


         Contingent Workers:
Data Analysis Limitations and Strategies


Michael Foley, M.A.
Washington State Department of Labor and Industries
John Ruser, Ph.D.
U.S. Bureau of Labor Statistics

Glenn Shor, Ph.D.
California Department of Industrial Relations
Harry Shuford, Ph.D.
National Council on Compensation Insurance

Eric Sygnatur, M.A.
U.S. Bureau of Labor Statistics


                                November, 2012



White Paper prepared for the Workshop on the Use of Workers' Compensation Data for Occupational Safety and Health, June 19-20, 2012
Introduction
The growth of the contingent workforce,  and those working on alternative work arrangement, including independent contractors, temporary leasing, day laborers and temporary agency employment, has become the subject of a burgeoning literature in public policy.  There has been a debate as to how rapidly these work arrangements are growing, the sources of this change and whether this change represents a fundamental transformation of the relationship between labor and management. The rise of contingent forms of labor contracting (also known as precarious or non-standard work) has been seen as one aspect of a larger process of "disarticulation" of organizations where some activities previously done in-house by permanent employees of the firm are contracted out and purchased through the market. In labor markets this takes the form of a "core-periphery" pattern of human resource management in which a core of permanently-attached full-time employees is supplemented by a periphery of part-timers, temps and contract workers [Belous, 1989; Gordon, 1996].  The existence of the periphery allows businesses to react flexibly to changing market conditions while maintaining a valuable long-term relationship with its core workers in whom the company has invested much in training. For the client firms, peripheral labor provides a source of "just-in-time" workers. Pensions, health insurance and other fringe benefits typically are not provided to many workers in the periphery, particularly independent contractors, and the firm only pays for their actual time on the job [Mangum, Mayall and Nelson, 1985]. Costs of recruitment, screening and training are shifted to the company supplying the workers, who handle the payment of wages, taxes, unemployment and workers' compensation insurance [Golden and Applebaum, 1992].  
Why should contingent workers and those on alternative work arrangements be a special focus for worker safety and health surveillance, particularly with workers' compensation data?  Short tenure at a particular workplace has been linked to an increased risk for injury. Some factors that may contribute to this include unfamiliarity with new work practices and surroundings, limited safety training, and disproportionately younger workers in this category.  In addition contingent workers may face an increased inability to refuse hazardous work or demand appropriate protective equipment for fear of dismissal. Employers may hire temporary workers as a means of shielding permanent workers from risky tasks, and they may invest less time in providing them with appropriate training and protection equipment. Temporary workers hired through an agency have two separate parties who are responsible for their safety, which raises the possibility that neither will take full responsibility to prepare the worker adequately.  Various occupational injury and illness data systems, including workers' compensation and those built on OSHA recording, may capture these workers differently, creating ambiguities interpreting data and compounding difficulties with occupational safety and health surveillance and epidemiological studies for this group of workers.
Contingent work and alternative employment arrangements
The U.S. Bureau of Labor Statistics (BLS) distinguishes between two, separate but related categories of workers that may be in scope of this paper: contingent workers and those on alternative employment relationships.  Contingent workers are defined as those who do not have an explicit or implicit contract for long-term employment and who hold jobs that are expected to last only a limited period of time.  The BLS identifies four categories of alternative employment arrangements, which may or may not involve contingent work: 1) independent contractors, consultants and free-lance workers, who may be either self-employed or wage and salary workers; 2) on-call workers who are called to work only as needed; 3) temporary help agency workers who are paid by a temporary help agency, regardless of whether the worker considers the job to be temporary; and 4) workers provided by contract firms, who are working for a contract company, usually work for only one customer, and usually work at the customer's worksite.
The BLS obtains estimates of the number of contingent workers and alternative employment arrangements in periodic supplements to the Current Population Survey (CPS), the most recent of which covered February 2005.  The CPS data are collected from households and the estimates that are generated come from the responses to a series of structured questions.  The CPS scope includes self-employed workers, as well as wage and salary workers.  
According to the February 2005 CPS supplement, contingent workers accounted for 1.8 to 4.1 percent of total employment, depending on how expansive was the concept of contingency.  These percentages were little changed from February 2001.  The supplement also identified that there were 10.3 million independent contractors (7.4 percent of total employment), of whom 87 percent were identified as self-employed in the main CPS survey.  The CPS supplement also identified 2.5 million on-call workers (1.8 percent of total employment), 1.2 million temporary help agency workers (0.9 percent of total employment) and 813,000 workers provided by contract firms (0.6 percent of total employment).
Another source of information on employment under some alternative work arrangements is data from establishments.  Unlike the CPS, these data include only wage and salary workers and exclude the self-employed.  Employment under alternative arrangements is identified according to the industry classification of the employing establishment.  In particular, North American Industrial Classification System (NAICS) subsector 561, Administrative and Support Services, includes establishments engaged in activities that support the day-to-day operations of other organizations.  Many of the activities performed in this subsector are ongoing routine support functions that all businesses and organizations must do and that they have traditionally done for themselves. Recent trends, however, are to contract or purchase such services from businesses that specialize in such activities and can, therefore, provide the services more efficiently.  NAICS subsector 561 includes a wide variety of industries engaged in supporting other organizations, including facilities support services (janitorial, maintenance, trash disposal, guard and security, mail routing, reception, laundry, and related services), temporary help services (establishments primarily engaged in supplying workers to clients businesses for limited periods of time to supplement the workforce of the client) and Professional Employer Organizations (engaged in providing human resources and human resource management services to client businesses).
Employment in Administrative and Support Services has been growing and is projected to continue growing.  While 7.044 million workers were employed in this industry in 2010, employment in the industry is projected to grow to 8.454 million in 2020 or 21 percent over the decade.  Further, within this sector, Employment Services (NAICS 5613), which comprises employment placement agencies, temporary help services, and professional employer organizations, is projected to add 631,300 jobs over the decade, at an annual rate of increase of 2.1 percent.  Employment in this industry will reach 3.3 million by 2020, placing this industry among those with the largest projected employment growth from 2010 to 2020.
Literature
