THE ECONOMIC ROLE AND IMPACT OF WORKERS' COMPENSATION

Document Sample
THE ECONOMIC ROLE AND IMPACT OF WORKERS' COMPENSATION Powered By Docstoc
					 THE ECONOMIC ROLE AND IMPACT OF
      WORKERS’ COMPENSATION




                 APRIL 8, 1998


                  Final Report
                PREPARED FOR:
  ROYAL COMMISSION ON WORKERS’ COMPENSATION
             IN BRITISH COLUMBIA
                VANCOUVER, BC



                  PREPARED BY:
                RON GIAMMARINO
                 GERALD GARVEY
         UNIVERSITY OF BRITISH COLUMBIA
FACULTY OF COMMERCE AND BUSINESS ADMINISTRATION
                                              TABLE OF CONTENTS
EXECUTIVE SUMMARY ........................................................................................................................... i

CHAPTER ONE: ECONOMIC FACTORS THAT INFLUENCE AND ARE INFLUENCED BY
WORKERS’ COMPENSATION .................................................................................................................1
   1.1 OVERVIEW .........................................................................................................................................1
   1.2 WHO PAYS FOR INSURANCE .................................................................................................................2
   1.3 INCENTIVE CONSIDERATIONS ...............................................................................................................3
   1.4 FINANCING CONSIDERATIONS ...............................................................................................................4
   1.5 THE RATIONALE FOR WORKERS’ COMPENSATION ................................................................................5
CHAPTER TWO:COMPULSORY PARTICIPATION BY EMPLOYERS AND EMPLOYEES.........7
   2.1 ECONOMIC RATIONALE S FOR COMPLUSION ............................................................................................9
   2.1.1 IMPROVING EMPLOYEE WELFARE ........................................................................................................9
   2.1.2 PROTECTING THE REST OF SOCIETY...................................................................................................10
   2.2 VOLUNTARY WORKERS’ COMPENSATION SYSTEMS .............................................................................13
   2.2.1 TEXAS ..............................................................................................................................................13
   2.2 .1.1 WHICH TYPE OF EMPLOYERS OPT OUT .........................................................................................13
   2.2 .1.2 HOW DO INJURED WORKERS FARE WITH NONSUBSCRIBING FIRMS ................................................14
   2.2.2 THE UNITED KINGDOM BEFORE THE WORKERS’ COMPENSATION ACT OF 1897..............................15
   2.2.3 WORKERS’ COMPENSTATION UNDER COMMON LAW IN THE US, 1897-1903 .....................................17
   2.3CONCLUSIONS.......................................................................................................................................17
CHAPTER THREE: FIXED COMPENSATION SCHEDULES............................................................18
   3.1 WHO ACTUALLY PAYS FOR WORKERS’ COMPENSATION BENEFITS? ..................................................20
   3.2 BENEFIT RATES AND CLAIM COSTS ....................................................................................................21
   3.2.1 DETERMINANTS OF MEDICAL EXPENSES ..........................................................................................22
   3.3 CONCLUSIONS......................................................................................................................................23
CHAPTER FOUR: THE ECONOMICS OF THE WCB ASSESSMENT SYSTEM...........................25
   4.1 A SKETCH OF THE CURRENT SYSTEM ..................................................................................................27
   4.1.1 PLACEMENT OF EMPLOYERS INTO SUBGROUPS.................................................................................27
   4.1.2 EXPERIENCE RATED ASSESSMENTS ..................................................................................................28
   4.2 EXISTING RESEARCH ON THE INCENTIVE EFFECTS OF EMPLOYER ASSESSMENT ................................30
   4.2.1 RESEARCH TO 1996 ..........................................................................................................................3 0
   4.2.2 HYATT’S STUDY OF THE INCENTIVE EFFECTS OF ERA......................................................................31
   4.2.3 ASSESSMENT AND CRITIQUE.............................................................................................................32
   4.3 SOME EVIDENCE ON THE DETERMINANTS OF ACCIDENT RATES AT THE SUBCLASS LEVEL .................35
   4.3.1 VARIABLES TO BE USED ...................................................................................................................35
   4.3.2 RESULTS: THE DETERMINANTS OF ACCIDENT RATES .......................................................................39
   4.4 CONCLUSIONS ....................................................................................................................................42
CHAPTER FIVE: GOVERNANCE AND CONFICTS OF INTEREST ...............................................44
   5.1 GOVERNANCE AND THE HETEROGENEITY OF CLAIMS ON AN ORGANIZATION ....................................44
   5.1.1 GOVERNANCE OF THE WORKERS’ COMPENSATION SYSTEM IN BC...................................................45
   5.2 CAN CONFLICTS BE REDUCED .............................................................................................................48




Garvey and Giammarino                                                        a
                          EXECUTIVE SUMMARY
A.   THE BASIC ECONOMIC                   CHARACTERISTICS             OF     WORKERS
     COMPENSATION.

1. Workers’ compensation is insurance, the value of which is recognized by both
   employer and employee as an important part of total compensation.
          Gainful employment inevitably involves some risk of injury. Workers take
             this risk into account in their labor market decisions: evidence supports the
             view that riskier jobs have higher pay and fringe benefits. Workers’
             compensation is insurance against the costs of injury and forms part of the
             overall compensation received by employees.

              Because it is mandated by statute, Workers’ compensation is often thought
               of as a benefit that is added to an independently determined wage rate. In
               fact, the B.C. statute explicitly prohibits employers from reducing wages
               to reflect increases in WCB assessments. This prohibition is, however,
               very difficult to enforce. Evidence from a number of jurisdictions shows
               that wages adjusted to take account of the value of the Workers’
               compensation package provided.

              When minimum wage laws effectively prohibit the wage from adjusting,
               evidence from the US shows that increased Workers’ compensation
               benefits leads to increased unemployment.

2. An important feature of the workers’ compensation plan is that participation is
    compulsory. We consider the arguments for and against compulsory participation and
    conclude that voluntary participation has sufficient benefits to make it worthy of
    consideration.
            It is possible that many of the concerns brought to the attention of the
               Royal Commission would be alleviated if employers could opt out of
               Workers’ compensation. For instance, those who feel that assessment rates
               are excessive could be invited to purchase insurance elsewhere.
               Employees who are dissatisfied with the manner in which WCB
               administers the program could obtain alternative coverage from an insurer
               that they deem more suitable. Most importantly, potential competition
               from other sources would help insure high levels of efficiency and
               customer service.

              The primary support for compulsion is based on a belief that there is a
               social cost of the decision to opt out of insurance that is not reflected in
               the private costs faced by employers and workers. For example, it may be
               socially costly to regulate private insurance or to resolve disputes that
               would result between employers and employees. Alternatively, the
               insurance market may simply not develop due to “adverse selection”
               where the behavior of high risk employers drives out low risk employers.

Garvey and Giammarino                        i
              A second argument in favor of compulsion is that employers would
               simply stop providing insurance and that this would reduce employee
               welfare. There are a number of reasons why this is unlikely to be the case.
               First, coverage can be made mandatory while at the same time allowing
               private provision of insurance. Second, even if coverage decreased, it is
               not clear that firms would opt out or that workers would be worse off if
               they did. There is considerable evidence to suggest that a decrease in
               injury insurance benefits will lead to an offsetting increase in wages.
               Third, evidence from Texas, where firms are currently allowed to opt out
               of mandatory coverage, and from other jurisdictions where voluntary
               participation had previously been allowed, indicates that reasonable injury
               compensation was often provided in the absence of compulsion.

3. Another important feature of the Workers’ compensation System is the fact that
   compensation for injury is independent of the underlying need for insurance or the
   extent to which the employee contributed to the risk of injury. We consider the
   following aspects of this system.

              The main benefit of this system is in the fact that it eliminates the need to
               assign fault and assess the cost of specific injury. Despite the expense of
               the WCB system, the alternative mechanism based on tort law may be
               more costly.

              Although the cost of the fixed benefits system is ostensibly paid for by the
               employer, there is considerable evidence that the cost is treated as part of
               the total wage and benefit package: Higher valued compensation benefits
               are related with lower wage levels.

              Employees do seem to respond quite strongly to the way in which benefits
               are determined. More generous wage benefits appear to significantly
               increase the number of claims made. The cost of treatment seems to be
               higher when being paid for as a compensation benefit than when it is not.
               Hence the overall cost of the system is quite sensitive to the benefits
               offered.

4. Employer assessments that are independent of the employer’s claim history may
   reduce the incentive to adopt costly safety measures. In practice two adjustments to
   assessments are used in an attempt to mitigate adverse incentives. First, employers
   are placed into subgroups based on the risk of the job. Second, an Experience Rated
   Assessment (ERA) system is used where employer assessments are adjusted relative to
   subgroup levels based on their relative claim history.

          Some fear that the introduction of ERA has driven employers to suppress
           legitimate worker’s claims in order to reduce their assessments. We do not
           know of any studies that support this view. Moreover, there is no data
           available that would allow us to evaluate this hypothesis. It is not clear to us,
           however, that workers would involuntarily refrain from reporting an injury. It


Garvey and Giammarino                        ii
           is also not clear how a worker is worse off if they are induced to voluntarily
           decide not to file a claim.
          Based on previous work and our own analysis of the data for B.C. we
           conclude that the introduction of ERA has had an effect on employer safety
           performance. The impact is, however, less than one might expect due to the
           fact that the ERA system in BC does not allow full experience adjustment to
           take place but instead caps the impact that a good or bad experience can have.
          There are considerable differences in the characteristics of employers within a
           given sub class. Even when ERA is included, this leads to considerable cross
           subsidization within a subgroup. Consistent with this, we find that the safety
           records of more homogeneous subgroups to be better both before and after the
           introduction of ERA.




Garvey and Giammarino                      iii
                       CHAPTER ONE
        ECONOMIC FACTORS THAT INFLUENCE AND ARE
          INFLUENCED BY WORKERS’ COMPENSATION
             Employment inevitably introduces some risk of injury to a worker. Injuries
       are costly to both the worker and society. Workers forgo income if required to be
       off the job, endure pain and suffering, and may obtain less enjoyment from their
       leisure activities. Society may lose a productive worker for some time, may lose
       the overall contribution of a productive citizen, may pay part of the costs of
       treatment, and may also experience problems that spring from the disruption to
       the worker’s life. A Workers’ compensation system provides partial insurance
       against these costs, thereby spreading the hardship of injury among other workers
       and society.

              The economics of Workers’ compensation starts with the recognition that it
       is valuable insurance that is part of the total employment agreement. In contrast to
       the view often presented, Workers’ compensation is not ancillary to an
       independently determined wage and benefit package but is an important part of
       that package. The basic question that we address in this report is: How does the
       specific form of insurance present in British Columbia affect the economic well
       being of its participants.

             The specific issues that we set out below and those covered in more detail
       in the chapters that follow are, we feel, central to the task of the Royal
       Commission in that they examine tradeoffs that are faced in designing a system
       that “ . . . meets the needs of the people of British Columbia for a high quality
       public system that is equitable, effective and efficient . . . “

             In the remainder of this chapter we provide an overview of the economic
       nature of a Workers’ compensation. This overview motivates the more detailed
       study of a number of features of the system that is found in the chapters that
       follow.

1.1 OVERVIEW

             On the job injuries are a by-product of production. The prospect of a work
       related injury decreases the benefit of employment in two ways. The first is the
       direct cost of an injury if one should occur. The second less direct impact is the
       personal cost of having to deal with more uncertainty in one’s life. If the
       employment arrangement merely provided wages in exchange for work, the
       employee would face the risk that an injury would involve large medical and
       rehabilitation costs at exactly the time when their ability to earn an income is
       suspended by the injury. Risk aversion can be defined as a willingness to pay in
       order to avoid this uncertainty.




Garvey and Giammarino                       1
              Risk averse individuals facing risk of on the job injury seek insurance to
       eliminate the uncertainty. Insurance translates the possibility of a catastrophic
       loss for one person into the certainty of a specific expense. In fact, research
       discussed in the next chapter shows that many employees in the UK and the US
       were covered by company sponsored plans before Workers’ compensation plans
       were introduced. Hence, Workers’ compensation plans are first and foremost
       insurance whose value to workers depends on the riskiness of the job and the
       ability of the worker to deal with the consequences of a workplace accident
       through personal wealth.


1.2 Who Pays for Insurance?
             Individuals take the risk of the job into account in deciding where to work
      and adjust their required compensation accordingly. The earnings of someone in a
      job that is considered relatively dangerous are typically higher than the earnings
      of those in safer jobs. This difference is referred to as a “compensating
      differential” and there is considerable evidence that compensating differentials are
      in fact paid. The size of the compensating differential is also affected by the value
      of the insurance benefits that are provided to the worker.

              Perrin, Thorau & Associates (1997) find that the funds for compensation
       assessments are literally taken from funds held by the employer in virtually all
       jurisdictions covered by their survey. It can be argued, however, that, to the extent
       that the employee’s wage is reduced by the premiums paid, the employee in fact
       pays for the insurance benefit received. Research reviewed in Chapter 3 provides
       strong evidence that wages are in fact reduced in the presence of higher Workers’
       compensation costs. For instance, Moore and Viscusi (1990) find that lower
       wages more than offset the employer’s total cost of workers’ compensation. This
       is consistent with the view that the insurance provided is part of the total
       employment contract.

              Also consistent with this view are studies which show that employment is
       not generally affected by Workers’ compensation insurance. If higher insurance
       costs are offset by lower wages, the total labor costs are the same and hence the
       demand for labor will be unchanged. An exception to this, however, is
       employment of youth and young adults. These groups are typically paid minimum
       wage so that an increase in Workers’ compensation payments cannot be converted
       into lower wages. In this case, higher Workers’ compensation costs correspond to
       higher wages and one would expect a lower level of employment. Kaestner
       (1996) finds evidence to support this prediction. In a study of the relationship
       between employment, the minimum wage, and workers compensation costs in the
       US, Kaestner finds that a 1% increase in Workers’ compensation insurance costs
       is associated with a 1.5% lower level of employment of youth and young adults.




Garvey and Giammarino                        2
             The relationship between employment and workers’ compensation costs
       implies that industries that rely on minimum wage labor are most likely to be hurt
       by increases in Workers’ compensation costs and are more likely to leave
       jurisdictions where these costs are high. Unfortunately, we are not aware of any
       studies that systematically examine this hypothesis in detail.


