Business structures, employment relationships, job characteristics, and worker outcomes have changed—in some ways, unpredictably—over the last few decades. The drive to be more flexible and to lower costs is often what motivates businesses to rethink their employment relationship strategies. In some cases, new technologies have enabled new arrangements. It was already apparent to researchers by the 1980s that changes were generating a need to develop new measures to track work and workers.
In response, the U.S. Bureau of Labor Statistics (BLS) developed the Contingent Worker Supplement (CWS), first administered in 1995.1 At the time, one concern was that the growth in contingent and alternative work arrangements, driven by the flexibility and cost-savings goals of businesses, signaled less long-term commitment by companies to their workforce. It was feared that this in turn might lead to worse outcomes for workers (Polivka, 1996). Some evidence suggested that work was becoming more precarious—that there was something more unstable about the workplace than had been the case in the past (for example, see Osterman, 1999). However, how work arrangements are actually changing and what the implications of those changes are for workers are empirical questions that require high-quality data to answer.
1 The economist and labor historian Audrey Freeman, whose original work helped motivate the CWS, first used the term “contingent” in 1985 as shorthand for the shift toward temporary or conditional employment with little or no attachment between the employee and the employer (Polivka and Nardone, 1989).
To begin addressing these questions, the CWS sought to measure key aspects of a worker’s employment relationship. Specifically, it sought to find out (1) whether the worker’s job was temporary (or contingent) in nature, and hence less secure; and (2) whether the worker’s main job—defined by BLS as the job associated with the most hours worked—belonged to a selected set of alternative work arrangements (AWAs) that differed from traditional employment arrangements in important ways that seemed likely to matter to workers. Information on five categories of AWAs was collected in all six waves of the CWS: (1) employees of temporary help agencies, which act as intermediaries by contracting out workers on their payrolls to client organizations on a temporary basis; (2) employees of contract companies that, like temporary help agencies, contract out employees or their services to clients; (3) independent contractors, independent consultants, and freelance workers, who provide services for customers and are self-employed; (4) on-call workers, who must be available to work when called on; and (5) day laborers, who are selected by employers from among workers who congregate at particular spots to work for a day. In 2017, questions on work performed through mobile apps or online platforms, such as Mechanical Turk, Uber, and Lyft, were added to capture this new and rapidly growing type of work arrangement. The BLS does not label part-time work as an AWA, but instead measures part-time work through the questions asked on the monthly Current Population Survey (CPS) questionnaire.
Although much has changed in the 25 years since the first CWS, the broad measurement objectives as originally outlined are still relevant. Nevertheless, as this report argues, modifications to the survey are needed to meet today’s policy and research needs. This report takes the position that improved and expanded measurements of the types of work arrangements covered in the CWS are needed, although research and experience with the CWS indicate that in some cases a household survey is not the ideal vehicle for collecting the information.
The work arrangements of particular interest may be broadly categorized into two types: those in which employers do not hire their workers as employees and those where work schedules are highly variable and unpredictable. With respect to the former, an organization or customer may contract directly with workers, in which case the workers are considered self-employed. Self-employed workers have at times been treated as a black box in economic statistics. This belies the reality that much diversity exists within the self-employment category. Self-employed workers, for instance, include owners of large and medium-sized businesses that have employees and substantial capital investment, independent contractors or freelancers working for an organization or through an online platform, and independent contractors or informal workers providing personal services directly to consumers.
It also has become apparent that workers’ perceptions are far from uniform regarding their own work status. There are systematic differences between those who do and do not identify as self-employed even among those who are independent contractors.2 For example, people who work at large distribution centers and are paid a piece rate often are independent contractors but may not see themselves as self-employed.