The question arises whether the increasing share in the labor force of temporary or flexible-contract workers leads to deterioration in health and safety outcomes. Outside of the United States, evidence from previous research has varied from no difference in health outcomes [Virtanen et al., 2001, Virtanen et al., 2003; Bardasi and Francesconi, 2004], to significant findings of higher rates and severity of occupational injuries [Saha et al., 2004; Benavides et al., 2006]. Studies have also shown differences in risk between temporary and standard employment to be based more on the industry than on the type of employment arrangement [Saloniemi et al., 2004]. One challenge is the inconsistent classification across countries of what constitutes a contingent worker. In Europe, contingent or flexible employment includes not only agency-mediated temporary work, but also apprenticeships and direct-hire fixed-term contract workers. Flexible employment, under this definition, accounted for about 18% of paid employment in the European Union in 2000, with agency temporary workers accounting for about 2% of employed workers [Paoli and Merllie, 2001]. 
Previous research on the question of whether the rise of temporary or contingent work increases the risk of worker injury has been focused largely on discrepancies in health outcomes rather than on the underlying mechanisms which lead to the differential. Studies have found that temporary workers had higher odds of muscular pain [Benavides et al., 2000]; that in a manufacturing setting, temporary workers had injury rates two to three times higher than permanent workers [Morris, 1999]; and that temporary workers had four to seven times the claim frequency compared to permanent workers [Park and Butler, 2001]. In the 2000 European Survey on Working Conditions, temporary agency workers reported greater exposure to physical hazards and a higher level of work intensity and pace than permanent workers [Paoli and Merllie, 2001].  However, new hires generally might encounter comparable issues.
 Most studies have not controlled for differences between temporary and permanent workers in their industrial distribution. Data from the 2005 CPS Supplement show that temporary help agency workers are disproportionately concentrated in the manufacturing and professional and business services industries, with relatively low shares in retail trade, education and health services, and leisure and hospitality [BLS, 2005].  Foley [1998], using a large cohort of Washington State workers' compensation claimants, showed that claim frequency and severity as measured by time loss were higher for temporary workers than for permanent workers even after controlling for occupation and industry. Furthermore, this study found that the excess risk increases with the underlying hazard level of the industry. There were similar findings when the analysis was restricted to claims resulting in more than 4 lost workdays. Smith et al [2009] confirmed these results, finding workers' compensation claims rate ratios twofold higher than permanent workers in construction and manufacturing. Neither study, however, controlled for the effect of differences in tenure between permanent and temporary workers. Smith et al went on to look at differences in type of injury and found temporary workers had especially high excess injury rates for "struck by" and "caught in" type injuries in construction (IRR 4.93; 95% CI 2.80-8.08) and manufacturing (IRR 4.05; 95% CI 3.25-5.00). In addition, this study found disparities between temporary help workers and standard employment workers in the adjudication of claims, proportion with employer protests, and delays before receipt of time-loss benefits. Finally, since the late 1990s, the temporary help services industry in Washington State has had a known high-risk for upper-extremity musculoskeletal disorders [Silverstein et al., 1998, 2002]. These studies found temporary help services workers assigned to assembly occupations and machine operator jobs had among the highest incidence rates of all industrial groupings for shoulder disorders as well as back disorders. 
Even after controlling for occupation or industry, it is important to control for differences between temporary workers and standard employment workers in other variables which may be associated with increased injury. First among these would be job tenure.  At this point, we do not know whether tenure would account for a significant amount of the differences between temporary employees and their permanent counterparts.  Evidence suggests individuals with shorter job tenure are at higher risk for injury or illness [BLS 1997; Breslin, 2006]. Reasons for this association may include unfamiliarity with physical processes and environment, safety procedures and resources [Mayhew and Quinlan, 2002]. The 1995 CPS Supplement found length of service in a given assignment was much shorter for temporary workers than for permanent workers [BLS, 1996]. Given the much higher percentage of temporary workers who are at the lower range of job tenure, it is important to separate the independent contribution of job tenure to injury rate from that of employment arrangement. A recent study on the impact of temporary status on the probability of work-related sickness absence, which did not control for industry, did not find any association after controlling for job tenure [Tompa et al., 2008]. The authors speculate that the disincentive to take sick leave when one's job tenure is short may be playing a role. Alternatively it may be that tenure is the most important factor in explaining the observed differences.  Much the same reasoning applies to the need to control for the age of the worker. The 2005 CPS Supplement found 19% of temporary workers were under the age of 25, and 30% were 25 to 34 years of age, as opposed to 14% and 22% of permanent workers in these two age groups. As young workers experience higher rates of workplace injury and illness than workers aged 25 to 44, it will be important to control for this factor [BLS, 2011].  Thus, it is important for future research to attempt to tease out the differential effects of tenure and age in determining risk factors.
In contrast to studies focusing on health outcomes, relatively few studies have examined directly the antecedent factors leading to the discrepant outcomes between temporary and permanent workers.   Among these factors may be: To what extent is this difference the result of temporary workers' relative youth as distinct from their brief job tenure? Are temporary workers given the more hazardous jobs in a given worksite?  Do they know what to do if they are exposed to hazards? Do they feel unable to refuse unsafe work? What kind of safety training do temporary workers receive at the worksite compared to permanent workers?  Do temporary workers underreport injuries more than permanent workers? 
One study focused on such factors as lack of supervision and training provided to subcontracted employees at a large petrochemical plant which sustained a multiple-fatality explosion in 1989 [Kochan, 1991]. In another study, a sociologist worked as a temporary worker for a year and reported on his experiences, including the assignment of temporary workers to the more hazardous tasks in a given workplace [Parker, 1994]. But these detailed studies of particular incidents and experiences need to be supplemented by a more comprehensive examination which quantifies the relative contribution of the various potential risk factors to the discrepancy that has been observed in health outcomes.  For virtually all of these potential exposures it would be interesting to determine whether the experience of temporary workers differs from that of new workers generally.
Based on this overview of the literature, what research questions regarding contingent workers and those on alternative work arrangements do we seek to answer with workers' compensation data?  Workers' compensation data can shed additional light on a variety of questions such as the following.  Are contingent workers and those on alternative work arrangements at higher risk of workplace injury and illness than are workers in non-contingent and traditional work relationships?  What accounts for any higher risk observed?  Can the differences be traced to differences in industry, occupation and worker characteristics, such as gender and age? 