1.3 Incentive Considerations
             We have argued that the formal manner in which the Workers’
       compensation is paid for does not alter the fact that these costs are part of the cost
       of production. Therefore they are essentially paid for by the consumer. However,
       the benefit coverage and the mechanism for determining assessments does affect
       the total cost of the system. This is due to the incentive effects of the system and
       these effects do depend on the legal structure of the contract. There are four
       channels through which incentives affect the costs of the insurance system.

              The first incentive effect is through behavior in the workplace. If the firm is
       legally responsible for the costs of injury, they will have an incentive to invest in
       accident prevention and training. If the worker bears some of the cost of the
       injuries sustained on the job, they will have an incentive to take preventative
       actions and investments. A similar principle applies to the costs of treatment and
       rehabilitation. If the worker does not share in any efficiency gain they will have
       no incentive to seek out efficient treatment and rehabilitation. Perverse incentives
       may even take hold where greater costs of treatment (for example the length of
       rehabilitation) are correlated with higher perceived benefits (for example time off
       of work).

              A second consideration is the firm’s decision to produce. As argued above,
       Workers’ compensation is part of the cost of production and this cost is higher for
       riskier endeavors. If a firm cannot cover all of its costs, including the costs of
       providing Workers’ compensation, then it should exit the business.

             To the extent that the system involves cross subsidies, riskier activities will
       be supported by safer activities. These cross subsidies might allow firms that
       should leave an industry to stay in business. We show in chapter 4 that even with
       experience rating, such cross subsidization does occur in BC. We are not able,
       however, to determine the extent to which this has allowed firms to continue
       producing even when they are not economically viable on a full-cost basis.

             Workers’ compensation may also affect the decision of individuals to enter
       particular lines of work. To some extent individuals are able to determine the
       extent to which they are able to perform certain functions safely. In the absence of
       Workers’ compensation they are likely to avoid jobs at which they are likely to be
       accident prone. Since Workers’ compensation attenuates the costs of accidents, it
       reduces the incentive for individuals to avoid jobs for which they are not well
       suited.



Garvey and Giammarino                        3
             A fourth consideration is the incentive that Workers’ compensation provides
       for individuals to report off the job injuries as being job related. This incentive is
       related to the degree that Workers’ compensation benefits exceed the benefits of
       private medical insurance. There are two main solutions to this problem. One is
       to monitor claims closely, hoping to detect misrepresentation and, through
       penalties, discourage such behavior. An alternative is to simply insure all injuries,
       whether or not they are incurred on the job, that is, to provide 24 hour coverage.
       Both of these solutions are very costly. According to Perrin Thorau & Associates
       (1997) only New Zealand covers all accidents.


1.4 Financing Considerations
             Workers’ compensation is financed through assessments. These assessments
       serve two purposes. On the one hand, they cover the administrative costs of
       operating the system and of providing services to injured worker. On the other
       hand, they provide a deposit to cover future, uncertain expenses.

              As with many types of insurance, difficulties arise when the contingency
       being insured is uncertain and will not be incurred for some time. One financing
       strategy is to charge the expenses to the employer group as they are incurred. That
       is, the worker is guaranteed that benefits will be provided while employers pay for
       the costs as they are incurred. The difficulty with this plan is, however, the
       possibility that the firm will default on its promise due to insolvency or simply
       closing down.

             The system in place for Workers’ compensation and most other insurance
       systems is to have employers pay enough each period to cover current and
       expected future benefit disbursements. This allows for a reserve fund to be
       established to cover future costs and the reserve fund provides some assurance to
       workers that the resources will be available to cover future benefits.

             The problem is that the amount paid is based on expectations about two
       risky variables; the costs of the benefits and the returns on the investments made
       on the reserve fund. More succinctly, the size of the reserve fund must be
       determined and the fund size will be dependent on the investment strategy
       followed by the reserve fund managers as well as the technological risks of
       workplace injury and treatment.

             Determining fund size in the presence of these risks is difficult enough. It is
       compounded by the fact that insurance funds of this sort also create two specific
       incentive problems. One is to ignore work place hazards that are difficult to
       monitor and that cause injuries over a long period of time. Since the treatment
       costs will appear in the future and will not increase assessments until then, they
       have both a lower present value and a lower probability of being paid since the
       employer may be out of business by then. A second related incentive effect stems
       from the possibility of bankruptcy. Firms that are in financial distress face lower
       expected future assessments since there is a good chance they will not be in
       business. These firms have a greater incentive to ignore injury prevention.

Garvey and Giammarino                        4
1.5 The Rationale for Workers’ compensation

             The Terms of Reference for the Royal Commission stated that the
       Commission should examine the structure of Workers’ compensation in order to
       provide the people of British Columbia with a system that is equitable, effective
       and efficient. Moreover, the system should deliver

              Accident prevention

              no-fault compensation

              collective employer liability

              industry funding

              universal coverage

              administrative adjudication

           In light of the discussion presented in this chapter, it is clear that important
       tradeoffs will emerge and we examine these tradeoffs in more detail in the
       following chapters.

           A primary result of basic economic analysis is that efficiency is enhanced by
       the forces of competition. Although the Workers’ compensation system is
       intended to be an efficient provider of services, it is currently administered by a
       monopoly supplier with a resulting large number of complaints about both its
       costs and its effectiveness. In the next chapter we examine the question of
       compulsory participation. We consider the possibility of allowing employers to
       opt out of the current system. This does not necessarily mean that universal
       coverage will be compromised. Employers may be required to provide insurance
       but be able to choose the insurance scheme they wish to belong to.

             Economic efficiency also requires that insurance benefits be commensurate
       with the economic need of the workers while preserving incentives to take care. In
       contrast, the current system is required to provide no fault coverage. This implies
       that benefits are based on a fixed schedule where compensation is partially
       adjusted for the general characteristics of the injury but worker specific factors are
       suppressed. In chapter 3 we examine the consequences of the fixed compensation
       system.

             The efficiency of the system is enhanced if the incentives of the insured are
       accounted for. On the other hand, the requirement of universal coverage and
       equitable treatment can be taken to mean that assessments are only partially
       adjusted for worker specific risks and decisions. In chapter 4 we examine the
       assessment system in order to consider the extent to which the use of subgroups
       and experience rating mitigates adverse incentives.

             Finally, the efficiency of the Workers’ compensation system depends on the
Garvey and Giammarino                          5
       motivation of the organisation that administers the system. This is in general a
       function of the governance system in place and the importance of governance is
       especially important when the market forces of product market competition and
       takeovers is absent. In chapter 5 we critically examine the governance of the
       Workers’ compensation Board of BC.




Garvey and Giammarino                     6
                                 CHAPTER TWO

   COMPULSORY PARTICIPATION BY EMPLOYERS AND
                  EMPLOYEES
                     Workers’ compensation resembles an ordinary “fringe benefit” of
       employment in many respects. Employers pay the WCB and in exchange their
       employees receive coverage. Moreover, employers in higher-risk businesses pay
       higher rates and their employees make more intensive use of the service. From an
       economic perspective, the outstanding feature of most workers’ compensation
       systems is not that they provide insurance nor that they are a non-wage form of
       compensation. What is noteworthy is that that workers’ compensation is both
       compulsory and highly standardised. It is compulsory in that an employer who
       wishes to operate in one of the covered businesses must pay the rates assessed by
       the WCB. It is standardised in that employees may have no more and no less
       coverage than is stipulated by the WCB. The “historic compromise” in which
       employers were required to finance the system and employees would agree to
       give up their rights to sue in exchange for a fixed compensation formula was
       reached over 80 years ago, and needless to say many employers and employees
       did not directly assent to the compromise even at its outset. Nonetheless, the
       compromise continues to bind parties today.

                    Compulsion occupies a central position in any economic analysis of
       workers compensation because free choice by individuals on the basis of their
       own perceived gains and losses is the essence of a competitive market system.
       Given a choice, individual employers would assent to the compromise if it
       provided cost-effective benefits to their employees and to themselves, while
       others would opt out for some alternative that they prefer.

                     Closer to the topic at hand, some of the most difficult issues that this
       Royal Commission is likely to encounter might be substantially ameliorated if
       workers’ compensation were voluntary. Employers who complained about
       excessive assessment rates would have the opportunity to purchase insurance
       elsewhere, or to explain to their workforce why they did not see fit to purchase
       WCB coverage. Similarly, employees who expressed their dissatisfaction with
       aspects of the WCB in submissions to this commission could turn their ire to their
       employer and demand that they seek alternative coverage. The need to attract
       “customers” should also provide a spur to increased efficiency on the part of the
       WCB. Finally, allowing parties to opt out could provide hard information as to
       the value of the various services provided by WCB. This final point, most
       commonly associated with Hayek (1945), is so basic that it is often overlooked.
       It asks, how we could really know whether the WCB is functioning well?
       Complaints, and studies by experts, are an imperfect guide for the fundamental
       reason that we do not know what alternatives could be made available in BC. A



Garvey and Giammarino                        7
        bona fide choice in favour of WCB coverage would be a strong and unambiguous
        signal that its services are valued.

                        The above does not necessarily imply that compulsion is wrong,
        but that it requires careful consideration and justification. That is, there must be
        compelling counters to the above arguments, which we summarise in the next
        section. Clearly, governments must see a compelling rationale for compulsion
        since, as summarised by Perrin, Thoreau and Associates (1997), it is the rule
        throughout the industrialised world with only one exception to which we will
        return later. Moreover, compulsory participation is not new. It appears to have
        originated with the UK Workers’ Compensation Act of 1897. Epstein (1982)
        notes that employers were allowed to opt-out in principle, but in practice the Act
        presented substantial barriers to so doing.     First, employers had to provide a
        substitute accident compensation plan and were not allowed to “merely” pay
        higher wages. Second, each private arrangement had to be registered, and
        reviewed every five years or whenever a complaint was received, whichever
        occurred first. Finally, employers could not provide a uniform plan for an entire
        workplace. Thus, a single employee could effectively force the employer, and all
        the other employees of the firm, to join the State-sponsored workers’
        compensation system.

                       Compulsion has been more extreme in the case of BC Workers’
        Compensation. Opt-out provisions were only ever allowed for a subset of
        employers. Moreover, Bill 63, in force Jan 1 1994, extended coverage to
        effectively all BC employers and workers. This involved an increase in coverage
        of approximately 18,000 employers and 150,000 workers. Just over 6000 of these
        new employers were previously voluntarily registered. The apparent position of
        the WCB is that compulsion is proper and necessary, and that broader coverage is
        better whether or not the employers and employees concerned wish to be
        involved. The only discussion we were able to obtain was from the June 7, 1993
        summary of the decision of the WCB governors which stated that: “The
        governors of the WCB consider that coverage under the act should be as broad
        and inclusive as public policy dictates, ...the most effective and publicly
        accountable way to achieve this broad and inclusive coverage under the act is for
        the minister of Labour and Consumer Services and his ministry to take the
        initiative in developing and presenting amendments to the Act in the Legislative
        Assembly”.1 From the content of Bill 63 it is apparent that the dictates of public
        policy imply universal and compulsory coverage.

                      The remainder of this chapter is organised as follows. Section 2
        summarises potential economic justifications for compulsion, and also indicates
        the potential associated costs. Section 3 provides some evidence on what would
        happen if firms were allowed to opt out of Workers’ Compensation coverage by
        summarising evidence from two cases in which workers’ compensation coverage

1
 Thanks again to Gerry Schive and Angela Welz for locating this document and for their thorough but
ultimately unsuccesful search for other evidence of debate.


Garvey and Giammarino                              8
       is (was) not compulsory.

2.1 ECONOMIC RATIONALES FOR COMPULSION


2. 1.1 Improving Employee Welfare
                     As noted above, the status quo in BC is that employers are required
       to pay assessments to the WCB and employees are required to forego their right to
       sue in exchange for the pre-set compensation schedules. The relevant question is
       therefore whether any employers should be allowed to opt out of the system in
       which they are not required to participate. Basic economic principles suggest that
       the ability to opt-out would have desirable consequences. First, participation in
       the WCB would only occur when WCB assessments and compensation schedules
       presented a cost-effective solution to the problem of safety at a given workplace.
       Also, the WCB would face competition which could both enhance their public
       image (parties now choose the WCB system over the available alternatives) and
       improve the mix of services they provide to employers and employees. To justify
       compulsion, this logic must somehow break down. We now summarise some
       ways in which it might do so.

                     The first and most obvious concern is that the existing form of
       workers’ compensation is good for employees but bad for employers, and absent
       compulsion employers would simply refuse to join the system or to provide
       adequate substitutes. There is an equally obvious response to this objection;
       workers compensation is inevitably part of an employee’s overall remuneration.
       Employers should provide accident insurance in order to attract and retain good
       employees if it is a cost-effective form of compensation, and as we shall see in the
       next section there is evidence that many employers have done exactly this. A
       common response to this argument is that workers without substantial bargaining
       power will be unable to demand adequate compensation for the hazards they face
       on the job. The fundamental weakness with this reply is that compulsory workers
       compensation does not increase the workers’ overall bargaining power. There is
       substantial evidence, summarised in the next chapter, that other aspects of
       compensation adjust. First, workers in more hazardous lines of business receive
       higher wages and more generous fringe benefits. Even more strikingly, Kaestner
       (1996) finds that increased workers compensation costs are almost always passed
       on to employees in the form of lower wages. The exception is the case of young
       employees where wages are at the legal minimum, where increased workers
       compensation costs reduce employment. For no type of worker does compulsory
       workers’ compensation serve to increase overall remuneration. This result is to
       be expected, since bargaining power and higher overall remuneration come from
       (1) the possession of highly valued skills and (2) modest numbers of persons who
       share these skills. Workers’ compensation and other legislation do not
       automatically create high-paying jobs, and if employers are mobile will do the




Garvey and Giammarino                       9
        opposite.2

2.1.2 PROTECTING THE REST OF SOCIETY

                      While there is good reason to doubt that compulsory workers’
        compensation reliably improves the lot of employees, there is still the rest of
        society to consider. Workers’ compensation may protect or benefit parties other
        than the employers or employees concerned. If so, we have what is known as an
        eternality problem; employers and employees do not take account of the full
        social costs and benefits involved in their decision to join or not to join a
        centralised workers’ compensation system. Put another way, if firms were
        allowed to opt out they would do so to the benefit of themselves and their
        employees, but to the detriment of society. 3 This is clearly a very broad and
        inclusive argument. We will summarise the three aspects that are most relevant to
        workers’ compensation in BC.