While organizations may engage workers as independent contractors instead of hiring employees to perform the same tasks, organizations also may contract out that work to companies whose employees perform it either at the client’s worksite or off-site.3 Temporary-help agencies are one type of contract company that acts as a labor intermediary. Temporary-help employment grew rapidly in the 1990s, accounting for about 10 percent of net employment growth in the economy during the decade, although the share of all wage and salary employment that is in the temporary-help employment sector has stabilized since 2000 at about 2 percent.4
Outsourcing, whether through independent contractors or contract companies, has always been a part of how U.S. corporations operate. Nonetheless, evidence suggests that among some leading corporations in the United States it has recently become more prevalent. Reports indicate, for instance, that temporary workers, contract employees, and vendors account for more than half of Google’s workforce, while Amazon relies heavily on independent contractors to take orders and to process and deliver goods and on contract companies to staff its warehouses.5 Moreover, the advent of online platforms and mobile apps such as Uber, Lyft, and Mechanical Turk represents a new, technology-enabled business model whereby the platform company, like a temporary agency, mediates the employment relationship by connecting workers to clients and handles their payment. Unlike the typical arrangement at a temporary help agency, at least under current law in most states, workers in these arrangements usually are not treated as W-2
5 The LA Times (“UC outsources thousands of jobs to private contractors. Is that a good idea?” December 1, 2019) reports that these categories of workers make up over 50 percent of Google’s global workforce, while also noting that outsourcing is a trend observed not just in the private sector but in the public sector as well. For example, The University of California system, the state’s third-largest employer, “spends some $523 million a year on outside contracts for an estimated 10,000 parking attendants, security guards, custodians, cafeteria workers, groundskeepers and patient-care technicians” (https://www.latimes.com/business/story/2019-12-01/university-of-california-outsources-jobs).
employees of the platform company.6 Although jobs carried out through online platforms and mobile apps still represent a small share of total employment, their number has grown rapidly, and continued growth could greatly increase the number of people working as independent contractors.
Both of the above categories of workers—independent contractors and people working for contract companies—may be subject to unpredictable work schedules at higher rates than the average across the labor market. Even in traditional employer-employee jobs, however, the nature of employment arrangements has changed in many industries and occupations. For example, the development of scheduling algorithms has transformed work in retail, restaurants, and other services industries. While using such technology may enable companies to better match workers with demand, it also means workers may be on-call or otherwise receive little advance notice of their schedules from week to week, and variable hours may translate into variable earnings.
Claims about upheavals in the way people now work notwithstanding, research is mixed regarding the extent of change in employment relationships in the United States in recent years. Two measurement problems are particularly important: (1) trends in independent contractor relationships vary considerably as measured across different datasets; and (2) levels of self-employment are higher in tax data than in household survey data, which appears to be accounted for mainly by independent contracting (Abraham et al., 2020; Collins et al., 2019; Lim et al., 2019). Cases of workers holding multiple jobs complicate the measurement of workers in both instances.
The CPS Annual Social and Economic (ASEC) Supplement should be measuring essentially the same construct of self-employment as tax records in the Social Security Administration’s Detailed Earnings Record. As Abraham and colleagues (forthcoming) found, however, “there is a great deal of disagreement” between the two.7 For the period 1996–2015, the researchers found that 66.7 percent of those whose tax data in the Detailed Earnings Record showed self-employment income reported no self-employment income in the CPS and, conversely, 51.5 percent of respondents indicating self-employment income in the CPS reported no self-employment income in the Detailed Earnings Record. Overall for the period, estimates of average annual levels of self-employment are much higher when based
6Telles (2016) posits the following list of features that characterize online platform, or “digital matching firms”: (1) they use information technology to facilitate peer-to-peer transactions, (2) they rely on user-based rating systems for quality control, and (3) they offer workers flexibility in deciding their typical working hours and rely on workers to use their own tools and assets to provide a service.
7 These records are provided to the Census Bureau by the Social Security Administration.
on the Detailed Earnings Record (16.4 million) than when they are based on the CPS (11.3 million).
