series of tasks must be executed with a high degree of repetition. This sector includes not only factory work, but much of the service economy: medical coding, pro­cessing insurance claims, making appropriate referrals in social service agencies, changing the oil in cars, etc. While there is a great deal of routine, there is also high variance in how much care, attention, and effort workers put in. Particularly when rewards are paltry—as they have been for most such workers since the late 1970s—a large proportion of workers can be expected to figure out ways to do as little as possible without being terminated. (Think of a classroom of students who know what it takes to get a C rather than an F, and who don’t see any great advantage in expending the effort required to get an A.)

Even when compensation is decent, if increases in pay or other valued benefits (e.g., time off) are not visibly tied to worker, department, or workplace performance, the incentive to expend effort and stay en­gaged is weak. This is how capitalism is supposed to work. Labor is a key factor input. For it to benefit, labor must be able to advocate for a return on its efforts, just as capital insists on one. Without the local union contract (and, therefore, in nearly all nonunion plants), workers have had no recourse but to disengage and withhold effort, with the unfortunate productivity results we’ve described. For labor to advocate for a return on its effort, it requires voice, a way to express its concerns and suggestions, and empowerment, a collective institution that has the skill and daring to make that voice protected, understandable, credible, and unignorable.

In this context, the Japanese HR model emulated by U.S. firms since the early 1980s can be seen as a clever workaround. Rather than reward effort and engagement with money, this approach favored grafting a set of new worker-management interac­tions on top of an unchanged compensation structure: set up quality circles, organize workers into teams, have workers and managers share parking lots and cafeterias; rename workers as “associates” or “techni­cians”; listen to their views on which first-line supervisors are despotic and unreasonable. For a while, these workarounds sometimes bore enough fruit to produce compelling anecdotes, but the HR literature of the time produced virtually no quantitative evidence that the cooperationist approach yielded lasting gains in productivity or, therefore, reductions in product or service costs. Even where cooperationism was rigorously attempted, it turned out not to have legs: very few workplaces have such programs in more than name only; and the idea that labor policy even matters has fallen into such disrepute that many employers, large and small, have outsourced most or all of the HR function to outside companies that use cookie-cutter algorithmic approaches to hiring, benefits administration, discipline, and termination.

Labor’s Voice and Productivity in the Auto Sector

Today, having lost confidence in the promise of HR gimmicks, as well as simply replacing labor input with more capital input, we must search for and experiment with new approaches.

With regard to labor, some now conventional practices will need to be rethought. Not only has de-unionization failed to restore productivity gains (and perhaps pushed in the opposite direction) in the Detroit Three, it has also revealed a major drawback to what is called pattern bargaining. In many industries, and particularly in manufacturing and trucking, national unions have sought to negotiate master agreements that apply to all plants of the companies with which the union bargains. Local unions still have agreements covering matters specific to their particular location, of course, but pay levels are almost entirely set by the national pattern. The logic of a national pattern wage for each kind of job was, as the UAW’s longest-serving leader, Walter Reuther, put it, “to take wages out of competition.” But it has had the unfortunate effect of limiting the ability of managers to reward workers for their contribution—arising from the resolution of the conflicts in which they engage—to higher productivity and lower costs. Prior to the 1990s, a portion of workers’ pay could effectively be bargained locally (partly because grievance bargaining and the daily effort-wage bargain could influence the jobs in which particular workers were classified), and there were still a few cents per hour at play locally; but today, there is little leeway for giving financial compensation for cost-saving im­provements. To enlist workers in the fight for meaningful cost reductions in traded-goods sectors, that will need to change. In the postwar United States until about 1980, it was a rule of thumb that about 60% of productivity gains went toward enhancing workers’ compensation.⁸John W. Kendrick’s 1991 article, cited previously, looks at U.S. productivity growth for the period 1889–1998 and comments on the well-documented productivity slowdown from 1973–81. See: Samuel Milner, “The Problem of Productivity: Inflation and Collective Bargaining after World War II,” Business History Review 92, no. 2 (Summer 2018): 227–50. The 60% figure makes a lot of sense, since labor income in the period accounted for about two-thirds of national income and GDP. Since 1980, on average, labor compensation has captured no more than about 20% of productivity gains, the rest split between owners of capital and consumers.

In short, the sharing of productivity gains and cost reductions within the adversarial collective bargaining system provides a bold policy framework. Productivity bargaining acknowledges the plant knowledge, contributions, and innovations of workers on the shop floor. Based on rigorous analysis of 1970s shop-floor data from the U.S. auto industry, it may show a way forward for unionized companies and their workers.