Coverage under workers' compensation insurance of contingent workers and those under alternative work arrangements
The usefulness of workers' compensation data for identifying and addressing workplace safety and health issues of contingent workers and those on alternative work arrangements depends on whether and how these data capture the injuries and illnesses of this group.  Two separate issues limit the usefulness of workers' compensation data for OSH surveillance of certain groups of contingent workers: where workers are not covered by workers' compensation, or where they are covered by workers' compensation policies of an entity distinct from where the employees physically work.
Under most states' laws, a person's eligibility for workers' compensation benefits depends upon them being an employee.  If a person is found to be an "independent contractor", they are typically not entitled to such benefits and so their injuries would not be captured in WC data.  For workers, the lack of "employee status" also may affect their coverage under the Occupational Safety and Health Act and under unemployment insurance, and may also impact their ability to establish eligibility for health insurance, social security benefits, Medicare, pensions, state disability insurance, Family and Medical Leave, and various other programs.  Thus, there may be savings in benefit costs to an employer in classifying a worker as an independent contractor as opposed to an employee.  These potential savings may create incentives to misclassify wage and salary workers as independent contractors, which DOL and State worker misclassification efforts are designed to address.  In the absence of needing access to safety net benefits, some workers may be convinced that independent contractor status is in their own benefit, especially if gross pay seems higher.  This seeming advantage can easily evaporate if there is an injury on the job.  On the other hand, employers who successfully establish that a worker was not an employee may also lose the exclusive remedy protections of workers' compensation law, and thereby be subject to tort liability if that worker was injured due to the negligence of the contracting employer.  
In contrast to independent contractors, temporary help workers are typically covered by workers' compensation insurance.  But, the temporary help agency, as opposed to the client company utilizing the services of the temp workers, provides workers' compensation coverage.  Key surveillance information may be lacking in the workers' compensation data about the client company and it may not be possible to track with workers' compensation data the injury and illness experience of workers according to the industry where they work.
Special issues arise in the workers' compensation coverage of employees of Professional Employer Organizations.  A Professional Employer Organization (PEO) is a firm that provides a service under which an employer can outsource employee management tasks such as employee benefits, payroll and workers' compensation, recruiting, risk/safety management, and training and development.  It does this by hiring a client company's employees, thus becoming their employer of record for tax purposes and insurance purposes. This practice is known as co-employment. As of 2010, there were more than 700 PEOs operating in the United States, covering 2-3 million workers.  PEOs operate in all fifty U.S. states. (Source National Association of PEOs - http://www.napeo.org/peoindustry/industryfacts.cfm).  PEOs are sometimes referred to as staff leasing entities.  Accreditation is currently under a self-regulatory body, the Employer Services Assurance Corporation.  
When a PEO obtains WC insurance for its client companies, it does so in one of two ways: a Multiple Coordinated Policy (MCP) or Master Policy, often depending on state insurance regulatory procedures. In a Master Policy arrangement, a single policy is issued in the name of the PEO, which provides coverage for all of the PEO's leased workers for each client. Each client is typically added to the policy by "endorsement", that is, the policy is amended with names of all employers added.  In an MCP arrangement, each client of the PEO has its own policy covering the leased workers. Each client's premium is based on its own class codes, rates, payroll and rating programs. All policies are assigned to the same insurance carrier whenever possible, and endorsements are used to coordinate coverage between the client and the PEO.
The goal of the MCP model is to keep the experience and reporting of each employer's payroll, premium and loss information unique to itself. The different models that are in effect differ slightly from state to state. The emphasis is on how the information is tracked and the interaction by the employers to secure the coverage in a particular state.  In multiple coordinated policies, data reporting can be more uniform and if an employer were to leave the PEO it is easier to reconstruct its experience rating factors.   For injury surveillance and resource allocation purposes, then, the MCP models that allows data attachment to a work location is a more useful source of information. (See further discussion on PEOs and proof of coverage requirements, below.) 
Recording on OSHA logs and the BLS Survey of Occupational Injuries and Illnesses (SOII)
The disparate definitions of "contingent worker" represent one challenge to measuring working conditions for this cohort, but there are structural limitations to existing safety and health surveillance systems that prevent those systems from producing reliable data on this group.  The BLS Survey of Occupational Injuries and Illnesses (SOII) and the OSHA recordkeeping logs on which it is based are not equipped to delineate injuries or illnesses incurred by contingent workers from other types of employees, and preliminary studies into employer recordkeeping practices indicate that such employees may be additionally underrepresented in the data for various administrative reasons.  Workers' compensation data have the potential to fill this void.
The SOII is the largest occupational health surveillance system in the U.S., with close to 250,000 establishments sampled annually.  Information gathered for the SOII comes directly from employers selected to report information from the OSHA 300 and 301 forms, on which the employer is required to keep a log of recordable injuries and illnesses throughout the year.
OSHA-recordable cases include injuries and illnesses that typically require medical treatment beyond first aid, as well as a number of other specific conditions that include loss of consciousness, hearing loss, and sharps injuries.  With specific regard to many employees considered part of the contingent workforce, OSHA recordkeeping requirements state:
"[Employers] must record on the OSHA 300 Log the recordable injuries and illnesses of all employees on your payroll, whether they are labor, executive, hourly, salary, part-time, seasonal, or migrant workers ... [as well as] employees who are not on your payroll if you supervise these employees on a day-to-day basis."
Thus, temporary, leased, and contracted employees are reported in the SOII based typically on where they are working, rather than for the parent firm (leasing agency, for example) for which they work.  This contrasts with many workers' compensation requirements, which mandate that leased employees be maintained on the parent company's compensation policy.  This reporting requirement and lack of other identifying variables for contingent workers makes it impossible for the SOII to disaggregate them and estimate the numbers, rates, and comparative health of these workers.  The differences in reporting likely also create confusion, and results in undercounts. The SOII does sample NAICS 56132 (Temporary Help Services) and 56133 (Professional Employer Organizations), but the firms in these industries are recording injuries and illnesses mostly for the executive, managerial, and administrative staff that run these firms rather than the employees they lease.  The numbers and rates are predictably small, as indicated in Table 1.
Table 1. Occupational Injuries and Illnesses, All US and selected industries, Private Industry, 2010
                                       