Surveys conducted by the Federal Reserve Board offer a different perspective on the labor market than the CWS because they use a more inclusive definition of AWAs, one that incorporates work other than that carried out on the main job. The Federal Reserve’s Survey of Household Economics and Decisionmaking8 and its Enterprising and Informal Work Activity Survey9 both reveal high shares of the workforce to be engaged in nontraditional work. The latter survey, for example, estimated that in 2015, 36 percent of workers did at least some freelance work. One of the justifications for expanding the CWS to cover secondary jobs is the high rates of independent contractor work found in that and similar surveys.
Findings based on administrative data or financial data have the potential to provide further complementary insights that may help in the design of household surveys. Tax data, for example, indicate that many people with wage and salary employment during the year also earn smaller amounts of money through self-employment (Collins et al., 2019). Personal financial accounts data indicate that participants in the online platform economy often are actively engaged in it for just a few months of the year but also that, between 2013 and 2018, transportation platforms have grown to dominate in both the number of participants and total transaction volume (Farrell, Greig, and Hamoudi, 2018).
The main point here is that different data sources uncover unique and sometimes contrasting portrayals of work activity in the United States. The contrasting results are not necessarily contradictory, however, as they are sometimes simply measuring different aspects of the changing work environment. Moreover, they can inform modifications to the CWS to better capture important aspects of the evolving nature of work.
Data collection should be driven by the research and policy questions that need to be answered, and because the questions change over time, data systems must be adapted to fill the information gaps that become exposed. Key policy measurement needs are identified and discussed in detail in Chapter 2, and solutions for improving measurement in the CWS are out-
8 The survey, its sixth iteration conducted annually since 2013, was last fielded from October 11 through November 12, 2018. Available: https://www.federalreserve.gov/publications/2019economic-well-being-of-us-households-in-2018-description-of-the-survey.htm.
lined in Chapter 3. Here, a few policy measurement needs are introduced in a preliminary way.
There is much interest among policy makers and researchers in addressing concerns about the future of work in the United States. These concerns are punctuated by the perceived fracturing of relationships between workers and employers, by the heightened importance of access to skills and education as the impacts of new technologies and automation are felt, and by the market-based pressure that companies face to produce short-term profits, sometimes at the expense of long-term value. Additionally, informal work, often done under the table, is not well captured in official statistics but is of interest for policy, since it may disproportionately be performed by the most vulnerable segments of the population—although data are needed to assess and quantify even this assertion.
Although problematic job characteristics such as insecure work hours and lack of access to benefits can be found in jobs across the labor market, various outsourcing and scheduling practices are of special interest. That is because evidence suggests their prevalence is high and growing, giving rise to new issues requiring attention from researchers and policy makers. Basic questions in need of answers include these: How many workers are in these arrangements? What is the impact on earnings of working in these arrangements? How many people engage in these work arrangements to supplement income from other employment? How do the compensation and benefits practices in these arrangements spill over and affect those practices in more traditional work relationships? In what industries are these workers engaged, and what is the demographic makeup of this worker population? And finally, what factors motivate people to pursue nontraditional work, including among people using AWAs to supplement income, and how do these motivations vary across income levels and other demographics?
Obtaining answers to the above questions requires more than measuring job categories. The real issues concern the nature and quality of modern jobs and how they are changing, whether people are on average worse off or better off in contract work arrangements than they are in traditional employer-based arrangements, and what characteristics of work most affect people’s lives in terms of economic security and, in turn, health, stress, and family life. It is these characteristics of work (e.g., access to social insurance, to employer-provided benefits, and to stable hours and earnings) and their links to outcomes that should drive measurement objectives, rather than the labels given to the arrangement. The goals of policy are to improve the economic security and well-being of workers, whether they are active in traditional or nontraditional jobs.