UAW-organized auto plants make for an excellent case study due to a long, well-documented history of local and national collective bargain­ing, powerful local unions, and a highly developed workforce. While working for eight years as an assembly-line worker at a General Motors plant, this essay’s coauthor Craig Zabala studied collective bargaining at the plant level using participant observation research. In doing so, he came upon data that seemed counterintuitive to academic researchers—namely that shop-floor bargaining had a major impact on plant output and product quality, and that so-called bad boy plants exhibited higher levels of shop-floor engagement, better quality audits, and in most cases higher levels of plant labor productivity. Militant unionism took the job seriously.⁹Many managers understood this and didn’t buy into the post-1980 union-busting prescription. “[When] asked which way they would like to see their own companies’ collective bargaining go in the future—toward a return to traditional negotiation or toward forcing concessions from labor, ‘even to the point of eliminating the unions’. Among executives of companies that are heavily organized and accustomed to dealing with entrenched and sophisticated unions, there is little incentive for pressing current gains. By a margin of 66 percent to 23 percent, these [managers] are anxious to return to traditional bargaining practices …. As [pollster Louis] Harris comments, ‘These people don’t expect to get rid of unions and figure that they might as well work with them.’ . . . [Managers] in companies with relatively little unionization split almost equally on whether to return to an adversary relationship.’” See: “A Management Split over Labor Relations,” Business Week 14 (June 1982): 19. Many managers understood this and didn’t buy into the post-1980 union-busting prescription. “[When] asked which way they would like to see their own companies’ collective bargaining go in the future—toward a return to traditional negotiation or toward forcing concessions from labor, ‘even to the point of eliminating the unions’. Among executives of companies that are heavily organized and accustomed to dealing with entrenched and sophisticated unions, there is little incentive for pressing current gains. By a margin of 66 percent to 23 percent, these [managers] are anxious to return to traditional bargaining practices …. As [pollster Louis] Harris comments, ‘These people don’t expect to get rid of unions and figure that they might as well work with them.’ . . . [Managers] in companies with relatively little unionization split almost equally on whether to return to an adversary relationship.’” See: “A Management Split over Labor Relations,” Business Week 14 (June 1982): 19. His observations led to significant breakthroughs in how economists, industrial leaders, and academics view productivity performance.

The tug-of-war among plant managers, general superintendents, foremen, and workers—with management trying to get more out of workers than the contract stipulated, and workers likewise pushing back—had fascinating results. Management needed workers to produce at high levels, but because they were organized, the workers would fight to be compensated for doing so. Evidence for their success could be seen in the hourly wage rate differences throughout the plant, which resulted from the to-and-fro of what Zabala called the “effort-wage bargain.”

Zabala took these observations further. He discovered that companies and unions collected historical data on this effort-wage bargain behavior, though they paid little attention to it. He constructed one of the first time-series datasets of shop floor bargaining at General Motors, spanning from the mid-1930s to the early 1980s and including a broad array of metrics—grievance bargaining, extralegal strike bargaining (wild­cats, walkouts), absenteeism, quits, quality audits (at the work group, departmental, and plant level), accident rates, and on-the-job deaths. These data had been collected by plant management and the union, and could be augmented with engineering data. Zabala asked how, without using this information, would management know what to do? On what information would it base decisions about which plants to invest in, and which to close?

When Zabala’s plant-level research came to the attention of the U.S. Department of Labor, he was hired to continue this work as part of the Department’s productivity research program. Plant-level productivity research was a missing link in the Department’s microeconomic research program; General Motors and the UAW supported the effort. Within a few years, Zabala entered into a collaboration with John Norsworthy, who headed the Division of Productivity at the Bureau of Labor Statistics. Zabala and Norsworthy spent more than ten years exploring whether there might be an alternative labor policy approach that could yield larger and more reproducible cost reductions in the auto industry and, by extension, in U.S. manufacturing.¹⁰The articles detailing this ten-year journey include, in reverse chronological order: Craig A. Zabala, “Sabotage at a General Motors Plant,” in Autowork, eds. Robert Asher and Ronald Edsforth (Albany: State University of New York Press, 1995), 209–25; Alice C. Lam, J. R. Norsworthy, and Craig A. Zabala, “Labor Disputes and Productivity in Japan and the United States,” Studies in Income and Wealth, vol. 53 (Cambridge, Mass.: National Bureau of Economic Research, 1991): 411–35; J. R. Norsworthy and Craig A. Zabala, “Worker Attitudes and the Cost of Production: Hypothesis Tests in an Equilibrium Model,” Economic Inquiry 28, no. 1 (January 1990): 57–78; Craig A. Zabala, “Sabotage at General Motors’ Van Nuys Assembly Plant, 1975–1983,” Industrial Relations Journal 20, no. 1 (Spring 1989): 16–32; Craig A. Zabala, “Information Systems and Labor-Management Cooperation: Negotiating for Productivity Improvement and Cost Reduction,” U.S. Department of Labor Bureau of Labor-Management Relations and Cooperative Programs, 1990; J. R. Norsworthy and Craig A. Zabala, “Responding to the Productivity Crisis: A Plant-Level Approach to Labor Policy,” in Productivity Growth and U.S. Competitiveness, eds. William J. Baumol and Kenneth McLennan (Oxford: Oxford University Press, 1985), 103–18; J. R. Norsworthy and Craig A. Zabala “Effects of Worker Attitudes on Production Costs and the Value of Capital Input,” Economic Journal 95, no. 380 (December 1985): 992–1002; J. R. Norsworthy and Craig A. Zabala, “Worker Attitude, Worker Behavior, and Productivity in the U.S. Automobile Industry, 1959–1976,” Industrial and Labor Relations Review 38, no. 4 (July 1985): 544–57; J. R. Norsworthy and Craig A. Zabala, “A Note on Introducing a Measure of Worker Attitude in Cost Function Estimation,” Economic Letters 10, no. 5 (October 1982): 185–91. Underpinned by Zabala’s participant observation research, the authors sought to demon­strate empirically how worker behavior on the shop floor affected plant- and company-level economic performance as constrained by company-wide labor policies.