                      In principle, employers and employees need not make any explicit
        agreement whatsoever to deal with the prospect of accidents on the job.
        Compensation and other arrangements could be arrived at in the event an accident
        actually occurred, and employees who were dissatisfied with the outcome could
        take the employer to court. This is in essence what is involved in the “common-
        law” alternative to workers’ compensation. Obviously, this alternative is
        desirable only to the extent the courts provide equitable and efficient judgements
        on the complex merits of on-the job injuries. To the extent that courts are
        excessively costly and/or unreliable adjudicators of individual compensation
        cases, there is an argument for an arrangement such as workers’ compensation in
        which workers forego the right to sue in exchange for a pre-specified benefit
        formula. While this argument appears to have substantial merit, it does not lead
        inevitably to compulsory participation in a universal, state-sponsored workers’
        compensation program. Individual employers and employees can, and in some
        cases have, made private agreements which involved exclusive remedies whereby
        employees forego the right to sue just as in the historical compromise in BC. As
        we discuss in more detail at the end of this chapter, English courts before the
        passage of the original workers’ compensation legislation refused employee suits
        in cases where a specialised compensation arrangement was already in place
        (Epstein, 1982, 789-93).

                 Another argument for compulsory participation is more specific to

2
  The BC workers’ compensation act forbids employers to deduct, directly or indirectly, any amount from
workers' wages to cover the cost of premiums. This provision cannot be effectively enforced, however, for
the simple reason that it requires knowledge of what workers would have been paid in the absence of the
system.
3
  Carmichael (1986) presents a model in which employers may invest too little in safety because imperfect
information on the part of employees makes wages insufficiently sensitive to accident rates. This argument
suggests that some form of insurance, or the operation of a tort system, may well be important but does not
provide a rationale for compulsory participation. In Carmichael’s model, both employers and employees
would gain by joining the system.


Garvey and Giammarino                               10
       countries such as Canada in which individuals bear only a small fraction of their
       own medical care costs. Such a system inevitably subsidizes those who make
       more intensive use of medical services. Particularly risky employers and their
       employees could effectively free-ride on the medical system. This provides a
       straightforward rationale for requiring employers to pay for expected medical
       costs on an ongoing basis through a system such as the WCB’s assessments. Of
       course, there are many practical hurdles to overcome of which the most basic is
       that of ascertaining whether or not a given injury was incurred in the course of
       employment. This question has proven so difficult that many US jurisdictions
       have considered introducing 24-hour coverage in which there is no attempt
       whatsoever to distinguish injuries on or off the job (see, e.g., Baker and Krueger,
       1993). We will return to this last issue in the next chapter as it raises important
       questions of how workers’ compensation benefits should be set. For the purposes
       of this chapter, it is important to note that there are many ways to address the
       problem of free-riding on the medical system. What is essential is that employers
       make a contribution towards the medical costs of their injured employees. This
       can be achieved by compulsory participation in a system such as BC’s but it need
       not. In many US jurisdictions, employers must carry medical coverage for their
       employees but can purchase such insurance from private companies. This
       approach has also been suggested in at least two independent submissions to this
       Royal Commission.

                 The above arguments suggest a role for compulsion in two directions;
       first, it may be prudent and efficient to require employers to carry a minimum of
       insurance. Second, it may be important to deny access to the courts as an avenue
       to seek out additional compensation. As indicated throughout this section,
       however, these arguments do not provide a justification for requiring all
       employers to participate in a single scheme administered by a government body.
       There are two sets of arguments which can justify such an arrangement. The first
       argument is that a competitive market can provide too little insurance due to a
       phenomenon known as adverse selection. As originally argued by Rothschild and
       Stiglitz (1976), safer employers can only prove their worth by agreeing to forego
       full insurance through sizeable deductibles or self-insurance. In this case, it is
       possible for matters to be improved if all parties are required to join a single
       system. Safer employers will wish to opt-out but they can only do so by
       accepting excessive risks on behalf of their employees. Thus, there is an
       argument that such employers should be compelled to join the system.

               While the adverse selection argument provides a potential, theoretical
       justification for compulsory participation in a centralised scheme, it does so only
       if some critical assumptions are satisfied. First, it must be the case that insurers
       have little ability to distinguish between risky and lower-risk employers through
       their industry, history, and safety practices. Only if this is the case would safer
       employers both opt-out of the centralised system and carry insufficient insurance.
       As we note below, in the one US state where employers can freely choose to opt-
       out of workers compensation, those who opt-out do not appear to differ from
       those who participate in terms of their overall safety. The second critical


Garvey and Giammarino                       11
         assumption for the adverse selection argument to justify compulsion is that the
         monopoly insurer will make every effort to be cost-efficient and offer reasonable
         prices, that is, that it will not exploit its monopoly position even though it is
         ideally placed to do so.

                        The final, and perhaps the most compelling argument against
         allowing BC employers to opting-out is the workers’ compensation system is
         simply that this would be a radical step. It is radical in terms of past policies, in
         terms of the mind-set that was evident in the passage of Bill 63, and in terms of
         practice in the rest of the industrialised world. Little is known about how opting-
         out would work in practice simply because there are so few cases where it is in
         fact allowed (see especially Perrin Thorau and Associates, 1997). Potentially
         complex administrative and legal issues could well be involved.4 On the other
         hand, these arguments can be used to defend any status quo. It is also important
         to recognise that the WCB currently encounters a host of administrative problems,
         including both employer appeals against classification decisions, employee
         appeals regarding compensation. It is possible that this burden could be reduced
         rather than increased by allowing those most dissatisfied with WCB practices to
         secure insurance elsewhere. Moreover, as we argue in the final section, overall
         governance of the WCB could be eased if the participants were all voluntary. It is
         not difficult to understand why employer, employee, and medical groups seek to
         influence the WCB’s decisions even when such pressures impede the
         management of the board as they apparently did in 1995 (e.g. Hunt et al, 1996).

                       As indicated above, compulsory participation in workers’
         compensation systems is the rule. However, there are exceptions which provide
         invaluable information in the actual outcomes experienced when parties are able
         to opt-out of the state-sponsored workers’ compensation system. We now turn to
         two such cases.

2.2 VOLUNTARY WORKERS’ COMPENSATION SYSTEMS


2.2.1 Texas
                It is routinely asserted that three US states allow employers to opt-out of
         the government’s workers’ compensation system: New Jersey, South Carolina,
         and Texas. We focus on Texas because very little information is available about
         South Carolina and the New Jersey Workers Compensation board’s public
         documents are careful to stress that opting-out of their system is not in fact
         permitted.5 Texas, by contrast has had a fully operational opt-out system since its
4
  We only recognise the presence of these costs and our inability to quantify them. In personal
communication, George Heinmiller indicated that there was little precedent or ground-work done on how
legislation might be changed.
5
   Specifically, their asserts that: “Any compilation of state workers compensation laws usually indicates
that New Jersey is an elective state when in reality such is not the case. In view of the confusion and lack of
understanding, there have been attempts to market other forms of coverage which have been purported to
satisfy the requirements of statute. These activities result from erroneous conclusions by reason of


Garvey and Giammarino                                 12
        inception in 1913. Moreover, employers are permitted to self-insure. In January
        1995, 44% of Texas employers covering 21% of Texas employees choose not to
        subscribe to the state workers’ compensation system (Public Policy Research
        Institute, 1995).


2.2.1.1 Which Types of Employers Opt-out?
               As the above numbers suggest, many of the firms that choose to opt out
       are small. The System Performance Update (1995) also notes that employer
       nonsubscription (opt-out) rates vary substantially across industries. Firms in
       mining are particularly likely to choose to participate in the sate-sponsored
       system. In fact, all mid-sized employers (100-500 employees) were subscribers.
       Firms in the retail trade industry were least likely to subscribe, especially smaller
       establishments with 50-99 employees.

                One concern with a system that allows employers to opt in and out of
        workers’ compensation is that policies might be unstable. It would be very
        difficult, for example, for an employee to assess her likely coverage in the event
        of an accident if her employer were to shift back and forth between private and
        public coverage. In fact, however, there is substantial stability in Texas
        employers’ participation decisions. Over half of the employers have always
        subscribed to the state system, and nearly a third have never subscribed.
        Moreover, less than 5% of employers have changed from not subscribing to
        subscribing, while less than 15% have left the system after some history of
        subscribing.

                 Perhaps the most important question is whether opt-out firms are safer or
        less safe than those which choose to participate in the state-sponsored system.
        As noted in the pervious section, if there is a serious adverse selection problem
        we would expect to see the safer employers opt-out. The fact that firms in retail
        trade tend to opt-out and that firms in mining tend to participate suggests that such
        a force may be at work. However, Butler (1996) finds that if we simply compare
        subscribers and non-subscribers in the same industry from 1993-1995, there is no
        statistically significant difference in accident rates. Indeed, nonsubscribing
        employers are less safe than subscribing employers in 3 of the 10 industries he
        studied. Interestingly, these are also the industries where subscribing firms are
        most prevalent (mining, finance and transport).


2.2.1.2 How do Injured Workers Fare in Nonsubscribing firms?
              Butler (1996) reports that in many cases the avowed reason for opting-out
       of the Texas workers’ compensation system is to improve control of medical
       costs. Many firms also opt-out of insurance arrangements entirely, with 70%


reference to only a part of the Law. There is no means for an employer or an employee to escape the
provisions of the New Jersey Workers' Compensation Law, Title 34, Chapter 15, Articles 1 to 9; R.S.
34:15-1 to R.S. 34:15-128. Neither an employer nor an employee may reject it.”


Garvey and Giammarino                              13
       paying wage replacement and 55% paying medical costs out of company funds.
       Not surprisingly, it is primarily the larger nonsubscribing employers who choose
       such self-insurance arrangements. The obvious concern raised by these facts is
       that employees of nonsubscribing employers are systematically disadvantaged.
       To address this issue, the Texas Workers’ Compensation board commissioned a
       telephone survey of 366 workers employed by nonsubscribing firms injured on or
       after January 1, 1993. The results are reported in Experiences of Injured Workers
       Employed by Nonsubscribing Employers (1997).

               The most important and immediate finding of the survey is that most
       employees are in fact covered by their employer or the employer’s insurance
       company. Over ninety percent indicated they had some coverage and 81%
       reported that their employer covered all medical costs. Moreover, 87% stated that
       coverage lasted for the entire duration of treatment. There is less evidence that
       workers received full income benefits for the time they missed due to injury,
       which is not surprising since few if any workers’ compensation arrangements
       replace 100% of wages. Nonetheless, 82% of injured workers reported that they
       received some pay for time missed, and fully 42% indicated that their pay was
       roughly equivalent to their earnings before the injury. There is also substantial
       evidence that, like most government-sponsored workers’ compensation plans,
       wage replacement is not indefinite. Specifically, 38% of the injured employees
       surveyed indicated that their wage replacement period was shorter than the term
       of their disability.

               Both wage replacement and medical benefits appeared to be slightly
       greater for employees with “severe” injuries, defined as those which involved
       surgery, hospitalisation or more than three months of lost work time. Butler
       (1996) provides additional evidence on the differences between injury types and
       rates in subscribing and nonsubscribing firms. Specifically, he finds that accident
       frequency is slightly greater in opt-out firms while severity, in the form of lost
       work day durations, is greater in participating firms. Also, chronic conditions
       such as strains and sprains are more prevalent in participating firms. Thus, it
       appears that nonsubscribing employers do have some ability to control the costs
       of at least some types of claims. These findings seem to mirror those of Hyatt’s
       (1997) study of the effects of experience rating of BC firms (which are of course
       all participating). Specifically, Hyatt finds that the introduction of experience
       rating appears to reduce claims rates but to extend their length. Since the Texas
       Workers’ Compensation system is also experience-rated, it is possible that the
       same forces are at work. In both jurisdictions, experience rating tends to penalize
       claims frequency more than claims length. Butler’s (1996) findings are to be
       expected if nonsubscribing firms face penalties for injury severity that are more in
       line with penalties faced for number of claims.

              Finally, it is important to note that the Experiences of Injured Workers
       Employed by Nonsubscribing Employers (1997) did not find evidence that private
       workers’ compensation arrangements involved massive litigation. Specifically,
       only 13% of the workers surveyed chose to hire an attorney, and 9% filed a


Garvey and Giammarino                       14
        lawsuit because of their injury. Of course, such numbers must be understood to
        reflect the employees’ perceptions of the court system in Texas as well as the
        generosity of their employer. We can only speculate as to the outcomes in
        jurisdictions such as California where awards to plaintiffs have been abnormally
        large.

                Texas provides an invaluable case-study of how workers compensation
        systems actually operate in the absence of compulsion. While the telephone
        survey conducted to study the experiences of injured workers involved the
        subjective views of a fairly small number of employees, it is difficult to imagine
        more objective ways in which this elusive but important topic could have been
        addressed. The other studies, by Butler (1996) and the Public Policy Research
        Institute (1995), are highly credible as they involved large samples, peer review,
        and world-class researchers in the field. Unfortunately, Texas is the only
        jurisdiction which permits such systematic study of voluntary workers’
        compensation systems in the industrialised world. We will, however, add two
        further “data points”; experiences in the UK and the US in the period before
        workers’ compensation legislation. These cases are not only useful as additional
        information, they also directly address the common perception that workers’
        compensation legislation emerged in an environment where employees were left
        to fend for themselves in the case of an injury.


2.2.2 The United Kingdom Before the Workers’ Compensation Act of 1897
               Much of what we know about workers’ compensation prior to 1897 comes
       from case-law. As Epstein (1982) stresses, much of the law addressed the
       difficult and fundamental problems of assigning fault for a particular accident,
       and assessing whether an injury was truly incurred in the course of employment.
       Of greater relevance for current practice are those cases which the courts
       dismissed because the employer had a scheme in place for compensating
       accidents. In two landmark cases, Griffiths vs Earl of Dudley (1882) (coal
       mining) and Clements vs London and Northwestern Railway (1894), an injured
       employees’ claim was set aside because the employer had already established a
       method of compensation which was operating at the time the employee joined the
       firm. Moreover, in both cases the schemes bore a remarkable resemblance to
       modern workers’ compensation; it was financed from worker contributions, and
       compensation was paid to anyone injured or killed in course of employment
       without regard to fault.6 In both cases, the compensation paid was less than 100%
       of the employee’s regular wages. Moreover, an unpublished paper (Valjanowski,
       1980) further reports that 25% of UK coal mines, most in Lancashire where
       mining was most dangerous, had similar plans.

