Takeaways from the Technology-Enabled Disruption Conference
For many who were adults in the late 20th century, the widespread adoption of the internet and related technologies may have seemed to represent the apogee of the disruptive effects of technological change, reshaping industries such as retail, music, travel reservations and news media. But for those who are in a position today to observe the potential for disruption from technologies such as machine learning and robotics, the web appears to be only a curtain-raiser on an era of even more dramatic disruption of business models and labor markets.
The implications of technology-enabled disruption for the economy—for firms, workers, governments and monetary policymakers—were the focus of the second joint conference of the Federal Reserve Banks of Atlanta, Dallas and Richmond, held at the Dallas Fed on May 22–23, 2019. Highlights of the sessions are below; selected videos and slide presentations from the sessions, as well as video interviews with insights from some of the presenters, are available on the conference page. The inaugural conference was held in 2018.
In his welcoming remarks, Mark Wynne, vice president, associate director of research and director of the Globalization Institute of the Dallas Fed, noted that technology has been disrupting economic life for as long as societies have engaged in organized economic activity. The First Industrial Revolution—also known as the British Industrial Revolution—brought technological innovations to textile manufacturing and other industries, leading to widespread displacement of workers. The Second Industrial Revolution in Great Britain and the United States in the late 19th and early 20th century likewise brought major changes. Technology-enabled disruption on a large scale isn’t new.
Today, Wynne observed, many are talking of a “Third Industrial Revolution,” driven by rapid innovation in information technology, with enormous changes to services, products and business models. At the same time, a variety of economic phenomena are presenting challenges to our understanding of the economy: slowing productivity growth, rising concentration in some industries and increasing polarization of opportunities in the job market, among others. Wynne said that the pace of innovation in information technology is an important part of the explanation for these changes in the broader economy, and he expressed hope that the conference would promote research into that connection.
Raphael Bostic, president and CEO of the Atlanta Fed, recalled that the conference series originated in conversations between him and Robert Kaplan, president and CEO of the Dallas Fed, in which they discovered a shared interest in the effects of current technological changes on microeconomic and macroeconomic relationships. He noted the addition of the Richmond Fed as a partner for this year’s conference.
Robert Kaplan said technology-enabled disruption means that workers are increasingly being replaced by technology. It also means that existing business models are being supplanted by new models, often technology-enabled, that bring more efficiency to the sale or distribution of goods and services. As part of this phenomenon, consumers are increasingly able to use technology to shop for goods and services at lower prices with greater convenience—which has the impact of reducing the pricing power of businesses. This reduced pricing power, in turn, causes businesses to further intensify their focus on creating greater operational efficiencies. These trends appear to be accelerating. It is likely that disruption is becoming a greater factor in the economic outcomes of workers. Increasingly, workers with lower levels of educational attainment are seeing their jobs restructured or eliminated. Unless they have sufficient math and literacy skills, or are retrained, these workers may see their productivity and incomes decline as a result of disruption. Thus, disruption may help explain the muted wage gains and overall labor productivity growth we have seen in the U.S. as well as in other advanced economies during much of the recovery from the global financial crisis.
Panelists in this session, moderated by Robert Kaplan of the Dallas Fed, presented their perspectives as senior business leaders from several major industries: oil, retail and railroads.
Greg Armstrong, chairman and CEO of Plains All American Pipeline LP, noted that not all disruption comes from information technology. The combination of horizontal drilling and hydraulic fracturing (commonly called “fracking”) has been a highly positive disruption for the U.S. economy and national security. Armstrong, who is also chair of the Board of Directors of the Dallas Fed, noted that although these new technologies increase workers’ productivity, the increase in volume has led to increased demand for labor and intensified competition for workers. As a result, Plains All American has started working with area colleges to train workers for its operations.