Norsworthy and Zabala’s model found quantitative links among worker performance, total factor productivity, and the cost of production. Their simulation results overturned a lot of widely held assumptions. Neoclassical economics had always assumed that unions are cost-increasing. Remarkably, this assumption had no measured basis. Prior to the work of Norsworthy and Zabala, the discipline had never measured the output- and capital-augmenting effects of worker behavior in a formal statistical model. The standard microeconomics “theory of the firm,” focused primarily on output measurement and cost accounting, postulated that workplaces and companies were simply a set of more or less identical institutions whose performance and profitability were controlled only by perfect competition, with new firms easily able to enter, and above-normal profits impossible to make. Capital, labor, energy, materials, and technology were assumed to be deployed in the fixed proportions of the production function. Getting past this gross oversimplification required the contribution of shop-floor studies.

Moreover, Zabala and Norsworthy found that the impact on productivity and the cost of varying the firm’s capital inputs wasn’t linear or invariant: it depended on how capital was applied, its cost in the credit markets, and on levels and types of labor-capital conflict. Conflict—measured, as already noted, by grievances, unauthorized strikes, and quits—moved the needle on productivity and cost by re­quiring negotiated improvements under enforceable local union con­tracts that addressed the issues over which workers and supervisors were at odds. The contract was a kind of living legal court: most of the conflicts signaled by grievances, walkouts, and so on could be “negotiat­ed to acceptable resolution, some had to be kicked up to higher levels of management, and a few required further concerted action to resolve. Every conflict and every ensuing negotiation was a learning opportunity. Over time, both sides accumulated a growing understanding of which changes advanced productivity and could reward those who contributed to its improvement.”

This shouldn’t be a paradox. To benefit from productivity gains, labor must be able to advocate for a return on its efforts. Without a seat at the negotiating table (typically provided by local unions), workers have no recourse but to disengage and withhold effort, with unfortunate results that make the chances of rebuilding American industry unlikely.

Productivity Bargaining and Labor’s Role in American Reindustrialization

Historically, total factor productivity has increased more slowly than labor productivity because businesses have replaced labor with capital and other inputs. While this replacement reduces labor costs, it raises the costs of other inputs, in some cases pushing total costs higher. Just as total factor productivity measures plant performance in physical terms, the unit cost of production measures plant performance in dollar terms.

Norsworthy and Zabala simulated the capital savings from a 10% improvement in worker attitudes. Since capital is a fixed amount, the value of capital is called a shadow value in their model and associated with constraint. Its value changes if constraints are intensified or relieved, reflecting the marginal value (or opportunity cost) of relaxing or tightening them. Most analysts consider the shadow cost of capital based on borrowing constraints to leverage production (e.g., working capital lines of credit) and compare them to the firm’s user costs. Norsworthy and Zabala introduce worker attitudes as a special type of technical change to see whether and, if so, how much those attitudes constrain capital.

They tested a synthetic 10% improvement in worker attitudes, measured as a 10% improvement in the number (and types) of grievances filed, the grievance settlement rate (fewer unresolved grievances, a better grievance bargaining and settlement environment in the plant), fewer unauthorized strikes, and fewer quits. Unauthorized strikes are typically due to a poor grievance bargaining environment (too many unsettled grievances) or a strike over health and safety. For example, Zabala’s plant had a wildcat strike after a worker was killed when he fell onto the conveyor belt due to unsafe machinery. Workers reacted spontaneously and shut the plant down. New safety policies were negotiated within days, and workers returned to work, but only after the production loss of thousands of vehicles, a terrible and unnecessary outcome of negative productivity bargaining over health and safety exacerbated by months of heavy overtime work.