                                   All U.S.
                                   All U.S.
                                  NAICS 56132
                                  NAICS 56132
                                  NAICS 56133
                                  NAICS 56133
                                       
                                 Number (000s)
                               Rate per 100 FTE
                                 Number (000s)
                               Rate per 100 FTE
                                 Number (000s)
                               Rate per 100 FTE
Total recordable cases
                                    3063.4
                                      3.5
                                      8.0
                                      1.9
                                      3.6
                                      2.6
Cases with days away, job transfer or restriction
                                    1598.1
                                      1.8
                                      3.8
                                      0.9
                                      1.7
                                      1.3
Cases with days away from work
                                     933.2
                                      1.1
                                      2.4
                                      0.6
                                      1.2
                                      0.9
Cases with only days of job transfer or restriction
                                     664.9
                                      0.8
                                      1.4
                                      0.3
                                      0.5
                                      0.3
Other Recordable Cases
                                    1465.2
                                      1.7
                                      4.2
                                      1.0
                                      1.9
                                      1.4
Source: BLS Survey of Occupational Injuries and Illnesses, 2010

Underrepresentation of contingent workers in SOII
A number of studies allege that the SOII undercounts occupational injuries and illnesses that it should collect by as much as two-thirds.  These studies match SOII sample data to workers' compensation census data for a particular state by narrowing the scope of the latter system to match the SOII, obtaining matching rates, and calculating a capture-recapture estimate from those figures (Rosenman et al., 2006, Boden and Ozonoff, 2008).  A Government Accountability Office report in 2009 also found that both workers and employers have incentives to avoid recording injuries and illnesses on OSHA logs; to whatever degree such incentives suppress reporting to OSHA, the SOII would similarly be affected.  [US GAO, 2009]  There is some evidence to suggest that underreporting for contingent workers may be particularly abundant.
Undercounting and underreporting have numerous definitions and interpretations amongst the various studies, but it is meant here as the failure to record an injury or illness on the OSHA log or SOII forms that meets the definition of recordability.  Undercounting can be a mistake or, at its worst, an intentional omission.  Underrepresentation differs in that the injuries and illnesses to workers are missed in the SOII due to limits on the scope of the survey rather than errors or omissions.  Contingent workers in the SOII are likely both underrepresented and undercounted.  A number of contingent workers, for example, can be designated, legally or illegally, as self-employed (independent) contractors or seasonal workers employed at small agricultural firms.  These two groups are excluded from the SOII by statute  -  they are not part of the sample and are therefore underrepresented.
The BLS has sponsored some research that suggests contingent workers are also undercounted in the SOII.  Beginning in 2009, the BLS began funding employer interview studies to better understand the issues and practices of both OSHA and workers' compensation recordkeeping.  The BLS pilot-tested an interview instrument among 26 establishments in the Washington, DC metropolitan area, followed by over 50 employer interviews in Kentucky.  The Washington SHARP program has interviewed over 100 establishments in the State of Washington.  Many questions were open-ended and covered a wide variety of recordkeeping and administrative issues, and the interviews typically lasted about an hour.  Final results of these studies became available in the Fall of 2012, and they were qualitative in nature, so drawing broad-based conclusions from these observations requires caution.  Despite this, the preliminary evidence exhibits recordkeeping issues affecting the SOII.
There are also issues about the training of those responsible for collecting and reporting occupational injury and illness data.  Generalizing across all three interview studies, about half of all respondents who were responsible for completing the OSHA log at their establishment had 5 or more years of experience in their current position.  Depending on the State, between 45% and 73% of the respondents interviewed had formal OSHA recordkeeping training.  On the surface, the respondents appear to be experienced at keeping OSHA logs.  Another generalization is that more than half of all establishments used temporary workers, including nearly three-quarters of the establishments interviewed in the DC area (Phipps and Moore, 2010), so the practice of using contingent workers is well-represented in the sample chosen for interview.
A majority of the establishments interviewed either did not record temporary help workers on their OSHA logs or did not know if that was the appropriate thing to do.  Many in the Washington, DC study "assumed that they were reported on the staffing company log" or whoever provided the workers' compensation insurance (Phipps and Moore, 2010).  Similar to temporary help employees, uncertainties regarding recordkeeping requirements for contract employees were also evident in the DC study.  
An additional confounding factor that may contribute to underreporting but is not addressed in the employer interviews involves injured contingent employees that may be quickly replaced with other contingent workers at the work site.  The injured employee may then return to work after recuperation and be assigned to another work site altogether, completely separated from the original establishment.  In this instance, OSHA expects a "good faith recordkeeping determination" by the first establishment that includes reasonable efforts to acquire information about the injured worker.  
It appears at first glance that when the rules governing workers' compensation and OSHA recordkeeping intersect and differ, the decisions made by employers may have an adverse effect on OSHA recordkeeping.  What remains uncertain is whether workers compensation records are similarly affected by different rules between the systems.