Craig Boyan, president of H.E. Butt Grocery Co., said that the grocery industry is normally not thought of as technologically advanced, but technology has brought significant changes to the industry in recent years. These have included the addition of new sales channels, such as online ordering with in-store pickup, local grocery delivery and shipments of groceries from the warehouse to customers’ homes (usually out-of-state customers who miss Texas food). Technology has brought new marketing channels and the shifting of marketing resources into digital marketing. It has brought business process improvements with big data machine learning. His firm maintains extensive data on customers’ shopping, which is helpful in tailoring inventory to target markets and in managing inventories and supply chains of highly perishable goods. He noted that his company is also testing the use of autonomous vehicles for local, last-mile delivery.
Brant Ring, vice president, business unit operations of BNSF Railway Co., said that one contribution of technological disruption to the freight railroad business has been improvements to safety. Machine learning and image analytics help the company anticipate and prevent failures. Automation at intermodal container handling terminals has increased as well. Over the course of the next 10 to 15 years, he expects the number of employees in the railroad industry to decline, as it has since the 1970s. But the pace of that decline will largely follow that of natural attrition. New employees joining the company, regardless of their roles, will be ones comfortable in the digital space.
This session, chaired by David Altig, executive vice president and director of research at the Atlanta Fed, presented academic perspectives on the potential of technology to disrupt some aspects of academia itself and the nature of work.
Richard Baraniuk, the Victor E. Cameron professor of electrical and computer engineering at Rice University, discussed OpenStax, a Rice initiative of which he is the founder and director. OpenStax is creating free, open-source e-textbooks in response to textbook price inflation. Baraniuk said textbooks now represent more than 40 percent of the cost of attending community college, creating access barriers. Although instructors historically haven’t given great weight to textbook prices when choosing the ones to assign, many became concerned beginning in the 2007–09 recession as they learned that their students were forgoing textbooks to save money. For several subjects, the OpenStax initiative uses data from students’ interactions with digital versions of the texts, in combination with machine learning, to generate feedback for the students about what to study.
Joel Mokyr, the Robert H. Strotz professor of arts and sciences and professor of economics and history at Northwestern University, addressed the relationship of technology and the future of work. He noted the concern of many that if technological progress continues at its current pace, the result will be widespread technological unemployment and a dystopian society. He said that such concerns are an old story; they have also arisen during earlier periods of highly disruptive technological change, leading to resistance to the new technology—the most famous example being the Luddites of early 19th-century England. Those fears proved unfounded as the displaced workers of the Industrial Revolution found work in new industries and occupations. Prominent economists warned in the 1960s of widespread unemployment from automation, and those predictions, too, proved unfounded. While such transitions were not painless in the past and will not be painless in the future, the newer jobs tend to be of higher quality. Today, moreover, labor supply is declining as a result of an aging population and other factors. To the extent that “idleness” may grow nonetheless, this is a lesser concern today as the options for leisure are increasing and improving.
Willy Shih, the Robert and Jane Cizik professor of management practice in business administration at Harvard Business School, considered how automation is affecting labor markets at different skill levels. He argued that automation is enabling middle-skill workers to take on traditionally higher-skill tasks and low-skill workers to take on traditionally middle-skill tasks. As an example, he presented a suit factory in which automation was enabling low-skill workers to do the work of high-end tailors. Such developments, by lowering prices and increasing the quality or complexity of the product, can raise demand for the goods. That, in turn, may partly offset the effect of the labor-saving automation on employment. In view of the likelihood of some displacement of workers, Shih also argued for rethinking the assumption that older workers can’t be retrained and for rethinking the approach of front-loading investment in human capital almost entirely within the first 25 years of a worker’s life.
Agustín Carstens, general manager of the Bank for International Settlements, was introduced by Roberto Coronado, senior vice president in charge and senior economist at the El Paso Branch of the Dallas Fed, and interviewed by Robert Kaplan of the Dallas Fed. Carstens was governor of the Bank of Mexico from 2010 to 2017 and Mexico’s minister of finance and public credit from 2006 to 2009. He argued that with the global economy having been stabilized, it is time to focus on policies that will increase sustainable growth, including in the areas of labor market reform and competition. He expressed concern about the ripple effects of trade disputes on investment and growth.
Participants in this session, chaired by Sylvain Leduc, executive vice president and director of research at the San Francisco Fed, presented recent research on incentives for innovation, on artificial intelligence and the future of the workforce and on participation in the sharing economy.