The statistical tests found that the shadow values, or implied cost savings, of inputs in the cost function (capital, labor, energy, materials) change significantly with a 10% improvement in worker attitudes. Specif­ically, the shadow value of capital input grew from about $1 billion in 1959 to over $5 billion dollars in 1979; in May 2024 dollars, that $5 billion would have been more than $24 billion. Worker attitudes—and hence the payoff to addressing the sources of withheld effort—became more influential over the time period studied. Consider the implications of this experiment for the ability of big American automakers like GM and Ford to compete with Tesla, BYD, and other innovative low-cost producers. In 1978, roughly 9.5 million cars and light trucks were produced in U.S. factories. In today’s dollars, the Norsworthy-Zabala experiment’s $24 billion in capital saving would have meant that each could have been produced for $2,600 less had there been such a 10% response to expressions of labor voice. It also means that a company such as Ford, which in 2023 produced almost exactly two million vehicles in its American plants, could have earned, and therefore could have reinvested, as much as $5.2 billion more.

A simple reading of our argument might suggest that highly unionized automakers in the 1970s and 1980s should have been highly profitable. One big reason they were not, of course, is that market share losses to newly capable foreign competitors were sapping their economies of scale. But it was also because the focus of union bargaining was on establishing a national wage pattern rather than addressing stagnating productivity. Both the automakers and the UAW International in Detroit strongly opposed giving local plants more power to negotiate, fearing it might cost them control of national bargaining. And particularly after 1980, the UAW was laser-focused on minimizing national contract concessions, and its leadership was in no mood for local union militancy or problem solving. Nor was that leadership aware of the root causes of the employers’ productivity doldrums: not since the early 1960s had the UAW made any serious attempts to get the automakers to share their production or financial data. The result was a major lost opportunity to maximize factor inputs and slash costs to address the emerging competitive horizon.

The productivity bargaining needed to achieve the kind and level of results found in the Norsworthy-Zabala experiment would require the company to open the books at the plant level for transparency, accuracy, and accountability as new policies related to production are designed and implemented by both management and labor. Pattern bargaining, which takes place primarily at the national union and corporate levels, rather than at the plant level, cannot produce comparable results. Pro­ductivity bargaining draws attention to interfirm and interplant differ­ences in productivity gains and cost reductions, and improvements become the basis for sharing gains between capital and labor. Of course, the Norsworthy-Zabala results were a simulation based on a narrow set of worker voice indicators. Using a much broader set of such indicators, and with companies opening the books to show output and financial information, would likely lead to even larger cost savings being identified and shared.¹¹Absenteeism is an obvious candidate to add to the roster of indicators. The data collected by Zabala showed increasing worker and local union militancy throughout the period 1976–79, with significant increases in grievances filed and in grievances not yet settled by plant managers and union representatives. Intensifying shop floor bargaining represents increasing worker engagement. Absenteeism was also increasing after 1975, as GM overtime set postwar records; in Zabala’s plant, where he worked on the door line in the body shop, nine-hour workdays and two Saturdays a month were standard production schedules from 1976 to 1980. That absenteeism was to be expected, as missing work gave workers some control over their life and take-home pay was ample; Zabala referred to this as counter-planning on the shop floor.

As Milton Friedman reminded us, a fully functioning neoclassical model of the firm requires that all factor inputs engage in profit-maximizing behavior.¹²Milton Friedman, “On Slavery and Colonialism,” in Is Capitalism Humane? (Ithaca, N.Y.: Cornell University Press, 1978). In today’s environment, in which the working class has been demoralized and in which unions represent only 10% of the U.S. workforce, labor’s ability to play its necessary role is badly damaged. This has bequeathed to us a sclerotic economic system, which can only be rejuvenated by more labor-management conflict, the application of scientific rigor to estimating how responses to conflict can boost productivity (especially the productivity of capital), and a state that invests heavily and intelligently in rebuilding American industry.

Organized labor must get bigger, stronger, and above all smarter if it is not only to share in but also to contribute to the big increases in productivity required for large reductions in cost. While simply fighting for a larger share of a pie that soon may be shrunk is desirable, it is not sufficient. Unions, and the companies that employ their members, must be players in a comprehensive plan in which union members share in the gains from qualitatively higher productivity. Such a plan may well be the only way to reindustrialize America.

This essay is adapted from “The Conflict Paradox: Productivity Bargaining and Labor’s Role in Reindustrialization,” which originally appeared in American Affairs Volume VIII, Number 4 (Winter 2024): 80–105.