Using multiple data sources to address undercounting in SOII
The finding of potential undercounting in SOII based on comparisons with other data systems has led some to argue that the gold standard for producing estimates of the total burden of workplace injuries and illnesses is a multiple data source system. Indeed, BLS has implemented such an approach in collecting workplace fatal injury data. The BLS Census of Fatal Occupational Injuries (CFOI) compiles a count of all fatal work injuries occurring in the U.S. during the calendar year. The CFOI program uses diverse state, federal, and independent data sources to identify, verify, and describe fatal work injuries. The use of multiple data sources assures counts are as complete and accurate as possible.  In 2010, there were 4,690 fatal work injuries. 18,774 unique source documents (or 4 documents per fatality) were reviewed as part of the data collection process that generated this count.
In comparison to the roughly 4,700 workplace fatal injuries counted by CFOI in all sectors of the economy, SOII estimated that in 2010 there were nearly 3.9 million OSHA-recordable nonfatal workplace injuries and illnesses in private industry and State and local government. The vastly greater number of nonfatal injuries and illnesses suggests that it would be quite costly to implement a multiple data source system across all States for all nonfatal occupational injuries and illnesses, and the variations in data systems between States creates practical barriers for a uniform approach nationally.  BLS has partnered with 3 states (California, Massachusetts, and Washington) to conduct multiple data source enumerations of amputations and carpal tunnel syndrome, in order to assess the cost and feasibility of using multiple data sources for enumeration of a limited set of injury and illness cases.  Workers' compensation data are one of the sources being used in the investigation.  Publication of research results from these studies is forthcoming.
The Potential Contributions of Workers Compensation Data to Understanding Workplace Safety and Contingent Workers
As indicated above there are concerns that the sources of data for the SOII may be ill-suited for capturing the experience of contingent workers.  In part this reflects the range of possible categories: leased workers (PEOs), temporary workers (short-term and longer-term), independent contractors serving as contract workers and consultants, and self-employed workers.  For some of these categories there is a sense that the lack of clarity as to who is the actual employer may result in incomplete data records or reporting.  Moreover, it seems likely that even if, for example, OSHA records are comprehensive, a survey of establishments may fail to capture a robust picture of the range of contingent employment categories that in total appears to account for only 2% to 4% of employment spread over numerous industries and occupations, or, if independent contractors are included, as much as 11% of employment.

Contingent Workers and Workers Compensation Data

Workers compensation data may help address some of the data limitations in traditional sources of information on workplace safety.  There are a variety of data from the workers' compensation system that might be used, including first and subsequent reports of injury, proof of coverage, and unit statistical reports.  The following indicates some of the challenges in coordinating these various data to address issues related to contingent workers and particularly with respect to PEOs.

All states except Texas require employers to provide proof that they have the financial resources deemed necessary to fulfill their obligations under state workers compensation laws.  In the case of PEOs, for example, an increasing number of states have explicit regulations related to mandatory reporting of workers compensation insurance coverage for both the PEOs and their client companies.  Thus the dual employer relationship should be captured in the Proof of Coverage (POC) data reported and maintained by state workers compensation regulators, typically part of the state labor department.

POC data often provide a way to identify PEOs and their client co-employers; in some states it is also possible to identify the workers compensation insurance policies issued to temporary employment agencies.  Data submitted separately by workers compensation insurers to designated statistical agents for rate making purposes (that is, the development of the premiums that insurers charge their insureds) are intended to provide detailed information on claim rates and the severity of injuries for individual employers.  There are three major challenges to using these POC and claims data to track workplace safety for contingent workers.  First, the data are not comprehensive; reporting requirements vary among states and not all categories of contingent workers are covered by workers compensation.  Second, the two key data sets are part of different regulatory processes  -  POC is part of employer regulation (typically in a division of the state labor department) while claims data are part of insurance regulation (typically in the insurance division of the department of banking and financial regulation or a separate department of insurance regulation).  Third, the statistical agents and rating bureaus that collect and analyze the claims data are largely prohibited by contract from sharing the employer specific data with outside entities.

As noted above, some states require multiple coordinated policies (MCP) and others accept master policies.  In some states master policies are acceptable in the voluntary market but MCPs are required in the residual market.  However, the interest of a state WC Division (usually part of the state labor department) is to be able to confirm that individual employers have workers compensation insurance (unless they are already a qualified self insured).  A master policy lists each client as an additional named insured; this list typically is used to verify that an individual employer is covered by a WC policy; the POC submission would show a single policy, the name of the insurance company, the name of the PEO and the names of each client employer.  In contrast under the multiple coordinated policy approach each client/employer is issued its own, separate policy; the POC submission would again list the insurance company, the client employer and the PEO.  These can also be used to verify that an individual employer is covered by a WC policy.  POC administration is relatively straightforward for MCPs  -  notification of a cancelled policy can be easily linked to the employer.  However, POC administration is a greater challenge with master policies  -  special reporting is required if an employer is dropped from the PEO's insurance policy. 
 