The policy attention currently demanded to address changing labor practices coincides with a critical time for our economy. With unprecedented economic disruption not seen since the Great Depression, it is
crucial that the BLS work to better understand AWAs in the United States. The onset and now deepening impact of COVID-19 has exposed how vulnerable our society can be when workers are participating in the labor market without an adequate and well-coordinated social safety net. This report’s recommendations are intended to improve policy makers’ ability to address labor market vulnerabilities to prepare for future economic downturns. Several of the more urgent policy issues are especially relevant for two categories of AWAs highlighted above: independent contracting and contract company work. The key distinction regarding the former (including most of those who work for online platforms or mobile apps) is that such workers, because they are self-employed, do not typically receive the protections afforded by employment and labor laws, are excluded from many social insurance programs, and are not eligible for employee benefits. These characteristics of work as an independent contractor give rise to key policy questions: Should some or all of the protections afforded to employees under existing laws be extended to independent contractors? What are the implications for rethinking the social compact in anticipation of future work structures? And, as mandated by the recent AB-5 legislation passed in California, should certain independent contractors, perhaps including those who find work through platforms such as Uber and Lyft, be classified as employees?10
Policy prescriptions may depend on our understanding of why workers take jobs characterized as AWAs. Do they take them by choice, or due to lack of choice? How often is independent contracting work a main job and how often is it done on a short-term basis to supplement income from some other primary source? Data collection on independent contracting is complicated by the difficulties survey respondents often are observed to have in accurately reporting whether they are contractors or not, an issue addressed in detail in Chapter 3.11
Similar policy issues arise for contract company workers. Outsourcing of certain tasks has always been a business practice and, in some circumstances, workers as well as firms benefit from these arrangements. In other
10 The enactment in California of Assembly Bill 5 (AB 5) tightens the definition of employment based on an “ABC” test, a guide for employers to determine if a worker should be considered an independent contractor or an employee. Its intended effect is to reduce the misclassification of employees as independent contractors and could affect the categorization of many workers—including ride-hailing drivers, construction workers, food-delivery couriers, nail salon workers, and franchise owners—in an effort to reduce insecurity associated with this kind of work. Similar legislation is currently being considered in a number of states including New York and New Jersey.
11 Lack of awareness of employment status can have important consequences for affected workers. With respect to unemployment insurance (UI) or workers’ compensation, for example, it may not be until people get laid off or injured on the job that they find out they are not employees and are not eligible for benefits.
circumstances, the literature shows that outsourcing may result in lower wages, reduced benefits, and compromised workplace safety.12 The joint employment status of contract company workers—wherein control and supervision of an employee’s activities are shared among two or more businesses—raises some significant questions: What are the obligations of a client firm to the workers it employs through an intermediary? Do these arrangements reap efficiencies? And, are there adverse consequences for workers in terms of wages, benefits, and job safety, and if so in what circumstances?13
Policy makers and researchers require information about worker preferences, such as why they work in various arrangements and what the implications of these arrangements are for their employment stability, wages, benefits, and other aspects of job quality. For example, having better data would be helpful to policy makers working on portable benefits plans, that is, plans that would be accessible to all workers regardless of work arrangement. Several states have introduced bills to create such plans and also to make existing programs, such as state paid leave programs and state auto-enrollment retirement accounts, more accessible to a wider range of workers. During the Obama administration, the U.S. Department of Labor identified independent contracting status and the misclassification of workers as major areas of policy concern (in a sense as a prelude to the state-level efforts described above).14 Another layer of worker well-being is affected when workers are pushed to seek secondary or even tertiary jobs to supplement income. Workers often are independent contractors or in some nonemployee arrangement in such secondary work activities.
In addition to raising the issues noted above, independent contractor and contract company work is sometimes (though not always) temporary and irregular in nature. “Contingency,” as measured by the CWS, is part of job insecurity; but other temporal aspects of work, such as flexible work scheduling and irregularity of hours, may also affect workers’ sense of security. High week-to-week variability in hours, or inadequate notice about the timing of hours available, occurs in many contract work arrangements, but also in standard W-2 type employment. As noted above, technology has enabled growth in the use of scheduling algorithms that attempt to closely
13 The U.S. Department of Labor has issued its interpretive rule (a form of regulatory guidance) under the Fair Labor Standards Act on determining joint employer status that provides a narrow definition of its application. Available: https://www.dol.gov/agencies/whd/flsa/2020-joint-employment/fact-sheet.