                 In some respects, this evidence suggests that modern workers’

6
  Of course, the costs of BC and other modern workers’ compensation schemes are formally borne by
employers. But as we argue throughout this report, employees do effectively contribute to the scheme in
the form of lower wages and fringe benefits than they would otherwise receive.


Garvey and Giammarino                              15
       compensation arrangements have an enduring economic logic. The assignment of
       fault was apparently viewed as excessively costly, and incomplete compensation
       for accidents encouraged workers to take steps to avoid injuries. If one adopts the
       view that the employers in Lancashire Coal mining who had compensation plans
       were appropriate role models for the rest of British Industry, it also supports the
       current practice of forcing employers to join the workers’ compensation system.
       However, as Epstein stresses, it is far from clear that such plans are appropriate
       for all cases. Indeed, the original Workers’ Compensation act covered only work
       around railroads, mines, quarries, and building construction. The extension of
       compulsory coverage to a wide range of industries, and certainly the extensions in
       Bill 63, appear to be a modern invention.

               As noted above, the fact that some British employers provided for their
       injured employees before the advent of legislation can be viewed as either
       supporting or opposing the current practice of mandating similar compensation
       across the economy. Certainly, we cannot justify current legislation by simply
       asserting that workers were uniformly ill-treated before its passage. However, we
       know very little about how employees who were not covered by an employer-
       specific plan fared. Fortunately, data which was gathered by reformers in the US
       at the turn of the century provide further information on the key question of how
       workers fared when they were reduced to using the courts to achieve
       compensation for on-the-job injuries.


2.2.3 Workers Compensation Under Common Law in the US, 1897-1903
               As in the UK, some large US employers such as Carnegie Steel and
       International Harvester developed private workers’ compensation schemes. A
       recent study by Cantor and Fishback (1995) provides insights into the experiences
       of employees who were not so covered but instead had to rely on the tort system
       and private settlements to secure compensation.

               Table 1 of Cantor and Fishback (1995) summarises evidence from five
       jurisdictions just prior to the adoption of workers’ compensation legislation
       (Missouri and Michigan in 1910, Illinois in 1908, Minnesota in 1909-10, and New
       York in 1907). On average, employees who litigated claims for injuries on the job
       received benefits more than 60% of the time, and compensation covered 70% of
       average losses. Taking both the probability of success and the expected payout
       into account, an employee who litigated a claim could expect to receive
       approximately one year’s earnings. While this represents a substantial sum, we
       of course have no way of knowing the underlying severity of the claim and who
       was primarily at fault. What is clear is that many employees were able to secure
       relatively large compensation amounts even in the absence of any special
       provision by the state or by their employer. On the other hand, there may well
       have been cases where legitimate claims were set aside and others which were not
       litigated at all.

               Some further insight can be gained from Cantor and Fishback’s (1995)


Garvey and Giammarino                      16
       detailed analysis of court cases and settlements in Michigan from 1897-1903.
       They find that, similar to modern practice, the great bulk of claims were settled
       out of court. They also found a worker was more likely to receive wage benefits
       during disability from such settlements when the worker was more experienced
       and more severely injured. Benefits were slightly more likely when the employer
       carried insurance, contrary to the view at the time that insurance firms were
       particularly likely to deny benefits to workers. Cantor and Fishback (1995) were
       unable to locate any additional patterns in the settlement data despite the fact that
       they had detailed information about both employer and employee. This can be
       seen as evidence against systematic discrimination by employers and the courts,
       especially since the payment of benefits was unrelated to whether the worker was
       old or young, married or single, and whether or not he/she was a native-born
       American. As Cantor and Fishback (1995) point out, these findings contrast
       strongly with assertions made by advocates of workers’ compensation at the time.

2.5 CONCLUSIONS

               Working from basic economic principles and available evidence, this
       chapter suggests that serious consideration be given to allowing some employers
       to opt-out of the workers’ compensation system. It is unlikely that any single
       compensation arrangement could be ideal for all employers, and voluntary
       participation provides the WCB with both credible information on how well they
       are doing and an incentive to respond. Given current practice, and the recent
       extensions of Bill 63, this may seem a radical or even retrograde suggestion. It is
       worth noting, however, that the goal of 100% compulsory coverage can also be
       seen as radical in a democratic society with a mixed or market economy.
       Certainly, the experience of Texas does not support the view that workers are
       seriously disadvantaged in firms which exercise the opt-out choice. Moreover, as
       noted in this chapter and more thoroughly in the next one, generous workers’
       compensation is not simply a gift to employees. There is evidence that workers
       effectively pay for workers compensation in the form of lower wages and, more
       tragically, in the form of youth unemployment.

              This chapter has addressed the broad and generally neglected question of
       why and in which cases the “historic compromise” of 1913, in which employers
       were to finance the system and employees would forego the right to sue, should
       be extended to parties who had no opportunity to disagree with the proposition.
       In the next two chapters we deal with the operation of the two sides of the
       compromise. The next chapter presents a brief analysis of the benefits provided
       to workers, and the subsequent chapter provides a more detailed study of the
       charges levied against employers under the system.




Garvey and Giammarino                       17
                                    CHAPTER THREE

                  FIXED COMPENSATION SCHEDULES
                   One of the defining features of workers’ compensation in BC is that
        employees’ benefits are fixed to cover medical expenses plus a fraction of salary
        (currently 75%) provided that this salary is less than the maximum (currently
        $55,800). This formula involves two major policy choices. First, compensation
        is the same for two employees of equal earnings, (1) regardless of how much they
        may differ in their underlying needs for insurance against workplace accidents,
        and (2) regardless of whether or not the employee contributed to the injury.
        Second, compensation is less than 100% of lost earnings. These two features are
        not unique to BC and indeed are ubiquitous throughout the industrialised
        countries studied by Perrin Thorau and Associates (1997).

                    As discussed by Epstein (1982), fixed, no-fault compensation
        provisions have some appealing features. First, they circumvent the intractable
        problems of assigning fault for accidents and instead take the position that gainful
        employment inevitably involves risks. Second, by providing less than full salary
        to an employee in the event of an accident, they preserve some incentive for the
        employee to take care.7 While the WCB experiences a substantial amount of
        controversy and expense in the process of appealing individual cases, it is easy to
        imagine that these costs would greatly increase if benefits formulae and grounds
        were more complex. Moreover, while we have argued in the previous chapter
        that the WCB compensation formula does not necessarily fit all situations and
        there is a strong case for allowing parties to explicitly opt-out of the system, there
        is little doubt that the basic philosophy of giving up the right to sue in exchange
        for fixed compensation is appropriate in many cases. This impression is
        buttressed by Epstein’s (1982) observation that the plans which British employers
        implemented prior to workers’ compensation legislation                involved fixed
        compensation without regard to fault.

                   This chapter analyzes two important aspects of workers’ compensation
        schedules, drawing primarily on research outside BC. We first ask the basic
        question, who effectively pays for the benefits that injured workers receive? We
        then turn to some important effects of the ways in which benefits are calculated
        and distributed.




7
  An important potential design problem in BC, however, was mentioned in many of the submissions to this
Royal Commission. Specifically, it was noted that payments under workers compensation receive such
favorable tax treatment that employees may receive higher after-tax incomes when they are hurt. This state
of affairs can be addressed by defining benefits on an after-tax basis.


Garvey and Giammarino                              18
         3.1 WHO ACTUALLY PAYS FOR WORKERS’ COMPENSATION
       BENEFITS

                  If we restrict attention to formalities, the answer to the question is
       obvious; the WCB decides on benefit levels and employers pay for these benefits
       through assessments. The original understanding appears to have been that the
       costs would eventually be borne by the consumers of the products. Employers
       were and are specifically forbidden in legislation from deducting any of the costs
       of assessments from workers’ wages. However, one of the most basic principles
       in economics is that the formal incidence of taxes and regulations does not change
       the actual incidence. Put another way, one cannot tell if an employer is indirectly
       deducting the costs of workers compensation from wages unless one knows
       exactly what the wage would be in the absence of the premium. The provision is
       simply unenforceable and, as we shall argue below, this may be fortuitous since
       otherwise employers could only escape the costs by hiring fewer workers.

              The above argument rests on the presumption that employers and
       employees care about the complete package of employee remuneration. More
       generous workers’ compensation arrangements are more costly to the employer
       and should be offset by a reduction in wages and other benefits. There is ample
       evidence, both direct and indirect, that pay does in fact change to reflect the costs
       of both accidents and arrangements to insure against such accidents. The first
       evidence of this effect is in the general area of employee safety. Employees in
       more hazardous lines of work (measured by accident rates at the 3-digit industry
       level and controlling for sex, age, experience, and education) receive not only
       higher wages but also more generous fringe benefits such as pensions and holiday
       pay (Dorsey (1983)).

               The evidence is actually stronger in the case of workers’ compensation
       costs. Gruber and Krueger (1991) estimate that, on average through the US from
       the 1970’s a dollar increases in workers compensation premiums reduced
       workers’ direct salary income by 86 cents. A more recent study by Kaestner
       (1996) finds more dramatic effects. He studies employment and wage data for all
       50 US States and the District of Columbia for the years 1982-89, and
       distinguishes between workers in the age groups 16-19, 20-24, and 25-34. He
       finds that an increase in workers’ compensation costs reduced wages more than
       proportionately, especially for the older age group. One explanation, to which we
       return in the next section, is that increased workers’ compensation involves both a
       direct and an indirect increase in costs. The indirect increase reflects more
       intensive use of the system by employees and, as we shall see, dramatically
       increased medical charges.

              The above studies suggest that wages inevitably adjust to offset workers’
       compensation costs. However, Kaestner’s (1996) sample of younger workers
       contains a large number of employees for whom it is literally true that wages


Garvey and Giammarino                       19
        cannot be reduced to offset the costs, as is meant to be the case in BC. These are
        employees whose wages were already at the legal minimum before any increase
        in compensation costs. Consistent with basic economic theory, he finds that
        increased workers compensation costs had a smaller effect on the wages of young
        workers but a larger effect on their employment. Specifically, he finds that a one
        percent increase in workers’ compensation costs reduced employment in the
        youngest age group by over 1.5%. Put another way, high workers’ compensation
        costs were a contributing factor to the tragedy of youth unemployment in the US.

               The evidence in this section is remarkably consistent.8 Workers’
        compensation benefits at least in the US are not a simple transfer from employers
        to employees. Employees shoulder a large amount of the costs in the form of
        reduced pay and in the case of youth in the form of increased unemployment.
        This is a clear but drastically simplified account of the effects of workers’
        compensation benefits and associated costs. We now turn to some more detailed
        evidence on the effects of workers’ compensation benefits.

            3.2 BENEFIT RATES AND CLAIM COSTS

           Perhaps the best-known and widely-documented fact about workers
        compensation benefits is that claims rates increase when benefits are increased.
        Krueger (1990) summarises a large number of studies which estimate that a one
        percentage point increase in workers’ compensation benefits increases reported
        injuries per 100 workers by between one and two percent. Krueger notes that
        many of these studies fail to control for changes in the mix of employees,
        earnings, and features of workers compensation benefits such as waiting time. He
        finds that when these factors are held constant, the estimated effects fall to just
        under 1 percent. He also finds that the waiting period has an even stronger effect
        on claims rates. Specifically, all US states have an average waiting period of 7
        days or less. Increasing the waiting period from three to seven days is associated
        with a reduction in claims of nearly 40%. Recent work on Canadian jurisdictions
        reported in Thomason and Chaykowski (1995) come to similar conclusions.

               This research has both obvious and indirect implications. The obvious
        implication is that an increase in benefits has a large effect on the costs of
        providing compensation; there is both the direct effect whereby a given injury is
        more costly plus the indirect effect of an increase in the number of injuries to be
        covered.      There are in principle at least two ways in which increased
        compensation benefits could increase the number of claims. The first is that
        reducing the loss to being injured reduces the incentive to take care on the job,
        thereby increasing the accident rate. The second is simply that some workers may
        expend more time and effort in filing fraudulent claims. One intriguing set of
        findings is that men and women employees show markedly different responses.
        Krueger (1990) reports that accident rates for women do not appear to increase in

8
  “Remarkably” because in many other cases such as increases in minimum wages it is extremely difficult
to isolate the effects of government policies; see eg., Card and Krueger (1995).


Garvey and Giammarino                              20
       the benefits rates. Thomason and Chaykowski (1995) report that higher benefits
       increase the duration of claims for women but not for men. Thus, both sexes on
       average respond in ways which increase overall compensation costs, but on quite
       different dimensions.

          3.2 .1 Determinants of Medical Expenses

               The previous section focuses on the generosity of wage replacement
       provisions, summarised by the ratio of benefits to wages. Invariably, workers’
       compensation systems also cover an employees medical expenses. From the
       perspective of an individual employer, both payments are equally important as a
       source of expenses. The WCB is in a similar situation since employer
       assessments are meant to cover the full costs of the system. Drastic increases in
       medical costs present a constant threat to the ability of workers’ compensation
       system to cover its full liabilities. The concern is most pronounced in Ontario,
       whose unfunded liability at present greatly exceeds that of all the other provinces
       combined. These concerns, and the proposed response thereto, are expressed
       recently in two publications from the Ontario Workers Compensation Board
       (1997a, b).    The fundamental contention of the Ontario WCB is that their
       workers’ compensation system currently covers an excessively broad range of
       complaints, and proposes (1997b) to: “restore the integrity of the workers'
       compensation system as a workplace accident insurance plan by precluding
       compensation for chronic occupational stress and limiting entitlement for chronic
       pain.”

               Clearly, reducing the range of coverage will reduce the costs of the
       workers’ compensation system. However, from the perspective of society as a
       whole, it is not clear that medical expenses should be a subject of special concern
       for workers’ compensation. Legitimate medical costs which are not covered by
       workers’ compensation are effectively covered by the ordinary medical system.
       One possible response is simply that that many chronic complaints are not a
       legitimate claim under either medical system. Another is that there is something
       special about work-related injuries.

               There are two reasons why work-related injuries are of special concern.
       The first was mentioned in the previous chapter on whether employers should be
       compelled to contribute to workers’ compensation; if employee medical expenses
       are covered by society at large then dangerous employers are effectively able to
       escape the true costs of their business. This subsidy to dangerous employers
       leads to reduced safety efforts and an actual disadvantage to safer lines of work.