None of this is linked to the reporting of claims.  This is part of insurance regulation. Claim experience is reported to a state's designated statistical agent in what are called "unit reports."  These typically contain detail on every claim associated with a given WC insurance policy.  The units, however, typically do not indicate whether a PEO is involved.  This requires linking to the policy data that is generated for POC reporting purposes.  Master policies make it difficult if not impossible to link claims experience with specific employers; with MCPs the employer is readily linked to the claim experience.  A hybrid is starting to appear  -  the PEO coordinated policy  -  as with the MCP a separate policy is issued for each client/employer but the PEO is listed as the primary insured and the client is an additional named insured.  This makes tracking of the claims experience of individual clients a little more difficult because it is more difficult in most systems to link policies based on additional named insureds over time.

To further complicate the issue, in some states, the reporting function actually goes both to the statistical reporting agent for ratemaking purposes, and to the state agency overseeing workers' compensation administration for assistance in tracking and understanding the body of cases and the scope of dispute resolution services.  Some states are currently collecting significant information on cases (first reports of injury, subsequent reports or injury, and medical detail data) through a standard process being coordinated through the International Association of Industrial Accidents Boards and Commissions, which may improve the usefulness of these data for the study of contingent workers. 

Overcoming the Obstacles to Effectively Using Workers Compensation Data

In the four states with monopolistic workers compensation funds (Washington, Ohio, Wyoming, and North Dakota) it should be reasonably straightforward to overcome the regulatory barriers separating POC and claims data.  Indeed, this is already being done in the State of Washington.  There the state insurance fund is part of the Department of Labor and Insurance (LNI); SHARP, the research arm of LNI has direct access to the data.  Recent research by SHARP on the experience of temporary workers is mentioned in this paper.

Workers' compensation rating bureaus, that is, the entities that develop premiums from claims data, are also interested in understanding the implications of contingent workers for the workers compensation insurance system.  Accurate reporting of exposure (i.e. payroll) and claims is essential for the integrity of workers compensation rates.  For example, there is concern that during the recent housing bubble some construction firms tried to reduce their workers compensation costs by classifying many of their regular workers as independent contractors. As a result of growing concerns (and complaints of reputable business owners) the Florida legislature made underreporting of payroll a criminal offense.  New York did the same in 2007.  

More generally, state insurance regulators and rating bureaus recognize that components essential to maintaining premium rates that are not "unfairly discriminatory" (i.e. class rates and experience rating) may not be reported appropriately or accurately by employers of contingent workers.  (see the appendix for a description of the types and intended uses of ratemaking data.) The National Council on Compensation Insurance (NCCI), the rating bureau for 35 states accounting for roughly half of the countrywide workers compensation market, has been working since the early 1990s to support state regulators with both POC and claims reporting issues.  NCCI recently began an investigation to determine whether it is possible to link the POC data with detailed claims data for the states where it has access to both.

To the extent that it is possible to link POC and claims data (or at least identify claims experience of contingent workers as in the case of SHARP) it may be possible within the contractual limitations on the use of their data for some rating bureaus to conduct analysis tailored to address specific workplace safety issues related to contingent workers.

Other issues with respect to contingent workers impinging on the use of WC data 
Worker reporting
In addition to those in jobs that have been designated, legally or illegally, as independent contractors and therefore ineligible for workers' compensation, information about work-related injuries and illnesses is unavailable where workers do not report work-related injuries.  This may apply to temporary workers who, despite legal access to benefits, do not report injuries because they feel that their employment is tenuous, or workers who fear retaliation or sanctions (e.g. immigrant workers fearing being reported to authorities).   Workers may not report injuries and therefore not enter workers' compensation in some situations where so-called "safety incentive" programs are in place that reward absence of claims.  Worker underreporting is not unique to workers' compensation.  Injuries and illnesses not reported by workers would also not be reflected on OSHA logs and thus would not be collected in the BLS Survey of Occupational Injuries and Illnesses.  
In contrast, there are also concerns, under certain conditions, about the over-reporting of workplace injuries or the "secondary gain" of not returning to work in a timely fashion after disability. Some have cited a so-called "Monday effect" related to non-work related injuries sustained over a weekend that are submitted as being work related the following Monday.  [Card and McCall, 1996; Ruser, 1998, Hansen, 2012] Similarly there is concern of other types of moral hazard in claims reporting: prior to strikes (WC benefits being more generous than strike pay) or following layoffs. An alternative explanation for the association between impending layoffs and workers compensation claims rates, is that for some injury types where there is some latitude for workers to decide whether or not to file a claim (e.g. musculoskeletal disorders), workers may prefer to continue working as long as their job is secure because of their higher earnings as compared to WC benefits. Such workers then might file for WC once they become aware of the impending layoff in order to obtain treatment for their injury and to receive wage replacement benefits, which exceed benefits under unemployment insurance.  Research studies that attempt to quantify the competing effects are rare. 
Conclusion
The growth of the contingent workforce and workers on alternative work arrangements presents many challenges in the occupational safety and health arena.  State and federal laws often entail obligations and rights between employees and employers, but contingent work creates a lack of clarity about who fits into the categories of "employer" and "employee."  This results in ambiguities concerning responsibilities to maintain a safe and healthful workplace, difficulties in collecting and reporting data regarding hazards and injuries and illnesses, and raises questions about which, if any, benefit programs are available if health care or disability pay is necessary.
Contingent work may involve uncertainty about the length of employment, control over the labor process, degree of regulatory or statutory protections, and access to benefits.  These arrangements introduce specific difficulties in the areas of coverage of occupational injuries and illnesses through workers' compensation, and the particular types and scope of data that the system needs to distinguish between employers' hazard levels and classification.
This paper discussed the various mechanisms under which contingent workers are hired and supervised and explained differences between types of labor contracting, staffing entities, self-employment, and other nonstandard work relationships, including how statistical data sources measure these work categories.   It reviewed the empirical literature on the relationship between workplace injury and illness risk and different forms of work, including the difficulty of distinguishing the differential impact of temporary work versus employee tenure at work.  
The paper highlighted differences in availability of regulatory protections and safety net benefits among various types of contingent workers and how these different arrangements incorporate safety and health incentives in the financing of programs.  Recognizing the growth in the use of Professional Employee Organizations (PEOs), the paper described different models for writing workers' compensation insurance policies to cover PEO workers.  Finally, it discussed challenges caused by contingent work for accurate data reporting in existing injury and illness surveillance and benefit programs, and opportunities for overcoming obstacles to effectively using workers' compensation data for occupational safety and health.  
  