14 See Administrator’s Interpretation No. 2015-1 July 15, 2015, issued by the Department of Labor’s Wage and Hour Division, concerning the Application of the Fair Labor Standards Act’s “Suffer or Permit” Standard in the Identification of Employees Who Are Misclassified as Independent Contractors.
match worker hours to firm needs, but the resultant flexible or unpredictable scheduling also shifts income risk onto workers. Variability in hours, particularly if it is unpredictable, may not be job ending (in the sense that is built into a “temp job”) but it may nonetheless lead to economic insecurity if a worker cannot count on enough hours to generate an income capable of meeting basic needs (Henley and Lambert, 2014).
Debate over the need for new worker protections, such as minimum advance-scheduling requirements, has led to legislation being passed in some cities to help offset negative impacts.15 To inform such efforts, key job characteristics need to be monitored through regular data collection, including variability in timing and number of hours worked and any associated volatility in earnings.
Along with policies that recognize and address the challenges created by emerging AWAs, policies are also needed that recognize and nurture the positives created by innovative new employment models. For some workers, schedule flexibility may increase productivity and wages. For participants in online platform work, attractive features may include the way the platform allows them to choose when they work, learn entrepreneurship skills, or transition back into work after extended absences. These flexible arrangements may be especially beneficial to people with otherwise limited labor market options, such as students, retirees, parents of small children, or those who can only find part-time work.
From the perspective of businesses, companies may use independent workers not just to reduce costs, but also to tap into talent and skills pools with the agility required to maintain competitiveness. The new economy requires policies designed to make new employment models work well not just for workers, but also for the employers who engage them and for the markets in which they operate. University of California officials have argued (see footnote 5) that, in addition to saving money as a means of curbing further tuition hikes, contractors give them the flexibility required to meet complex hiring needs. One policy goal is to protect the lower-paid, lower-skilled workers who are vulnerable to abuse as part of a “race to the bottom” while simultaneously enabling on-shore economic growth for the higher-paid, higher-skilled jobs in which workers tend to have greater choice. Despite contrary positions on different sides of these debates, it remains an open question whether the platform model—with the wider use of contracting and flexibility for workers that arises from it—is incompatible with providing workplace protections like minimum wage or overtime.
15 The following site tracks states and localities that have adopted “predictive scheduling” requirements with the idea of helping workers better plan their schedules and budgets: https://www.hrdive.com/news/a-running-list-of-states-and-localities-with-predictive-scheduling-mandates/540835.
For all the reasons stated above, policy makers are highly engaged in topics related to AWAs and, more broadly, in understanding how jobs are changing and what the future of work holds. A report issued by the Senate Committee on Appropriations on June 28, 2018, notes that “the Committee is pleased BLS is reporting on the contingent workforce during the current fiscal year. The Committee directs BLS to continue capturing data on contingent work and alternative work arrangements by conducting the Contingent Worker Supplement to the Current Population Survey on a biennial basis.”16
At the same time, Congress and other policy makers will move forward on legislation and advocacy in some capacity with or without proper data-based evidence to replace the assumptions currently in vogue. This policy climate makes it all the more urgent to improve the data infrastructure for studying how employment relationships are changing and the implications for workers and firms. Reliable data are particularly important for policy makers as they attempt to develop a consensus on how to move forward in a way that is helpful to the population engaged in this kind of work. More is known now than several years ago, but there are still many unanswered questions and much confusion among policy makers and the general public about how work arrangements are changing. To improve the knowledge base, a range of sources will need to be tapped, including those from government, academia, and the private sector.