               Recent health cost experiences in the US has presented an additional set of
       concerns. A 1990 study of health costs by the Minnesota Department of Labor
       and Industry (summarised in Baker and Krueger, 1993) found that expenditures
       for workers’ compensation injuries were more than twice those for similar injuries
       treated outside the system. These results cannot be explained away by any
       obvious difference in the nature of injuries treated, since the study included a


Garvey and Giammarino                      21
       nearly identical control group of claims covered by a major private insurer in the
       state. The apparent “overcharging” of workers’ compensation claims led some to
       call for 24-hour health care coverage which would effectively do away with the
       health benefit portion of workers’ compensation.

               Durbin, Corro, and Helvacian (1996) provide an even more detailed study
       from 1988-91 for the 15 largest insurers in Florida, Illinois, Oregon, and
       Pennsylvania pairing workers compensation claims with a control of group health
       claims. This study replicates the Minnesota finding that workers’ compensation
       claims are at least twice as costly. They also examine potential reasons for the
       cost difference. No difference is found in the charges levied for equivalent care,
       that is, physicians did not opportunistically price-discriminate against workers’
       compensation cases. Rather, patients covered by workers compensation used
       services more intensively and chose a more expensive mix of medical care
       providers. These findings make it clear that changing to 24-hour coverage will
       not simply reduce overall medical expenses. There would also be a change in
       the mix of care provided.         Whether such changes would be desirable is
       impossible to say since it is always possible that work-related injuries are
       inherently more severe or at least fundamentally different from those that occur
       off the job. What is clear is that it matters whether or not an injury is covered by
       workers’ compensation.



          3.3 CONCLUSIONS

               The previous chapter examined the question of whether or participation in
       workers’ compensation should be mandatory. In this chapter, we take
       participation in the system for granted and examine the value of providing fixed,
       no-fault compensation as part of this system. In addition, we note that employees
       do respond quite strongly to the ways in which benefits are calculated and
       changed. More generous wage benefits appear to significantly increase the
       number of claims, thereby drastically increasing the “bottom-line” impact of such
       changes. Furthermore, medical conditions that are covered by workers’
       compensation appear to be far more costly than similar conditions which are not
       covered. Thus, the overall cost of the system, which we have argued is ultimately
       borne by employees to a large degree, is extremely sensitive to changes in the
       benefits offered by the system.




Garvey and Giammarino                       22
                  CHAPTER FOUR
    THE ECONOMICS OF THE WCB ASSESSMENT SYSTEM
                Workers’ compensation is designed to be funded entirely by employer
         contributions, called “assessments”. While previous chapters have presented
         substantial evidence that workers do in fact bear at least part of the costs in the
         form of reduced wages and fringe benefits, payments to the scheme are formally
         assessed on employers. This is important not because the formal incidence of
         taxation matters, but because the actual assessment rates are determined by
         attributes of the employer rather than those of employees. In this section we
         investigate the effects of the way in which in practice the WCB assesses costs on
         individual employers.

                In the pure “collective” form of workers’ compensation, individual
         employer attributes are ignored altogether; all pay the same rates. This extreme
         approach was modified in the original Workmen’s Compensation Act of 1917 by
         placing employers into 12 groups for the purposes of assessment, and as of 1992
         there were 64 such categories, termed “subgroups” (Hunt, 1992, p.43). A further
         departure from pure collective employer liability was introduced in 1986 with the
         ERA (Experience Rated Assessments) system. This system, which is common
         throughout the world, assesses different amounts to employers in the same
         subgroup based on their own accident rate compared to the subgroup average.9
         This chapter seeks to shed light on the costs and benefits of charging differential
         assessment rates to employers based both on the observable attributes which
         determine their subgroup, and on their own claims experience.

            While in principle and in practice, there are many ways to distinguish among
         employers, the three relevant features for Workers’ Compensation are the nature
         of the production process (captured, albeit imperfectly, by industry classification),
         outcomes (actual accident rates and claims costs), and behaviours (equipment and
         work practices aimed at increasing safety). We follow the approach of the WCB
         Internal Audit and Evaluation (1995) by focusing on the first two features, and
         also in using accident outcomes to judge the performance of the system. It has
         been noted that in BC there is the potential for a conflict between the regulation of
         outcomes versus behaviours; an employer could have a merit experience rating
         and be in violation of specific safety rules or vice versa.10 This is because BC
         experience ratings are just that; reflective of the actual accident experience of the
         employer. Recently, Alberta, Ontario, and Manitoba have begun linking
         assessment rates to specific safety guidelines monitored by on-site audits as well
         as accident rates. Hunt, Barth, and Leahy (1996, p. 216) characterise the
         introduction of behaviours as well as outcomes into assessment as “the current

9
  These groupings have never been absolute, since (a) assessment rates are moved gradually and fluctuate
less than costs; and (b) unfunded liabilities for the entire system have emerged at times. Thus, there are
cross-subsidies between safe and unsafe subgroups, and over time.
10
   Hunt (1992), page xiii.


Garvey and Giammarino                                23
       frontier in experience rating in Canada” and suggest that BC has fallen behind.
       Similarly, a submission to the current Royal Commission by the BC Provincial
       Council of Carpenters suggests that the ERA “be further modified away from
       being tied to the cost of compensation claims and towards an assessment based on
       employer's injury rates and diligence in having proper safety processes at the
       work site” (p. 3).

               One could equally well argue, however, that WCB has been wise to avoid
       extensive and potentially misguided application of additional specific safety
       standards. Simply put, it is often difficult for outside regulators to adequately
       define “proper safety processes”, much less enforce compliance with such
       standards. Viscusi (1986) argues that experience rating in the US is far more cost-
       effective than the specific standards required by the US Occupational Safety and
       Health Administration. Similarly, Hyatt’s (1997) study of experience rating in
       BC finds that inspections and safety orders are unrelated to accident rates and
       costs. Of course, it is possible that employers who were subject to safety orders
       would have been even more unsafe in the absence of such orders. We will have
       little more to say about such matters because they involve technical and industry-
       specific knowledge of which observable actions and practices are most likely to
       contribute to safety. We focus instead on the design and effects of current and
       alternative assessment systems.

              Our focus on outcomes tends to de-emphasise explicit treatment of equity
       issues. We believe this de-emphasis is warranted, because different notions of
       equity can be used to argue for or against any assessment system, from pure
       collective liability to individual employer responsibility. For example, the
       originators of workers’ compensation in BC evidently viewed equity in terms of
       equal treatment for employers regardless of their safety record, a view that
       equates risk-sharing with equity. By contrast, the Phase II Statistical Profile of the
       WCB’s Internal Audit and Evaluation Study on pages 17-18 effectively maintains
       that equity requires employers to pay according to their accident rates, a view that
       equates equity with individual responsibility and incentives.

              Finally, consistent with our agreement with the WCB, we leave aside the
       question of whether overall employer assessments are sufficient to cover future
       accident costs. While this is an important issue (for instance, elimination of the
       shortfall has been the overriding concern in Ontario’s recent reforms) it may not
       be the key issue for BC. At least compared to other provinces, BC does not
       appear to have a large unfunded liability or exorbitantly high overall assessment
       rates at this time (Thomason and Chaykowski, 1995; Hunt, Barth and Leahy,
       1996). This does not imply that the BC assessment system is ideal. Just as a
       government’s budget can be balanced despite distortionary taxes and wasteful
       expenditures, so a solvent Workers’ compensation system does not imply that
       good employers are not penalised and unsafe work practices are positively
       encouraged.




Garvey and Giammarino                       24
              The remainder of this chapter is organised as follows. Section 1 presents a
       brief description of the BC assessment system, focusing on the two ways in which
       employers may either bear or avoid the costs of accidents; (1) assessments based
       on subclasses which share the employer’s safety characteristics; and (2)
       assessments based on the employer’s own safety experiences. Section 2
       summarises existing research on the effects of such systems, which has been
       focused almost exclusively on experience rating, henceforth ERA. Section 4
       presents some new evidence on how subclass formation as well as ERA affects
       accident rates, and section 5 concludes.

4.1 A SKETCH OF THE CURRENT SYSTEM


4.1.1 Placement of Employers into Subgroups
             The most fundamental determinant of an employer’s contributions to the
      WCB accident fund is placement into subgroups. Each subgroup is assessed a
      different payment rate expressed as a fraction of payroll. In principle, each
      subgroup should be self-financing so that low-accident industries and businesses
      do not bear the costs of other less safe pursuits. This has not been the case in BC.
      Table C.16 of the Internal Audit and Evaluation Study, Statistical Profile Phase II
      (1996), clearly shows that some subclasses pay more than their accident costs
      every year, while others pay less.

               Assessment rates also differ greatly across subclasses and a trade article in
       the Vancouver Sun (1997) encourages employers to pay careful attention to their
       WCB classification. Subgroup classification will inevitably be contentions
       because (1) assessment rates differ more than 3-fold between subgroups and (2)
       there is no objective scientific criterion for classification. Hunt (1992, pp. 22-26)
       makes it clear that many employers are dissatisfied with the process and outcomes
       of subgroup classification. Doubtless this is in part due to a natural drive for
       lower costs, and to a businessperson’s appreciation of the subtle differences
       between his/her own organisation and others which may appear similar to
       outsiders. But even an outsider quickly detects substantial heterogeneity within
       subclasses, in terms of such obvious factors as firm size and industry membership.
       Some BC assessment subgroups are small and relatively homogeneous while
       others are large and diverse. For example, stevedoring, subgroup 902 in 1992,
       contained only 32 firms all of whom were in the same industry while
       miscellaneous manufacturing and services, subgroup 602, contained over 1500
       firms in 40 industries. Moreover, Hunt, Barth and Leahy (1996) report the
       controversial case of class 621, Retail Stores, in which a single identifiable group
       of employers was responsible for essentially all the sub-group deficit. We will
       examine both the overall performance effects of fine versus coarse subgroupings,
       and the way such groupings affect outcomes produced by experience rating
       system.




Garvey and Giammarino                       25
4.1.2 Experience Rated Assessments (ERA)
               Since 1986, assessment rates for many employers have been based on their
       own accident record and costs as well as that of their subgroup. Smaller
       employers do not participate as fully in this experience rating system (ERA), a
       fact which we account for in our study below. The performance measure used in
       ERA is firm claim costs per dollar of payroll, which is the same unit in which
       base assessments are made. In a pure individual liability system, each employer’s
       assessment would be based on her most recent claims experience. Clearly,
       however, yearly accident costs can and do fluctuate quite widely for reasons out
       of the employer or her employees’ control. An ideal incentive system would
       isolate that part of accident experiences which are due to employer policy and
       safety decisions. In our view, the actual ERA method is if anything too careful to
       dampen “fluctuations” in accident experiences and thus errs on the side of
       insurance as opposed to safety incentives. In a technical appendix to this report
       we present an explicit formal model of the incentive effects of the WCB
       assessment scheme, including ERA.

           There are three basic ways in which ERA departs from a pure experience
       rating system. The first is that ratings are based on deviations from the subclass
       average injury cost experience. Thus, fluctuations that affect the entire subclass
       are netted out from ERA. Firms which have less than average accident costs
       receive “merit” reductions and firms which exceed the average are in “demerit”.
       That such “benchmarking” can improve an incentive scheme is well-known (see
       for example, Lazear and Rosen, 1981). The second way in which ERA reduces
       fluctuations in employer assessments is by using two years of past accident
       experience, so that assessment rates for 1997 will be based on employers’
       experience from 1994-1995. Similarly, long-term disabilities (over 30 months)
       are excluded from ERA in an attempt to filter out the effects of actions taken in
       the relatively distant past. Finally, new participants in ERA must accumulate
       more than one year of experience before it can affect their ratings.

           The above measures serve to dampen the incentive effects of ERA but also
       provide valuable insurance against uncontrollable fluctuations. ERA also places
       strict caps on the use of merit and demerit ratings in determining assessments.
       No matter how low (high) a firm’s two year claims cost experience may be, its
       assessment will never be adjusted more than 33% below (above) the subclass
       average. Such “capping” of the ERA system serves in part to smooth employers
       assessments over time. Phase I of the ERA Evaluation study (1995) shows that
       between 1 and 2% of employers in most subclasses actually “swing” between
       maximum merit to maximum demerit each year. While such violent swings may
       reflect employer decisions to some extent in some cases, they may also be
       primarily due to chance. However, Phase II of the Study (1996) shows that
       capping ERA at 33% involves serious cross-subsidies between employers.
       Remarkably, over 1/4 of the firms that are classified as maximum demerit (33%
       above industry average) are in fact more than 900% below the industry average.
       Moreover, they document a subset of employers who had been continuously at the
       maximum merit and maximum demerit for the entire period of the study, 1988-92.


Garvey and Giammarino                      26
       Clearly, many subclasses contain employers who are persistently and significantly
       accident-prone as well as those who are always safer than the average. Concern
       with such long-term imbalances led Canadian Forest Products to recommend in
       its submission to this Royal Commission that it “expand ERA to provide for a
       maximum 50% merit and a maximum 200% demerit”.

               While capping of the ERA system leads to some anomalous outcomes,
       change may be difficult for the simple reason that substantial cross-subsidies are
       consistent with the 1913 Meredith Commission’s contention that “..it is important
       that the small employer should not be ruined by having to pay compensation”
       (Workers Compensation Board, 1996). It is clear, however, that this principle
       comes at a cost. Basic economic principles suggest that firms that cannot cover
       their costs should go out of business since their continued operation reduces social
       welfare. Accidents and injuries are certainly a part of the true social cost of a
       business endeavour, and the capping of ERA effectively subsidises dangerous
       employers. It also places greater stress on the problems of classifying employers
       into subgroups since the safer ones provide subsidies to the more accident-prone.
       Indeed, ERA as currently practised has only a small effect on employers’
       assessment costs compared to subclass assignment. ERA does not allow an
       employers assessment to deviate more than 33% from the average. To get a sense
       of the importance of intra-subclass subsidies, consider first subclass 657, which
       contains firms engaged in lumber-yard style businesses. One of the more
       dangerous pursuits in this subclass is travelling wood-saws. The Internal Audit
       and Evaluation Study (1996), Table C.16 reports that in 1992, employers in this
       business paid less than one-fifth of their claims costs in assessment, with similar
       outcomes in 1991. At the other end of the lumber-yard spectrum, firms in the
       manufacturing of sawdust logs and firewood paid over four times their claims
       costs as assessments in 1992. Subclass 107, the manufacture of doors and beams,
       provides a similar example. Manufacturers of Box Works and Wooden Pallets
       paid only one-fourth of their accident costs as assessments while firms producing
       Wooden pails, tubs, and barrels paid over twice their accident costs. The above
       examples are certainly evocative, but inevitably paint an incomplete picture
       because accident rates fluctuate over time. It is always possible that in the next
       decade or so the cost/assessment picture will be reversed and that those firms
       which are currently subsidised will return the favour. Whether or not this turns
       out to be the case depends on the two incentive and selection effects of
       assessments to which we now turn.