Zorina Khan, professor of economics at Bowdoin College, shared her historical research on innovation in Britain, France and the United States. She compared two approaches to encouraging innovation: administered systems, such as prizes, in which decisions about rewards are made by administrators or panels, and market-oriented systems such as patents. She found that the market-oriented patent system outperformed administered systems, which induced rent-seeking and the misallocation of resources. She also showed that major innovations such as the telephone led to spikes in litigation until legal doctrines were able to adapt to the new technology and reduce its negative consequences.
Susan Lund, a partner at McKinsey & Company and a leader of the McKinsey Global Institute (MGI), presented research conducted by MGI on progress in AI and robotics and how this progress is likely to affect current jobs. This research concluded that less than 10 percent of jobs can be entirely automated (that is, fully replaced by automation). Examples of these include assembly line workers and sewing machine operators. At the other end of the spectrum, there are also few jobs with no activities that can be automated. In the aggregate, across the economy, around 40 percent of work activities have the potential to be automated using currently demonstrated technologies within current business models. The analysis indicated that the shift will favor workers with technology skills or social-emotional skills.
Erica Deadman, vice president and consumer research lead of the JPMorgan Chase Institute, shared research conducted using JPMorgan Chase account data to estimate participation in the “online platform economy,” a term encompassing a variety of platforms for the selling of labor or services or for leasing. Many of the platforms are part of what is commonly called the sharing economy or gig economy. The platforms fall into four sectors: transportation (such as ride-sharing services), nontransportation work (other types of labor), independent leasing (such as home rentals) and independent selling of goods. The research yielded information on trends in participation by families in the various sectors. Some 4.5 percent of families engaged in platform work in 2018; platform revenue represented a major source of income for families in the months during which they participated, but they commonly participated only a few months out of the year.
This session, moderated by Raphael Bostic of the Atlanta Fed, offered the perspectives of leaders in higher education on the roles of their institutions in preparing a new workforce.
William Serrata, president of El Paso Community College, noted that more than 88 percent of his institution’s enrollment is in academic transfer coursework for students preparing to transfer. He said that while there is nothing wrong with transferring to four-year institutions, he considers that percentage out of step with the labor market. He suggested that at least 30 percent of enrollment should be in career technical education. He described developments in career technical education programs at his institution, including in automotive technology, fashion design, allied health, welding and machining and information technology. He said it is important for college matriculation to happen directly after high school; of those students who wait, only 1 percent will eventually enroll and earn a credential.
Joe May, chancellor of the Dallas City Community College District, said that only 35 percent of high school graduates nationally from the lowest income quartile earn any postsecondary credential. Within high schools in the Dallas area, the figure is only 27 percent—yet 65 percent of jobs created in North Texas require a postsecondary credential. His institution has learned that many high school students, especially minorities, are deterred from enrolling in college by the inflexibility of college in relation to the work hours of the jobs that they need to earn money. He described the Dallas Promise Network, a network of high schools, regional universities, employers, nonprofits and success coaches in place to increase college matriculation by low-income students. He also shared efforts to link information systems across institutions so that high schools, community colleges and four-year institutions can access the combined records of enrolled students, an effort that will reduce barriers to access by simplifying the students’ transition from one institution to another.
Carine Feyten, president and chancellor of Texas Woman’s University, suggested that both displacement by technology and the need of workers to obtain new skills to move up in their careers are creating a greater role for adult education. She proposed an “extended warranty” or “subscription” model of higher education in which graduates could come back for further training as they need it. She also suggested that employers should become more deliberate in creating career paths for Hispanic employees so that they do not remain stuck in lower-level jobs. She described the value of educational institutions such as hers that primarily serve women, particularly in encouraging women students to take risks and move into leadership roles.
In this session, chaired by Marc Giannoni, senior vice president and director of research of the Dallas Fed, the panelists presented research on labor markets and how they may be affected by the adoption of disruptive technologies.