The type of data needed to address these policy concerns does not rely on formal, established categories of employment, because often the categories themselves are shifting or the terminology describing them is changing. Instead, what is most relevant are the characteristics of the jobs held by American workers, such as their level of earnings, whether they are working on an employee or nonemployee basis, whether the job is temporary, whether their work schedules are regular and predictable, and whether the jobs provide health insurance, retirement benefits, or time off. These are the characteristics most relevant to understanding and ultimately trying to improve the outcomes and well-being of workers engaged in AWAs. For example, having a count of the number of jobs categorized as “on-call” according to some specific definition is less important than knowing how many jobs have an unpredictable work schedule. Similarly, having a count of the number who consider themselves independent contractors is less useful than knowing how many are hired by an organization or client on a nonemployee basis, regardless of the term used to describe the arrangement.
The CWS is designed, in part, to measure the temporary nature (contingency) of jobs.17 BLS defines contingent workers as those “who do not have an implicit or explicit contract for ongoing employment.”18 To date, the CWS has been conducted in 1995, 1997, 1999, 2001, 2005, and 2017. As described above, the creation of the CWS was motivated by concerns during the 1980s about the changing nature of employment and the implications for work policies. Today’s concerns about the changing nature of employment are therefore not new. The characteristics of alternative work have evolved since the 1980s, however, and some of the issues of concern also have changed since the CWS was conceived. Although the panel recognizes the value of maintaining consistency across surveys and believes that the broad measurement objectives of the original CWS are still appropriate, the survey requires updating to continue to be relevant for policy and research purposes. For example, as alluded to above, the modern labor market dictates that more attention be given to measuring the irregularity and unpredictability in workers’ hours.
Of course, no single survey or survey supplement can measure all aspects of the changing nature of work. As noted above, a central focus of the CWS has been on measuring jobs that are temporary, or contingent, along with measuring work arrangements whose characteristics differ from those of standard employee jobs in ways that likely would matter to workers. The latter cluster includes the work of independent contractors, contract company workers, on-call workers, day laborers, temporary help agency workers, and, in the 2017 survey, electronically mediated work (that is, work obtained through online platforms or mobile apps that mediates the payment from the customer to the worker). For the most part, the categories of work arrangements are defined as mutually exclusive, although there is a small overlap between on-call and contract company work in all CWS waves, and in 2017 information on electronically mediated work was collected for all jobs and could be associated with any arrangement on workers’ main jobs.
When the BLS received funding to field a new round of the CWS in 2017, a primary objective was to assess how the number of workers in
17 The CPS, of which the CWS is a part, is a monthly survey of about 60,000 households that provides data on employment and unemployment in the United States. Special supplements to the CPS have been used as an efficient way to collect additional targeted data. The statistics produced from the supplements, and from the CPS in general, are often considered to be the gold standard. Currently, supplemental questions are asked on a wide range of topics, including veterans’ employment, displaced workers, and students’ employment.
alternative employment arrangements had changed since 2005, the previous time the supplement was fielded. Another objective was to “measure an emerging type of work—electronically mediated work, defined as short jobs or tasks that workers find through websites or mobile apps that both connect them with customers and arrange payment for the tasks” (BLS, 2018). Four questions, listed in Box 1-1, were added to the 2017 CWS for this latter purpose.19
As currently constructed, the CWS is well positioned to help researchers answer questions about the number of workers in temporary jobs and the AWAs covered by the survey, various characteristics of those jobs, workers’ job tenure, and workers’ reasons for being in a particular work arrangement. In considering how the CWS could be reshaped to better measure additional aspects of AWAs, in the chapters that follow the panel addresses a number of questions, which are previewed next.
Within the broad measurement objectives of the CWS, are the work arrangements or characteristics of work covered in the survey the most rel
19 Full documentation of the CWS survey, including the development and results for the four questions about independent work done through online platforms, is provided in Current Population Survey Staff (2018).
evant ones, or should some new questions be added and others be dropped to make space?
In other words, what is it that policy makers and researchers most need to know about work arrangements and work characteristics to better understand the modern labor market?