References
Bardasi, E. & Francesconi, F. (2004). The impact of atypical employment on individual 
	wellbeing: evidence from a panel of British workers. Social Science & Medicine, 58(9), 
	1671-1688.

Belous, R.S. (1989). The contingent economy: The growth of the temporary, part-time and subcontracted 
	workforce. Washington, DC: National Planning Association.
Benavides, F. G., Benach, J., Diez-Roux, A. V. & Roman, C. (2000). How do types of employment relate to health indicators? Findings from the second European survey on working conditions. Journal of Epidemiology and Community Health, 54(7), 494-501. 

Benavides, F. G., Benach, J., Muntaner, C., Delclos, G. L., Catot, N. & Amable, M. (2006). 
	Associations between temporary employment and occupational injury: What are the 
	mechanisms? Occupational and Environmental Medicine, 63(6), 416-421.
Boden, L. I. & Ozonoff, A. (2008). Capture - recapture estimates of nonfatal workplace injuries 
	and illnesses. Annals of Epidemiology, 18(6), 500-506.
Bonauto, D., Silverstein, B., Adams, D. & Foley, M. (2006). Prioritizing industries for occupational injury and illness prevention and research, Washington State workers' compensation claims, 1999 - 2003. Journal of Occupational and Environmental Medicine, 48(8), 840-851. 

Breslin, F. C. & Smith, P. (2005). Trial by fire: a multivariate examination of the relation between job tenure and work injuries. Occupational and Environmental Medicine, 63(1), 27-32. 

Bureau of Labor Statistics. (1996). Current population survey, February 1995 supplement. Washington, DC: United States Department of Labor.

Bureau of Labor Statistics. (1997). Survey of occupational injuries and illnesses, 1995. 
	Washington, DC: United States Department of Labor.
Bureau of Labor Statistics. (2005). Contingent and alternative employment arrangements. 
	Washington, DC: United States Department of Labor.
Bureau of Labor Statistics. (2011). Survey of occupational injuries and illnesses, 2010. 
	Washington, DC: United States Department of Labor.
Card, D. & McCall, B.P.  (1996).  Is Workers' Compensation Covering Uninsured Medical Costs?: Evidence from the 'Monday Effect". Industrial and Labor Relations Review, July 1996
Fan, Z. J., Bonauto, D. K., Foley, M. & Silverstein, B. (2006). Underreporting of work-related 
	injury or illness to workers' compensation: Individual and industry factors. Journal of 
	Occupational and Environmental Medicine, 48(9), 914-922.
Foley, M. P. (1998). Flexible work, hazardous work: The impact of temporary work arrangements on occupational safety and health in Washington State, 1991-1996. Research in Human Capital and Development, (12), 123-148.
 
Golden, L. & Appelbaum, B. (1992). What was driving the 1982 - 88 boom in temporary employment? The American of Economics and Sociology, 51(4), 473-493. 

Gordon, D. M. (1996). Fat and mean: The corporate squeeze of working Americans and the myth of managerial downsizing. New York, NY: Martin Kessler Books.

Hansen, Benjamin. (2012).  The "Monday Effect" in Workers' Compensation: Evidence from the California Reform.  Working paper, University of Oregon.

Kochan, T. A. (1991). Managing workplace safety and health: The case of contract labor in the US petrochemical industry. Dallas, TX: Lamar University Press.

Mangum, G., Mayall, D. & Nelson, K. (1985). The temporary help industry: A response to the dual internal labor market. Industrial and Labor Relations Review, 38(4), 599-611. 

Mayhew, C. & Quinlan, M. (2002). Fordism in the fast food industry: Pervasive management control and occupational health and safety risks for young temporary workers. Sociology of Health & Illness, 24(3), 261-284. 

Morris, J. A. (1999). Injury experience of temporary workers in a manufacturing setting. Factors that increase vulnerability. Journal of the American Association of Occupational Health Nurses, 47(10), 470-480. 

NORA Services Sector Council. (2008). National Occupational Research Agenda (NORA): National |
	Services Sector Agenda for Occupational Safety and Health Research and Practice in the U.S.
Occupational Safety and Health Administration.  Part 1904: Recording and Reporting Occupational 
	Injuries and Illnesses. In: OHSA Regulations (Standards-29 CFR). Available at: 
	http://www.osha.gov/pls/oshaweb/owasrch.search_form?p_doc_type=STANDARDS&p_toc_lev 
	el=0&p_keyvalue=.   
Paoli, P. & Merllie, D. European Foundation for the Improvement of Living and Working Conditions, Luxembourg: Office for Official Publications of the European Communities. (2001). Third European survey on working conditions 2000.