Henry Siu, professor of economics at the University of British Columbia, described how changes in skill demand—arising in part from technological change—have affected different categories of occupations and how these effects have fallen differently on male and female workers. It is useful to sort occupations along two dimensions: as cognitive vs. manual and as routine vs. nonroutine. Routine occupations, both routine-manual and routine-cognitive, have been shrinking as a share of overall U.S. employment since the early 1980s and that decline has been accelerating. Examples of routine-manual include machine operators and auto mechanics; examples of routine-cognitive include secretaries, bank tellers and travel agents. He reported that women in routine occupations are becoming more educated and moving into better jobs while men in routine occupations increasingly are not working.
Ekkehard Ernst, chief macroeconomist of the International Labour Organization, showed evidence that advances in digital technology since the mid-1980s have been associated with increases in inequality rather than increases in unemployment. He said four underlying forces related to technology are influencing the increase in inequality: the rise in demand for higher-skill workers, the hollowing out of middle-tier jobs (job polarization), winner-take-all markets that concentrate rewards on a small number of companies and, most recently, the ability of data-driven tools such as AI to engage in more exact price discrimination.
Katharine Abraham, professor of economics and professor of survey methodology at the University of Maryland, presented research on the effect of new technologies on the structure of work—specifically, a firm’s decision whether to hire or to contract out. She noted that firms contract out for a number of reasons, including greater ability to adjust wages, avoidance of some legal responsibilities that would be owed to an employee, easier adding or shedding of capacity in response to fluctuations in demand and taking advantage of specialized expertise. Online platforms that reduce the transaction cost of contracting out may lead firms to favor it more often. Overall, it is ambiguous whether contracting out is better or worse for workers; most contract workers appear to prefer it. She shared research in progress on one occupation, taxi drivers, indicating that the entry of Uber into a local labor market increases the rate at which incumbent drivers exit and lowers the earnings of those who stay. From 2010 to 2016, ground transportation (mostly taxi-type services) has seen huge growth in self-employed drivers, coinciding with the period when ride-sharing services were introduced. The rise of technologies that aid in contracting out creates issues for policymakers, as labor market laws and regulations in a range of areas—including safety, social insurance and discrimination—are designed for scenarios in which most workers have an employer.
Craig Hall, chairman and founder of HALL Group and serial entrepreneur, was introduced by Mine Yücel, senior vice president and senior research advisor at the Dallas Fed, and interviewed by Mark Wynne of the Dallas Fed. Interests of the HALL Group include real estate development, management and finance; hotels; winemaking; and early-stage startup investment. He described his early experiences in entrepreneurship, his observations of how technology is disrupting winemaking, his view of technology as a boon to today’s entrepreneur by offering access to information and to lower-cost services, the need for banks to have greater incentives to lend to smaller, younger companies, and America’s difficult regulatory environment for starting businesses compared with those of many other countries.
Patrick Bajari, vice president and chief economist of Amazon and professor of economics at the University of Washington, was introduced by Pia Orrenius, vice president and senior economist at the Dallas Fed. Bajari shared two internal Amazon research projects based on machine learning. The first project used machine learning to seek to improve quality/hedonic adjustments in price indexes. It fed Amazon product images and the text of product descriptions and reviews for clothing into a machine learning system to create a model of changes over time in clothing and combined this information with transaction prices. The second project investigated the extent to which large datasets are needed to build highly accurate machine learning models. The research found that increasing the number of observations per product is valuable. Increasing the number of products is helpful when the number is relatively small (fewer than 5,000 products or so); beyond that, the benefits are negligible as the number of products increases. He said this finding is inconsistent with the view that accumulating large datasets automatically improves a company’s models.
This session, chaired by Kartik Athreya, executive vice president and director of research at the Richmond Fed, offered a case study in the optimization of the results presented by an online search platform, research on whether the contributions of new general-purpose technologies to productivity have been mismeasured and research on the effects of information technology on economic output and on different occupation categories.
Will Wang, research economist at Microsoft Research, shared his work on rank optimization in online search. He described approaches to understanding users’ search behavior and methods of ranking search results to maximize consumer surplus. He then showed the application of these methods to a production application at Microsoft, the Microsoft Partner Center, a search tool that matches Microsoft partners among themselves for the exchange of goods and services.