4.2 EXISTING RESEARCH ON THE INCENTIVE EFFECTS OF EMPLOYER
     ASSESSMENT


4.2.1 Research to 1996
          The incentive effects of any assessment system operate through two distinct
       channels: (1) accident-prevention efforts of existing firms; and (2) entry and exit
       of firms with different safety characteristics. In both cases, we can expect more



Garvey and Giammarino                       27
       desirable outcomes when employers bear, directly or indirectly, the costs and
       benefits that their decisions impose. Thus, systems which impose accident costs
       on other employers or on the rest of the community fail to encourage existing
       employers to take sufficient care (channel 1); and allow some fundamentally
       unsafe firms to continue operation long after they should have been displaced by
       safer rivals or new businesses (channel 2).

                The above effects are broad, pervasive, and very difficult to quantify.
       Existing empirical research on the incentive effects of assessment system has
       inevitably focused much more narrowly. No one, to our knowledge, has
       systematically studied the effects of workers compensation assessments on the
       entry and exit of firms. While the continuous demerit firms documented in the
       Phase II Statistical study suggests some adverse effects, we do not know what
       would happen if ERA limits were lifted or if subclass definitions were narrowed
       so that merit and demerit firms were more carefully distinguished. The problem
       is that exit and entry are major discrete decisions that likely reflect more than just
       workers’ compensation related factors. The incremental effect of accident costs
       would be almost impossible to identify. Empirical research has instead focused
       on channel 1, the effect of assessment systems on the claims costs of existing
       firms.

               While in principle there are many ways in which assessment systems
       affect employers’ safety incentives, existing research has focused on the effects of
       various approaches to experience rating. Experience rating is an obvious attempt
       to provide individual employer incentives, and has also attracted the most
       controversy. In the next section we seek to provide some evidence on the
       incentive effects of the classification system as well as that of ERA.

               In our view, the Internal Audit and Evaluation Study document
       “Experience with Experience Rating”, section 2-4, provides a thorough, critical,
       and highly satisfactory review of the existing evidence. We concur entirely with
       its conclusion that there is an appreciable body of evidence that ERA does have
       some negative effect on accident rates and costs, but that such evidence is indirect
       and therefore inconclusive. Ruser’s (1985) study exemplifies the achievements
       and shortcomings of existing research. He first replicates the well-known and
       exceedingly important fact that worker claims in the US increase when Workers’
       Compensation Benefits increase. He then argues quite correctly that this effect
       should be attenuated when the employer directly bears some of the associated
       increase in claims costs. The weakness of his and most existing studies lies in the
       next step; there is no clear, precise way to isolate the effects of ERA. Ruser
       (1985) uses firm size to capture the effects of experience rating, and finds that the
       claims are indeed less sensitive to benefits for larger firms. While it is true that
       larger firms in his sample were more likely to be experience rated, it is also true
       that large firms differ from smaller ones in many respects that have nothing to do
       with ERA. For example, they tend to be older, pay higher wages, have lower
       turnover, and grow more slowly (e.g., Krueger and Summers, 1988). All of these



Garvey and Giammarino                       28
       features can be expected to reduce employees’ injury claims and to some extent
       align their interests with those of their employers (e.g., Dickens and Katz 1990).

               The fact that ERA was introduced for a large number of BC firms for the
       first time in 1986 and became effective in 1988 provides a more satisfactory
       laboratory for examining its incentive effects. While human societies never
       present a prefect controlled experiment, the introduction of ERA to previously
       base-rated firms allows us to attribute at least part of observed changes in safety
       to ERA. Hyatt (1997) takes advantage of this fact, and we devote the next
       subsection to his study.


4.2.2 Hyatt’s Study of the Incentive Effects of ERA
          Hyatt’s (1997) study is well-designed to isolate the overall effects of ERA on
       accident rates and costs. He collects detailed data on over 24,000 firms that
       experienced the introduction of ERA and were continuously registered with the
       WCB between 1983-92. His research design is simply to estimate the changes in
       claims rates and claims costs that can be attributed to the introduction of ERA.
       His data control for some key firm-specific factors other than ERA participation
       that should affect claims rates. The most important of these controls are: whether
       or not the firm was in violation of other safety provisions; time; the possible
       savings offered by ERA; total firm employment; and benefits paid by the WCB to
       injured workers.

               Most of these controls had expected effects on claims rates and costs.
       Higher benefits marginally increase claims and larger establishments have fewer
       but more costly accidents. Inspections and orders had no effect on claims or
       costs, indicating either than such measures are ineffective or that they interact in
       more complex ways with other firm characteristics (as if, for example, an
       inspection brings an otherwise laggard firm back up to the average).

               Hyatt’s estimates of the effects of ERA provide a clear and consistent
       picture. Accident rates and costs respond to both the desirable and to the
       (unintended) negative effects of ERA. Specifically, ERA reduces first paid claims
       by nearly 10% or 1.1 claim per 100 workers. But ERA also significantly
       increases long-term disability claims costs. This is, unfortunately, consistent with
       the incentives provided by ERA since claims outside the 30-month window are
       excluded from experience rating calculations.


4.2.2.1 Assessment and Critique
              Hyatt’s (1997) study indicates that ERA is a potentially powerful tool
       which can both improve and reduce workplace safety. He documents both an
       obvious and a relatively subtle change, both of which can quite plausibly be
       ascribed to ERA. There are two major shortcomings with the report, one of which
       we address with additional evidence in the next section and the other which we
       can only evaluate theoretically.


Garvey and Giammarino                       29
                The key issue which neither we nor Hyatt can fully address is the problem
        of “claims suppression” or more generally, “unwarranted employer attention to
        claims”. This concern is raised in Hunt (1992) and in submissions to this Royal
        Commission, in both cases by union representatives. For example, page 17 of the
        BC Provincial Council of Carpenters submission maintains that:               “The
        Experience Rated Assessment program which bases employer assessments on
        claims costs motivates many employers to suppress legitimate compensation
        claims and motivates them to support return to work attempts of injured workers
        that are either premature or unrealistic.”

           We are unable to provide any direct evidence for or against the presence of
        claims suppression, nor are we aware of any systematic research on the issue.11
        Our research, like Hyatt’s, focuses on reported claims and so cannot distinguish
        true safety improvements from unwarranted claims suppression. Hyatt’s (1997)
        documented shift from short to long-term claims provides indirect evidence that
        employers do not simply increase their overall safety efforts. In general, however,
        the clandestine nature of the alleged employer behaviours make them difficult to
        study systematically.

           The first concern is that employers who are subjected to ERA will tale a
        greater interest in appealing and opposing claims to the WCB. Indeed, Hyatt and
        Kralj (1991) find evidence of increased appeals after the introduction of ERA in
        Ontario. It is not at all obvious that this is undesirable. Employers who do not
        bear the costs of increased claims arguably have too little interest in appealing and
        pursuing fraudulent claims which, after all, do impose costs on society at large.
        This is especially true since the costs of such claims will largely be borne by other
        members of a subclass, some of whom are in turn likely to be competitors.
        Employer interest in appealing claims is of concern only to the extent that the
        underlying claims process is either overly costly or biased in some fashion, an
        issue on which the Commission will be receiving expert advice from the task
        force. Since, as the original WCB legislation recognised, the costs of workers’
        compensation (including fraudulent or excessive claims) are in the final analysis
        borne by parties such as the consumers of the products and services produced,
        heightened employer interest in appeals may well be a desirable development.
        This is not to argue that the appeals process should not be as cost-effective and
        neutral as possible, but merely to recognise that use of the process is not in itself
        undesirable.

           It is also alleged (e.g., Hunt, 1992 p.41) that ERA employers may discourage
        the reporting of accidents. The only obvious ill-effect of such behaviour is that it
        reduces the integrity of the WCB database. It does not harm the employee or
        society at large, if the affected worker is aware of his/her rights under the WCB.
        An employee who understands and is willing to exercise his/her rights can only be

11
  Walker (1986) presents an informal survey of the impressions of union affiliates. It confirms that union
representatives believe there is more claims suppression in experience-rated workplaces. Clearly, this does
not constitute reliable evidence of actual claims suppression.


Garvey and Giammarino                               30
       discouraged by a sufficient inducement from the employer, and there is anecdotal
       evidence of some claims being paid privately, just as with minor auto accidents
       not being reported to ICBC. Such private settlements do not seem objectionable
       so long as the employee is at least as well off as he/she would have been under
       WCB coverage. To be sure, the BC Provincial Council of Carpenters’ submission,
       maintains that “The non-union workers generally do not have access to strong
       advocacy support for their compensation claims problems (p.8)”, but they hardly
       constitute a disinterested party in so asserting. The recently observed hostility of
       non-union construction workers to recent proposed legislative changes is equally
       consistent with the conclusion that such workers prefer not to have such “strong
       advocacy support”. The counter-argument that workers have little bargaining
       power in a highly competitive industry like construction is, for this issue, beside
       the point. Employers have ample incentive and ability to take advantage of such
       labour market conditions on more direct dimensions such as wages and work
       rules.

          There is, however, another potential lesson to be gleaned from the notion of
       claims suppression. One obvious reason why employers and employees might
       voluntarily agree to suppress a claim is that they can avoid the costs and time of
       receiving compensation from the WCB to their mutual benefit. Essentially,
       claims suppression constitutes a form of “opting-out” from the otherwise
       compulsory WCB system. We argue elsewhere in this report that the WCB
       should seriously consider allowing such actions to take place more openly. While
       this might (or might not, depending on the quality of alternatives) erode some of
       the WCB’s reach it is not clear that it undermines worker welfare.

          Some of the arguments we have made above will doubtless be controversial.
       One conclusion which we believe will be less so is that little is known about the
       many dimensions on which employers and employees can and should respond to
       the explicit and implicit incentives provided by the overall             workers’
       compensation system. The submission by Canadian Forest Products Ltd. makes
       the point that “Safety officers who lack the knowledge or interpersonal skills
       should be retrained and their progress toward an acceptable standard monitored.
       Inspectors should be have experience in the industry that predominates in their
       area. For example, sawmill inspectors should come from a sawmill background.
       (p. 3)” Another area which merits further study is the devolution of safety
       incentives to the “shop-floor”. We have heard anecdotes of prizes or bonuses
       awarded to workgroups with exemplary safety records in the construction
       industry. Such schemes may have promise when safety is effectively a “team”
       output, that is, when one worker’s carelessness could harm his/her fellows.

               Clearly, there is much to learn about the specific desirable and undesirable
       practices that ERA and the assessment system in general encourage employers to
       undertake. Hyatt presents compelling evidence of changes in injury rates and
       types, but there is much more that might be done with more time and data.
       Hyatt’s study is also incomplete in that it effectively assumes that employers
       respond in much the same way to the incentives in ERA and in assessment more


Garvey and Giammarino                       31
       generally. Specifically, he only allows claims to differ according to the
       employer’s participation rate and the claims rate. While these two variables are
       important determinants of the potential cost savings an employer might reap from
       ERA, there are many other factors which matter. Indeed, the Phase I statistical
       profile stresses that there is wide variation between subclasses in the apparent
       incentive effects of ERA. This, we argue, is to be expected since ERA and
       assessments more generally are fundamentally determined at the subclass level.
       First of all, an employer’s merit or demerit are all defined with respect to the
       subclass average. Equally important, base rates are determined by the experience
       of the overall subclass. The key point, which we formally model in a technical
       appendix, is that cost-reducing incentives are far stronger for an employer whose
       subclass consists of firms which are similar to himself. Any effort or cost an
       employer expends towards reducing accidents has a much smaller expected
       impact on his assessment costs if he is grouped with a large number of
       fundamentally dissimilar employers. The reason is that both the employer’s
       relative rank (which determines their ERA merit or demerit) and the overall
       subclass experience (which determines their future base rate) are much more
       likely to be determined by unrelated economic and behavioural factors. We now
       present some systematic evidence on how and why accident experience differs
       across subclasses, both before and after ERA.



4.3 SOME EVIDENCE ON THE DETERMINANTS OF ACCIDENT RATES AT
     THE SUBCLASS LEVEL.

          Our study will focus on unexamined aspects of the data presented in the 1995
       and 1996 Statistical Profiles. Specifically, we will examine whether the observed
       variation across subclasses in overall claims rates, and the change in such claims
       due to ERA, reflect the incentives provided by the assessment system. We
       propose to test whether the determination of subclasses’ experience rating matters
       by asking whether subclasses which should, in theory, show better safety
       performance do in fact exhibit such performance. Similarly, we ask whether
       subclasses which should show a larger response to ERA do so in fact.

               While our focus on subclasses rather than individual firms as in Hyatt
       (1997) is partially due to time constraints, the fact that both base and experience
       ratings are computed with respect to the subclass average makes the subclass an
       appropriate economic unit to study. Under ERA, a firm’s relevant benchmark is
       not its own past performance but its performance relative to its subclass average.
       Moreover, the base rate from which ERA adjusts is primarily determined by the
       need to cover past and prospective claims of the subclass.

             The fundamental feature of any ERA system is to distinguish between
       employers on two separate dimensions: (a) industry and subclass; and (b)
       accident experience of the individual employer, relative to the subclass average.
       We are not aware of any systematic research on the effects of finer versus coarser


Garvey and Giammarino                      32
       groupings. As with explicit behavioural regulations and standards, making
       appropriate groups requires detailed industry knowledge which we do not possess.
       We will, however, provide some evidence on the effect of fine versus coarse
       grouping, exploiting the fact that the composition of actual subclasses varies from
       small numbers of employers in the same industry to large numbers of employers
       in quite different lines of business.


4.3.1 Variables to be used
Safety Performance Measures

          All our data come from the Phase I and Phase II statistical profiles, which
       provide rich information at the subclass level. Our approach is to ask whether
       accident rates, and changes in these rates over time, differ across subclasses in
       ways which are consistent with the incentive effects of ERA and the assessment
       system in general. We are also able to provide estimates of the quantitative
       effects of some of the critical features of assessments.