Park, Y. & Butler, R. J. (2001). The safety costs of contingent work: Evidence from Minnesota. Journal of Labor Research, 22(4), 831-849. 

Parker, R. E. (1994). Flesh peddlers and warm bodies: The temporary help industry and its workers. Brunswick, NJ: Rutgers University Press.

Phipps, P. & Moore, D. (2010). Employer Interviews: Exploring Differences in Reporting Work Injuries 
	and Illnesses in the Survey of Occupational Injuries and Illnesses and State Workers' 
	Compensation Claims. Proceedings of the Survey Methods Research Section of the American 
	Statistical Association.  American Statistical Association, 2910-2924.
Polivka, A.E. (1996). Contingent and alternative work arrangements, defined. Monthly Labor Review. 
	119(10):3-9. Washington, DC: Bureau of Labor Statistics.
Rosenman, K. D., Kalush, A., Reilly, M. J., Gardiner, J. C., Reeves, M. & Luo, Z. (2006). How much work-related injury and illness is missed by the current national surveillance system? Journal of Occupational and Environmental Medicine, 48(4), 357-365. 

Ruser, J.W.  (1998). Does Workers' Compensation Encourage Hard to Diagnose Injuries?  The Journal of Risk and Insurance, 65(1), 101-124.
 
Saha, A., Ramnath, T., Chaudhuri, R. N. & Saiyed, H. N. (2004). An accident-risk assessment study of temporary piece rated workers. Industrial Health, 42, 240-245. 

Saloniemi, A., Virtanen, P. & Koivisto, A. M. (2004). Is fixed-term employment a new risk for adverse physical working conditions? International Journal of Occupational Safety and Ergonomics, 10(1), 35-42. 

Silverstein, B., Welp, E., Nelson, N. & Kalat, J. (1998). Claims incidence of work-related disorders of the upper extremities: Washington State, 1987 through 1995. American Journal of Public Health, 88(12), 1827-1833. 

Silverstein, B., Viikari-Juntura, E. & Kalat, J. (2002). Use of a prevention index to identify industries at high risk for work-related musculoskeletal disorders of the neck, back, and upper extremity in Washington State, 1990 - 1998. American Journal of Industrial Medicine, 41(3), 149-169. 

Smith, C. K., Silverstein, B. A., Bonauto, D. K., Adams, D. & Fan, Z. J. (2010). Temporary workers in Washington State. American Journal of Industrial Medicine, 53(2), 135-145. 

Tompa, E., Scott-Marshall, H. & Fang, M. (2008). The impact of temporary employment and job tenure on work-related sickness absence. Occupational and Environmental Medicine, 65(12), 801-807. 

 U.S. Government Accountability Office. (2006).  Employment Arrangements-Improved Outreach Could Help Ensure Proper Worker Classification. Retrieved May 15, 2012 from http://www.gao.gov/products/GAO-06-656

U.S. Government Accountability Office. (2009). Enhancing OSHA's records audit process could improve the accuracy of worker injury and illness data (US GAO-10-10). Retrieved from U.S. Government Accountability Office website: http://www.gao.gov/assets/300/298510.pdf 

Virtanen, M., Kivimaki, M., Elovainio, M., Vahtera, J. & Cooper, C. L. (2001). Contingent employment, health and sickness absence. Scandinavian Journal of Work, Environment, and Health, 27(6), 365-372. 

Virtanen, M., Kivimaki, M., Elovainio, M., Vahtera, J. & Ferrie, J. E. (2003). From insecure to secure employment: Changes in work, health, health related behaviours, and sickness absence. Occupational and Environmental Medicine, 60(12), 948-953. 

Washington State Department of Labor and Industries. (2007). Washington state department of labor and industries workers' compensation insurance manual. Olympia, WA: Washington State Department of Labor and Industries.

Appendix A
The Types of Data Typically Collected by a WC Rating Bureau 
Aggregate ratemaking for workers compensation relies on financial data - total number of lost time claims, total medical and indemnity payments, total medical and indemnity incurred costs (payments plus outstanding case reserves on open claims)  -  submitted annually by all workers comp insurers in a state.  Currently the financial data submitted to NCCI covers the most recent 20 years  -  each year is reported separately and an update of the financial data of each year is reported at annual increments.  For example, in 2009 there was one report for the year 2008 but there were 20 report periods for the year 1989. Typically financial data are provided for both "policy years," which track results for all insurance policies written in a calendar year (a policy year, therefore, covers 24 months as policies written in late December of one year cover claims that might occur in the following calendar year), and for "accident years," which track results for claims due to injuries or illnesses occurring in a calendar year.  
Class ratemaking relies on analyzing the claims experience for each policyholder.  NCCI receives these data  -  "unit statistical records" - from insurance carriers annually.  There is a report for each policyholder for the year indicating the status of each claim associated with that policy: claims cost data  -  e.g. payments, incurred costs (i.e. payments plus case reserves); injury detail  -  e.g. part of body, nature and cause of injury; occupation of injured worker.  In NCCI states, a unit record for a given WC insurance policy is submitted annually for 5 years; this time frame recently was expanded to 10 years.
Experience rating calculates debits or credits ("experience mods") for individual policyholders based on their claims experience relative to other policyholders in their industry group.  This analysis is based on the three most recent years of unit statistical records. This analysis of individual policyholder experience is more heavily influenced by frequency as the cost of individual claims is capped.  This addresses some statistical concerns but also is thought to encourage increased attention on workplace safety.