Chad Syverson, the Eli B. and Harriet B. Williams professor of economics in the Booth School of Business at the University of Chicago, presented research on a possible factor behind the phenomenon of missing productivity improvements—that is, the absence from productivity data of major gains that might be expected from computer technologies. The research considers whether the missing gains might be partly the result of mismeasurement of intangible capital related to the technologies. Syverson and his co-authors addressed this question by estimating the intangible capital of firms (taking advantage of the relationship between a publicly traded firm’s market value and its tangible capital, which is known, and its intangible capital, which is unknown). They found that the mismeasurement of productivity appears to be small and does not explain the post-2004 slowdown in measured productivity growth.
Nir Jaimovich, professor of economics at the University of Zurich, elaborated on the research he conducted with Henry Siu and other co-authors and which Siu presented on the first day of the conference. After showing data highlighting the trend of job polarization, he addressed the question of what has happened in the labor market to workers with “routine” characteristics. The research treated the classification of workers into routine vs. nonroutine as a machine learning classification problem. This analysis indicated that in 1989, two-thirds of workers with routine characteristics were in routine occupations—but by 2015, that share had fallen to 56 percent, a decline of 11 percentage points. The analysis further pointed to where those 11 of 100 people had gone: four of the 11 moved to nonroutine manual jobs, and the remaining seven dropped out of the labor force. These and other statistics underscored the distributional consequences of the shifts in skill demand resulting from automation.
In this session, chaired by Alfreda Norman, senior vice president at the Dallas Fed, panelists discussed what communities are doing and could be doing to respond to technology’s disruption of labor markets.
Stuart Andreason, director of the Center for Workforce and Economic Opportunity at the Atlanta Fed, addressed the need for governments to consider the mechanics of investment in adult training. He highlighted the 2018 book Investing in America’s Workforce: Improving Outcomes for Workers and Employers, which sets out research, best practices and resources in this area; it is published by the Federal Reserve System with several partner institutions.
Rob Grunewald, economist at the Minneapolis Fed, described research from Harvard University and other institutions on the role of nurture during the first three to four years of a child’s life, particularly its role in the formation of synaptic connections that shape the child’s ability to succeed later in life. He also discussed long-term studies on the effects of high-quality early childhood education programs for disadvantaged children combined with parent education and support. Estimated annual rates of return to such programs range from 7 percent to 20 percent.
Sandy Dochen, manager of corporate citizenship at IBM, said that IBM has found that many jobs at the company in technology areas are well-suited for people with a two-year associate’s degree, jobs that IBM calls “new collar.” He then described an IBM educational model known as P-TECH, initially implemented in Crown Heights, Brooklyn, involving collaboration among high schools, community colleges and businesses to prepare students for technology jobs or continuation to a four-year institution.
Mark Flanagan, director of IT, innovation and solutions for the Dallas County Promise at the Commit Partnership, described the Dallas County Promise, which was also addressed by Joe May on the first day of the conference. He noted that the Dallas County Promise is founded on tuition-free college, success coaching and career mentoring and the development of technological infrastructure to wrap around siloed systems at different levels of institutions. He reported that the first cohort of Dallas County Promise students has had a 6 percent increase in college enrollment and a 12 percent increase in retention from the fall to the spring semesters.
At this session, four Reserve Bank presidents—Tom Barkin of the Richmond Fed, Raphael Bostic of the Atlanta Fed, Mary Daly of the San Francisco Fed, and Robert Kaplan of the Dallas Fed—engaged in a lively discussion of the upsides and downsides of technological disruption for the economy and how technological change is affecting the work of the Fed. The roundtable was moderated by Rana Foroohar, global business columnist and associate editor at the Financial Times.
For More Information
Please contact Sharon Wallace at firstname.lastname@example.org.
Session I: The Disruption Challenge Facing Business
Session V: Implications of Disruption for Labor Markets
Session VIII: Policymaker Roundtable