          Our definition of safety performance is similar to that used in Tables C-19 and
       C-20 of the Phase I statistical profile, which details first-paid claims per 100
       workers for each subclass over the years 1983-92. We supplement this with more
       recent information on claims rates from 1993-95 provided to us by the Royal
       Commission. Consistent with the very purpose of forming subclasses, there are
       large differences in the accident rates across subclasses which persist over time.
       The average over the entire time-period is just over 8 claims per thousand
       workers, ranging from a high of over 30 to less than 1/3.

           Not only are there large differences between subclasses in overall accident
       rates, the changes around the introduction of ERA also seem to differ greatly.
       Restricting attention to the subclasses which moved from Base to ERA in 1986
       and comparing averages for the years before ERA had an effect (1983-86) and
       after (1988-95), there was an average fall of only .37 claims per year per hundred
       workers, but the maximum safety improvement is a fall of over 3.5 claims and the
       maximum safety deterioration is an increase of over 3.1 claims. These numbers
       reflect absolute changes in accident rates, while the Statistical Profiles focus on
       changes in claims rates as a percentage of the overall subclass average level. The
       latter adjustment presumably reflects the concern that for high-accident subclasses
       a given absolute change is more likely to reflect random influences than actual
       incentive effects. For example, a fall of 1 claim per hundred workers is far less
       “impressive” for a subclass with an average claims rate over 20 than for a
       subclass where the claims rate is 2. While dividing by the subclass average is a
       defensible procedure, it is not ideally suited for indicating overall performance
       effects. A fall of 1 claim per hundred workers is, after all, an improvement in that
       worker’s lot regardless of his/her occupation. Moreover, if a subclass showed an
       accident rate of exactly 20 every year for 10 years and then dropped to 19, we
       should be quite confident that there was a change. A more appropriate adjustment
       to indicate the effects of ERA is to adjust changes for the variability of claims


Garvey and Giammarino                       33
         rates for each subclass. When we define variability as the standard deviation of
         subclass claims rates over the time period, and adjust for degrees of freedom, the
         resulting measure is a “t-statistic” which indicates how confident we can be that
         the change induced by ERA is not just due to chance. The 2 subclasses which we
         are most sure enjoyed a reduction in accident rates ( less than 3% by chance) are
         713 (Floor and Window Cleaning) and 726 (Road Making, Excavation and well
         work). The 2 subclasses which showed the most significant increases in claims
         rates are 811 (Operation and maintenance of bus lines) and 909 (Marine Shipping
         Services). Overall, the effects are modest but equally important there is
         substantial variation across subclasses.

Variables Related to Safety Performance

                 The Statistical Profiles gathered by the WCB summarise many attributes
         of the different subclasses, and this section describes the variables we construct to
         explain why accident rates vary across subclasses. Unless otherwise indicated, all
         variables are for 1991-1992. As we have only this snapshot for these variables,
         our study relies on the assumption that they are reasonably stable over time.

            The most obvious feature which affects safety performance besides subclass is
         firm size. The statistical profile decomposes each subclass into Large (over $100
         million), Medium ($100 to $ 1 million) and Small (below $1 million). On
         average, 2/3 of the firms are small but again the percentage varies from 86% to
         12%. Our expectation is that accident rates will be both higher and less
         responsive to ERA in smaller firms. A measure which is closely related to size is
         the fraction of each subclass which actually participates in ERA. We created a
         variable, Participate, which provides a weighted sum of each subclass’’ total
         participation in ERA. It is calculated as:

            Participate = 1(Fraction of firms that are 100% participating) + 0.5(fraction
         of firms that are 50% participating) + 0.25(fraction of firms that are 25%
         participating).

            It is difficult to differentiate the effects of the size and participation because
         they are so highly correlated, reflecting the fact that larger firms participate in
         ERA to a much greater extent. Indeed, only 15% of the variation in Participate
         can be explained by factors other than size.12

            While firm size may have some power to explain firms’ accident rates, it is
         certainly not a factor that the WCB can control. The other variables we use, by
         contrast, are to a large extent design features of the assessment system. As
         indicated earlier, the fundamental determinant of assessment rates is how firms
         are grouped into subclasses. In terms of safety incentives, the most important
         consideration is the extent to which firms share a subclass with others who are
         fundamentally different from themselves. Firms which are in tightly defined
12
   An Ordinary Least Squares regression of Participate on the fraction of small firms, generated an R2
(fraction of variation explained) of more than 0.85.


Garvey and Giammarino                                34
       subclasses have greater safety incentives because their base rates are determined
       by factors that are more likely to be under their control. While there are many
       ways in which firms differ from one another, the only reliable measure available
       to us is industry classification. To characterize the extent to which a given
       subclass is either tightly defined or, by contrast, comprised of many very different
       firms, we construct what is known as a Herfindahl Index. This is a well-known
       and reliable measure of dispersion (e.g., Carlton, 1989). The Herfindahl index
       ranges between zero and one with zero being the most broad and general subclass
       and one the most tightly defined. Specifically, when there are F firms in a given
       subclass and j = 1 to I industries with nj firms in each industry, the Herfindahl
       index is computed as:
                                2
                     I
                        nj 
           Herf  	  
                  j 1  F 


          To see how the index captures dispersion, consider a subclass with three
       member industries. Suppose the firms are almost all concentrated in one of the
       industries, so that the first industry has 98% of the firms while the other two have
       only 1% apiece. Such a subclass is little different from one with only one
       industry, and the Herfindahl index appropriately takes on the large value of
       0.9654. If by contrast the firms are equally distributed across the three industries
       the Herfindahl index drops to 0.27. The Herfindahl index is a sensible way to
       characterise the ways in which subclasses differ across other dimensions as well.
       Unfortunately, the only reliable measure available to us is by industry. On the
       other hand, the index we calculate is clearly one which can be easily changed
       simply by creating new subclasses or reallocating firms and industries between
       subclasses. In the 55 subclasses where we have adequate data, the average
       Herfindahl index was 0.59 with a minimum of .096 and 8 single-industry
       subclasses which had the maximum value of one. We also experimented with
       other measures of subclass diversity, including the raw number of industries and
       of firms in each subclass. Neither had any effect on accident rates once we
       accounted for the effect of the Herfindahl index.

          As described above, a key feature of the ERA design is the “capping” of merits
       and demerits at 33%. The Statistical Profiles summarise the fraction of firms
       which are below the maximum demerit, and also give some information as to the
       extent to which such firms are below the maximum demerit. 2.4% of firms in the
       average subclass were 900% or more below maximum demerit, 2.2% were
       between 400 and 899%, 2.5% were between 200 and 399%, and 2.7% were from
       100% to 199% below maximum demerit. Clearly, all else equal there should be
       higher accident rates in subclasses which have more such high-risk firms. Also,
       ERA should have a somewhat weaker effect on such subclasses because firms
       which are “deep in demerit” do not realise any reduction in their rates until they
       pierce the 100% barrier.




Garvey and Giammarino                       35
          A weakness with the above measure is that firms do not necessarily remain in
       maximum demerit over our entire period. To help correct for the fact that firms
       may migrate between categories, we also included the fraction of firms which
       “swing” between maximum demerit and merit or vice versa in the years 1991-
       1992. In the average subclass, 1.7% of the firms went from maximum demerit to
       maximum merit and 1.15% swung the other way, but the maximum fraction
       swinging was over 50%.           The effect of such swings on average safety
       performance is not straightforward. On the one hand it is associated with a larger
       fraction of small firms and firms with volatile accident histories which should
       imply higher accident rates and smaller responses to ERA. On the other hand, it
       may represent a sensitivity of accidents to incentives.

          The final variable which we believe should be related to overall accident rates
       and to the response of accidents to ERA, is the extent to which the subclass as a
       whole finances its claims costs each year. In practice, the assessment system does
       not make every subclass self-financing; some pay far more than their claims costs
       and some pay far less. The measure which is both most reasonable to us, and
       which turns out to have the strongest relationship to accident rates, is the absolute
       subsidy rate, computed as:

                            Claims Costs
           Asubsidy  1 

                             Assessment

           The reason we focus on absolute values is that, in terms of incentives, there is
       little difference between a subclass which pays the costs of other subclasses and
       one which is able to shift its costs onto other subclasses. In both cases, incentives
       to reduce one’s own accident rates are reduced. For this same reason, however,
       the marginal effect of ERA on safety should be greater for subclasses which have
       such diluted incentives from the base rating system. The average subsidy is quite
       substantial, over 40%. While many subclasses had no subsidy at all, there were a
       few with subsidy rates over 250%. The subsidy measure is positively correlated
       with our Herfindahl index (Pearson correlation coefficient = 0.24), so that more
       homogeneous subclasses are more likely to pay their full costs (and less likely to
       pay more than their full costs). It is also positively correlated with the fraction of
       firms at maximum demerit, (Pearson correlation coefficient = 0.24), presumably
       reflecting the fact that such firms are carried along by the system at large as well
       as by their subclass fellows.

          Finally we include a “dummy” variable which takes on the value one for each
       of the firm categories (Forestry, Construction, and Voluntary). This is done to
       account for the fact that these subclasses were not purely base-rated before the
       introduction of ERA.


4.3.3 Results: The Determinants of Accident Rates
         Our research design highlights the overall incentive effects of the BC workers
       compensation assessment system, including but not restricted to ERA. As argued


Garvey and Giammarino                       36
        above, the fact that subclass base rates are adjusted at least in the direction of
        actual claims costs has always provided firms with some incentive to control
        costs. ERA is therefore an additional incentive system.

           Presumably, the key overall question for the Commission is how to design or
        alter the assessment system for the future. We will accordingly focus on the
        determinants of accident rates in the most recent years, after the introduction of
        ERA. We measure subclass average first claims rates averaged over the longest
        available post-ERA period (1988-1995) to reduce yearly noise. Nonetheless, our
        results change little if we use shorter time-windows.

           Our results are reported in Table 1. They are all Ordinary Least Squares
        estimates with first paid claims rates of each subclass as the dependent variable.
        Results are similar with alternative estimation procedures and functional forms.13
        As indicated in the previous section, annual claims rates are more variable for
        some subclasses than for others. In statistical terms, this indicates the presence of
        heteroskedasticity, for which we correct by computing standard errors according
        to the standard White (1980) approach. To economise on the presentation of
        “non-results” as well as our major findings, we present a specification which
        includes variables which do not turn out to be important. Since there was reason
        to believe they should be important, non-results are also worthy of attention.

           The first column of numbers in Table 1 shows the relationship between first-
        paid claims in each subclass over 1988-95, and various attributes of the subclass
        in 1991-1992. Some of the stronger results are quite intuitive. The voluntary
        sector has lower claims rates, presumably because these are professional practices
        and other primarily “white-collar” lines of business.14           Also, as expected,
        subclasses with a preponderance of small firms tend to exhibit higher claims rates.
        Since ERA is less likely to apply to such subclasses, this may also be taken as
        evidence of the incentive effects of ERA. However, our direct measure of
        participation in ERA has an insignificant but positive effect on claims rates. Since
        the two variables (ERA participation and size) are so strongly related to one
        another, our interpretation is that it is simply difficult to reliably disentangle the
        effects of ERA participation and raw size.

                As predicted, a larger Herfindahl index, indicating a more homogeneous
        and tightly defined subclass by industry, is associated with lower claims rates. As
        we have argued above, and will soon buttress with additional evidence, this
        should not be taken as a pure ERA effect. That is, we do not interpret the result as
        indicating that ERA is more effective in tightly defined industries. Rather, we
        will see that a tighter definition of subclasses has a positive effect on safety
        regardless of the presence of ERA. Whatever the source of the effect, it is quite

13
   A Logit estimate which breaks up the sample at the mean of the dependent variable produces
qualitatively similar results, as does a “Box-Cox” estimate which allows the functional form of the
variables to be determined simultaneously with the regression coefficients.
14
   As described in Chapter 2 of this report, participation became compulsory in 1993-94. We do not have
sufficient data to determine the effects of this change.


Garvey and Giammarino                              37
        large. For example, decomposing a subclass with two industries of equal
        importance into its constituent industries would increase its Herfindahl index from
        (0.5)2+(0.5)2 = 0.5 to 1. According to our estimated coefficient on this index,
        such a change would reduce the accident rates for the average firm by nearly
        30%, from 8.1 to 5.8 claims per hundred workers. It is also instructive to note
        that subclasses which are more heterogeneous by industry (and thus have a low
        Herfindahl index) also tend to be the ones where some employers are below
        maximum demerit. For example, the correlation between the Herfindahl index
        and the fraction of employers who are more than 900% below maximum demerit
        is 0.25.

           While our Herfindahl index has the predicted effects on accident rates, some
        other results are puzzling, at least to us. First of all, among the four classes of
        firms which are below maximum demerit, only the relatively mild cases, from 199
        to 100% below maximum demerit, have any appreciable effect on increasing
        subclass accident rates. The fraction of firms which are further below maximum
        demerit have essentially no effect. As mentioned above, it is also surprising that
        subclasses which participate more fully in ERA have, if anything, higher accident
        rates. The effect is not, however, statistically significant and changes sign when
        we omit the fraction of small firms as an additional explanatory variable.

           Our first column of results, summarised above, are illuminating but clearly do
        not present a complete picture. Approximately 2/3 of the difference in claims
        rates across subclasses is driven by factors we do not capture, and it is even
        possible that some unobserved attributes of subclasses are driving the
        relationships we do observe. In the second column of results, we include the
        average claims rate in the four years before ERA was in effect. This controls for
        all durable features of each subclass. One result follows immediately from the
        fact that the coefficient on the four years pre-ERA is nearly one and is highly
        significant; ERA did not fundamentally change the underlying determinants of
        accident rates. This impression is strengthened by the fact that over 95% of the
        variation in claims rates across subclasses in 1988-95 is explained by the average
        claims rates from 1983-86 alone.15 Not surprisingly, many of our other results
        change or fail to hold when past accident rates are accounted for. In essence, we
        are now isolating the incremental effect of these variables on the changes
        associated with ERA rather than with overall accident rates. The most striking
        and surprising result, is that participation in ERA now shows a positive and highly
        significant effect on claims rates. We have no satisfactory explanation for this
        result, except to note that again participation is highly correlated with size and
        small firms continue to exhibit higher accident rates even when we control for
        past claims.