Information about work categories and job characteristics is needed from the statistical system, although it is possible that some can be better captured through administrative data or surveys other than the CWS. As argued above, the changing nature of employment arrangements has created a measurement need to collect more comprehensive information about specific job characteristics faced by workers than is currently done in the CWS. A clear example is unpredictable work schedules. Evidence from other surveys and research (presented in Chapter 3) indicates a high prevalence of schedule unpredictability, giving rise to a highly salient policy issue. Collecting this information in a large nationally representative survey would help to inform regulatory policy decisions.
Many issues of interest to policy makers and researchers hinge on the distinction between employees (W-2 workers) and the self-employed, which includes independent contractors. Thus, it is important to accurately classify workers as employees or self-employed/independent contractors or other nonemployees. For many questions, however, it is necessary to capture the characteristics of jobs rather than capturing only a type of work arrangement (e.g., working on-call rather than identifying someone as an “on-call worker”).
The current CWS collects information only on the main job. Should this be expanded to include secondary work activities?
This question relates to the appropriate scope of work to be covered in the CWS. Throughout this report, the panel considers the merits (and pitfalls) of collecting information on more than one job or work activity. Several studies suggest that independent contractor or informal work arrangements, including work for online platforms, are often secondary work activities that are important to household income (Abraham and Houseman, 2019; Farrell, Greig, and Hamoudi, 2018; Robles and McGee, 2016). More fully understanding the various sources of income for households is important for research and has potentially important implications for whether and how policy makers should respond to changing work arrangements.
The current CWS collects information on work that has occurred only during a very recent time period. Should this reference period be expanded?
This question also relates to the appropriate scope of work to be covered in the CWS, which asks respondents who identified as employed
“last week”: “Did you do ANY work for pay (either pay or profit)?” This report’s recommendations address the benefits of asking about work for a different reference period—for example, to capture work done by participants only sporadically—as well as the potential negatives, which include considerations about the ability of respondents to report accurately about activities that took place further in the past.
Is the CWS, a household survey, the best vehicle for collecting the various pieces of information needed to inform policies related to work arrangements? And if not, what are alternative data sources?
Because space for questions in the CWS is limited, it is always worth assessing whether resources could be shifted to maximize its value for the purpose of measuring AWAs. In some cases, topics could be deemphasized not because they are unimportant but because other data sources have a comparative advantage over household surveys in their collection. An example, alluded to above, is the case of temporary agency and contract company workers. Workers often may not be fully informed about the complicated nature of the business relationships underlying their “employment.” For example, an individual being paid by a staffing firm and assigned to a distribution center of a major retailer may report being employed by the retailer. The difficulty household survey respondents have in accurately reporting such intermediated work arrangements is well documented, and information about these arrangements may be better collected through other means.
For data that should continue to be collected in the CWS, does the current survey instrument fully and accurately capture the desired information or are there ways the data could be improved?
In Chapter 3 of this report, the panel identifies situations for which cognitive testing and altered question wording could improve the accuracy of information collected in the CWS, including for questions about temporary, independent contractor, and online platform or mobile app work.
Recognizing resource limitations, how should the information collected in the CWS be prioritized?
Although BLS has decided to reshape the CWS to accommodate the measurement needs created by new labor market dynamics, the agency also recognizes that there is value to fielding a supplement over time that maintains a fixed set of core questions, so that valid time-series comparisons can be made. Almost immediately after BLS released estimates from the May 2017 CWS, it received feedback from users (academics, business leaders, politicians, and other government researchers) who agreed that the capacity to compare estimates from the May 2017 and earlier iterations of
the CWS was helpful. Users also expressed the view, however, that there were important questions that were still not possible to answer and that needed to be addressed.
Much of the current survey instrument is devoted to measuring the temporariness of the respondent’s job in order to produce statistics on contingency. While contingency is an important job characteristic, the panel recommends (in Chapter 3) simplifying this set of questions to make room for asking about other aspects of the employment relationship. To allow room for higher-priority questions, the panel also recommends that BLS consider dropping questions that have been little used by researchers or the policy community or that might be better measured using other sources.
The Value of Multiple Data Sources
Going forward, the data strategy for measuring AWAs will not be compartmentalized in one survey or even necessarily one statistical agency. The Foundations for Evidence-Based Policymaking Act of 2018 urges statistical agencies to seek opportunities to innovatively combine data sources with the goal of improving measurement.20 In this spirit, while the first part of the measurement strategy is to improve the BLS surveys, a second part, described in Chapter 4, involves exploiting and improving coordination among complementary data sources.
Reflecting the potential of a multiple-source data approach, economic researchers have used surveys, tax reports to the Internal Revenue Service, data from the Social Security Administration, and transactions information from commercial banking accounts to measure different aspects of AWAs and, in particular, web platform work activity and income. Although new measurement challenges have arisen, especially with the use of “organically” generated commercial data—for example, issues regarding their representativeness and their transparency21—there are also distinct advantages. Commercial data, for example, offer detailed, high-frequency information based on transactions posted daily. This data characteristic can allow
20 The law requires, among other things, that statistical agencies become more transparent with their data and share their datasets internally and with other government agencies for research purposes to the maximum extent possible under the law. Available: https://www.congress.gov/bill/115th-congress/house-bill/4174/text.
21 Unlike survey data, whose properties are well understood as the result of decades of methodological research, use of administrative and commercial data is relatively recent. But approaches are being developed to assess the quality of new types of data. Japec et al. (2015), for example, outlines a “Total Data Error” framework which includes traditional methods (which parse potential sources of bias and error into sampling and nonsampling errors) but expands the sources of nonsampling error to include measures of error capturing how commercial or organic data are generated, extracted, transformed, loaded, and ultimately analyzed.
researchers to identify even work that is performed sporadically. Elsewhere, insights gleaned from research using tax data have been generated from records capturing payments by firms to unincorporated individuals for nonemployee services (e.g., Collins et al. ). For example, tax data indicate that most 1099 online platform work supplements primary W-2 jobs.
During the first meeting of the panel, held March 29, 2019, BLS leadership outlined their goals for the study and their motivation for commissioning it. Presentations by BLS experts described the history, measurement objectives, and past performance of the CWS. Strengths and weaknesses of household surveys, and the CPS in particular, for the purpose of measuring contingent and alternative work arrangements were identified. Time was also allocated for a discussion among panel members and sponsor representatives to sharpen the project statement of task. Box 1-2 lays out the scope of the study, which was refined collaboratively during the first meeting of the panel.
In addressing that charge, this panel report is organized into three chapters (beyond this introductory chapter). Chapter 2 assesses the measurement needs for monitoring the changing employment landscape and broader economy and for informing policy designed to mitigate negative effects while preserving the benefits from these changes. The chapter stresses the need to measure all work-based sources of income, not necessarily just income from “primary” jobs. The chapter also emphasizes the importance of monitoring key job characteristics, which often reveal more about effects on worker outcomes and well-being than does a sorting of categories of alternative work.
Chapter 3 addresses the role of the CWS within the spectrum of measurement needs. There are many potential sources of information about changing work arrangements. Some sources, like the CPS/CWS, are government household surveys, but other kinds of surveys, as well as administrative and commercial data, can play complementary roles. Within this context, the comparative strengths and weaknesses of the CWS are assessed so that an optimal role for the survey can be considered. Here, recommendations about the conceptual and content scope of the CWS are presented, as well as guidance regarding CWS questionnaire design.
Chapter 4 describes in detail the role of complementary (non-CWS) survey and nonsurvey data sources in measuring AWAs and the characteristics of workers in these arrangements. Findings emerging from these other sources—which include non-BLS household surveys, firm/establishment surveys, government administrative data sources, and commercial data sources—are reviewed, and lessons for the CWS are drawn. The promise
of combining multiple data sources to leverage the strengths of each source is also explored.
An appendix to the volume provides a summary of an open meeting of the panel, which included presentations highlighting the perspectives of data users and policy makers concerned with the issues created by changing work arrangements in the modern economy.
This page intentionally left blank.