           There are two other striking changes in the results between column one and
        column two. First, subclasses which paid either more or less than their claims

15
   This result is obtained by regressing Average First Paid Claims in 1988-95 on Average First Paid Claims
in 1983-86.


Garvey and Giammarino                               38
       costs received what we term an Absolute Subsidy. The presence of ERA appears
       to have improved safety incentives for such subclasses.        This result is not
       surprising once we recognise that subsidies between subclasses reduce safety
       incentives in the absence of ERA. For such firms the introduction of ERA could
       have represented a stronger incremental change in safety incentives; exact
       conditions for this effect are provided in the technical appendix. Intuitively, the
       idea is that firms bearing or receiving absolute subsidies were to some extent
       performing further below their potential safety levels before ERA, and so would
       be expected to respond more dramatically to its introduction

          A similar interpretation applies to the fact that the Herfindahl index shows
       essentially no effect on accident rates once the past average is taken into account.
       Indeed, it appears that more tightly defined subclasses perform worse in terms of
       safety. In the final column of table 1, we show that this is not the correct
       interpretation. The reason the Herfindahl index shows a positive effect on claims
       rates in the second column is that it had an even stronger safety effect before the
       introduction of ERA. This is to be expected since the incentive effects of ERA
       should to some degree compensate for the weakened incentives experienced by
       firms in loosely defined, heterogeneous subclasses.

4..4 CONCLUSIONS

               When combined with past work, our research suggests that the overall
       assessment system definitely has an effect on employer safety incentives. ERA
       clearly has some effect, and it is encouraging that the estimated effects are largest
       in the study (Hyatt) which uses the most detailed data. Nonetheless, we believe
       the effects of ERA are attenuated by two key features, also highlighted by the
       Internal Audit and Evaluation. First, employers often have limited knowledge or
       understanding of ERA. This observation is surely related to the fact that ERA is
       not allowed to have its full effect on actual assessment rates but is instead capped
       at 33% above or below the subclass base rates. Such attenuation is desirable only
       if concerns about “claims suppression” are paramount. We provide no new
       evidence on this factor, but would encourage a more systematic and critical
       evaluation of such concerns.

              It is also not surprising in this light that classification and its effect on an
       employer’s base rates are of far greater importance to employers. Not only do
       rates differ widely between subclasses, as indicted in section 2.2 there is
       substantial cross-subsidy within subgroups. Such subsidies reflect both the
       grouping of quite different employers, and the limited effect of ERA in tying
       assessment rates to individual employers’ claims history. Consistent with this
       view, we find that more homogeneous subgroups exhibit better safety
       performance both before and after the introduction of ERA.

         These results serve to highlight perhaps the key trade-off in the design of the
       overall assessment system. Increasing the number of subclasses would inevitably
       compromise the objective of similar assessments for all employers; high-risk


Garvey and Giammarino                        39
         industries would pay higher rates. But the objective of “equal treatment” for very
         different employers, like any system of cross-subsidies, reduces the safety
         incentive of all employers. Moreover, inherently high-risk employers are
         subsidised and remain in business when they would otherwise perish, while safer
         employers are discouraged from entering and expanding. Given that the WCB
         can in fact identify separate industries within many subclasses, serious thought
         should be given to breaking them up into additional subclasses.16 Hunt, Barth and
         Leahy (1996) maintain that the cross-subsidisation involved in the current
         classification system reduce its “perceived fairness” (p 259). As we have argued
         at the outset, various fairness arguments can be used to justify or to condemn the
         current system. Our results suggest the much more tangible and objective
         problem of increased claims rates.




16
  While there may, in principle, be a concern of “actuarial integrity” with smaller subclasses, we should
note that in practice assessments have allowed subclasses which perform better to subsidize badly-
performing subclasses. While such policies may be prudent from the perspective of the overall system, it is
important that such subsidies reflect insurance rather than subsidies, that is, that they tend to add up to zero
over time.


Garvey and Giammarino                                 40
                   CHAPTER FIVE
        GOVERNANCE AND CONFLICTS OF INTEREST
5.1 GOVERNANCE AND THE HETEROGENEITY OF CLAIMS ON AN
    ORGANIZATION

                     Economic theory views an organization as “a nexus of contracts”, a
       legal fiction created as a clearing house for implicit or explicit contracts by a
       group that chooses to deal with each other in pursuing common goals. In this way
       an organization is an intermediary. For instance, rather than have customers
       contract directly with workers, suppliers, and government for the inputs needed to
       provide goods and services, they contract with a firm for delivery of the final
       product. A business firm consists of a set of contracts written with suppliers of
       various inputs to the production process. Workers, including entrepreneurs and
       managers, contract with the firm for the supply of services in exchange for
       remuneration. Investors’ contract with the firm for the supply of capital in
       exchange for a future stream on income. Society as a whole effectively contracts
       with the firm to establish the common property that can be used and establishes
       standards that will be followed. For instance, fresh air is a common property that
       firms are allowed to use even if doing so results in pollution. This allows those
       who work for the firm to provide a livelihood for themselves and, in exchange,
       they pay taxes and agree to abide by standards set out by government including
       pollution standards.

                      The goals of those contracting with each other through an
       organization will inevitably conflict to some extent. Higher wages for some, for
       example, will mean lower wages for others and/or higher product prices; less
       expenditure on pollution abatement will mean higher wages and/or returns to
       capital but greater pollution of a common resource; greater expenditure on
       accident prevention can mean lower returns to capital but higher return from work
       and lower social costs for medical treatment, rehabilitation, and social disruption.
       Therefore, decisions made on behalf of the organization will often make one
       participant better off relative to some other participant and participants will take
       actions to influence those decisions. Sometimes those actions will be disruptive to
       the organization as a whole.

                     Since organizations are legal entities, there is no one person or group
       of people that have exclusive title to decision making power. Instead, the ability to
       make decisions for an organization is delegated to an individual or group of
       individuals by the other participants. The process that determines which
       individuals will be given the power to make decisions and the limits on the
       authority that they have is commonly referred to as the governance system.




Garvey and Giammarino                       41
                          The degree to which conflicts develop among members of an
           organization and the efficiency with which conflicts are resolved is crucially
           dependent on the governance system. Decisions must be made that determine how
           well one group does relative to another. The resulting conflicts of interest will
           ultimately be disruptive if one or more groups participating in the organization
           feels that their interests have not been given enough attention by decision makers.
           Hence, providing input to all ‘stakeholders’ helps alleviate these concerns. At the
           same time, however, representation of narrow interests may inhibit the ability of
           the organization to achieve broader goals.

                          A governance system is not typically a static set of rules that apply
           in all circumstances. As circumstances change due to either specific decisions or
           changes in socio/economic conditions, some conflicts of interest intensify under
           one set of governance rules and another set of rules may be introduced. For
           instance, the governance of a corporation may change as its solvency changes.
           When solvent, suppliers of equity capital are allowed to elect the primary decision
           makers in a corporation. The conflicts between suppliers of debt and equity are
           usually not large for a solvent firm while the conflicts between equity holders and,
           for example, employees may be relatively larger. When a firm is insolvent,
           however, the conflict between equity holders and debt holders intensifies and
           becomes potentially disruptive. To deal with this, society has established
           debtor/creditor law and bankruptcy law. These laws allow for a change in
           governance to accompany the change in circumstances. In other cases, the change
           is more unanticipated and changes in charters or legislation may be needed. This
           is illustrated by the governance changes made to the Workers’ Compensation
           Board of BC in recent years.


5.1.1 Governance of the Workers’ Compensation system in BC
          The general role of governance and the associated conflicts of interest, as
       outlined above, apply to both private corporations or public entities such as the
       Workers’ Compensation Board of British Columbia. The Workers’ Compensation
       Board of BC is, however, different in some important ways from both other
       organizations and other Workers’ Compensation Systems.

              Workers’ Compensation Systems in general are different from corporate
           organizations in the social role that they are given. In particular, in all but one of
           the jurisdictions surveyed by Perrin, Thorau & Associates17, Workers’
           compensation is compulsory. That is, by law, workers’ compensation insurance
           must become part of the private employment arrangement between a firm and its
           workers. Moreover, in 91 of the 136 nations that had workers’ compensation
           insurance systems, the insurance provided was part of a general social insurance
           system that covered work and non work related injuries. Hence, it is a mandatory
           feature of the employment relationship that is often part of a more general social


17
     The one exception is Texas.


Garvey and Giammarino                           42
       program. In some jurisdictions, including BC, not only is coverage compulsory
       but the insurance must be purchased from a official monopoly insurer.

          The system in BC is typical of those found throughout Canada, but Canadian
       systems are themselves somewhat unique. As Perrin, Thorau and Associates
       note:

             “The Canadian Model, a comprehensive, compulsory, no fault
          system with an exclusive provincial government fund . . . is the
          minority approach throughout the world (only 10 countries or less
          than ¼ of those with separate Workers’ compensation systems have
          exclusive government funds).”

          This characteristic of our system implies that government’s interest in the
       provision of compensation insurance has a dimension that is not typically found
       elsewhere. In addition to the social goals that governments have, Canadian
       provincial governments, as the sole underwriter of the financial obligation
       involved, have a direct concern for the financial status of the fund.

          Hence the Workers’ Compensation System in BC has an extensive set of
       legitimate stakeholders who seek input to the governance mechanism. The
       government’s interest is at least as large as it is in the corporate sector. Workers
       and employers have an obvious stake due to the direct impact of benefit
       determination and operating efficiency. Less direct but equally important are the
       customers who ultimately pay for the benefits and the suppliers of other inputs
       whose viability is linked to the efficiency of the firms that it deals with.

          The need for inclusion in the governance process is further complicated by the
       fact of compulsion. Those who are unhappy with the governance of a corporation
       are typically free, as customers, workers, or suppliers of capital, to take their
       business elsewhere. Where this is not the case, conflict seems to increase. For
       instance, Karpoff and Rice (1989) study Alaskan Native owned firms where
       transfer of shareholdings was prohibited and find a high level of disruptive
       conflict among shareholders.

             The desirability of a governance system that both recognizes conflicting
       needs and is not tied to a particular interest group was recognized early on. The
       Meredith Commission in 1917 proposed the establishment of a Workmen’s
       Compensation Act that would be administered by a Board that, while appointed
       by the state, would be free of political influence. This was further emphasized by
       the Pineo commission in 1916. As noted in the Workers’ Compensation Board of
       BC (1996)

                “ The original recommendation of Pineo, which was acted upon
          by B.C. Legislature, was that the Workmen’s Compensation Board
          consist of three Commissioners. The primary characteristics that these
          Commissioners were intended to have were expertise and independence



Garvey and Giammarino                       43
          in performing the duties authorized by the Act . . . This intent was
          underscored by Pineo’s recommendation that the terms of office of the
          Commissioners should be ‘ at least 10 years . . .’ “

          Independence, however, can undermine accountability. In 1988 the Minister of
       Labour established a committee (the Munroe committee)                     to make
       recommendations that would allow employers and workers to have a greater input
       into the governance of the WCB. Apparently, the independent Commissioners
       were no longer seen as representing the interests of stakeholders sufficiently well.

          After studying a number of alternatives, the Munroe committee recommended
       a dramatic departure from the reliance on independent, benevolent
       commissioners. Through legislation that came into effect on June 3, 1991, the
       Act provided for a Board of Governors that consisted of five representative of
       workers, five representatives of employers, two governors who were to represent
       the public interest, and the chairman.

          Despite the conceptual appeal and initial support for this structure, it did not
       last long. Conflicts of interest arose and were exacerbated by the fact that many
       employers were new to the WCB, having been brought under WCB coverage in
       January of 1994. Conflicts over policy prompted the government to commission
       Judi Korbin and Patrick O’Callahan to review the structure and operations of the
       board of governors. Their report was presented to government on April 18, 1995.
       The report made several recommendations and concluded that

             “With the goodwill and cooperation of all stakeholders and the
          implementation of our recommendations, we believe that the current
          governance system can function effectively for the benefit of all
          stakeholders.”

          Unfortunately, the requisite goodwill was apparently inadequate. On July 11
       of the same year the President and CEO of the WCB tendered his resignation due
       to the lack of support from the Board of Governors. The government’s response
       was to create the current structure which consists of a Panel of Administrators.

          Hunt, Barth and Leahy (1995) conclude that “at this point, the experiment that
       put representatives of workers and employers at the table to govern the WCB
       together must be regarded as a failure.” Similar views have been expressed by
       others, most recently by the Business Council of BC in its July 31, 1997
       presentation to the Royal Commission.


5.2 Can Conflicts be Reduced
         The history of governance of the Workers’ Compensation Board reflects
      tradeoffs that are simple to identify but difficult to resolve. Ideally, organizations
      would be governed by bodies that recognize the needs of the various participants
      and are able to adjudicate conflicts in a way that best advances the goals of the



Garvey and Giammarino                       44
       organization. To achieve this ideal, the emphasis was initially placed on
       independence. Presumably this structure was based on the notion that the
       independent commissioners had the will and the ability to internalize the interests
       of the various participants. Based on this, independence would then allow them to
       pursue the goals of the organization as a whole. Unfortunately, independence
       brings with it a certain amount of isolation and power. In consequence, the more
       narrow but nevertheless important interests of individual groups may be
       inadequately recognized. The long term experiment with this model was ended
       when the lack of accountability was seen as critical. It was followed by an
       experiment which stressed accountability through interest group representation,
       but the result was perceived as a failure to achieve the goals of the organization.

          It is important to recognize that the conflict between the interests of the
       collective and the interests of individuals has raised serious concerns about
       existing governance mechanisms in both the private and public sectors
       throughout the world. Evidence of this is found in the “blue ribbon” committees
       that have been struck in Canada, the US and the UK to examine corporate
       governance. In their report entitled “Where Were the Directors?” the Toronto
       Stock Exchange highlighted tradeoffs that are essentially the same as those
       outlined above. Not surprisingly, among their main recommendations were those
       that encourage firms to have more independent directors that are concerned with
       the interests of all stakeholders. As two segments of the WCB history illustrate,
       however, a mechanism that relies heavily on goodwill and independence must
       also be designed to support such behavior through accountability.

          No existing solution to the governance problem is perfect and is clear that
       organizations will continue to experiment with various models in order to bring
       about improvements. There are, however, a number of factors that should be
       considered in the design of a governance system for WCB.

          A fundamental source of conflict is cross-subsidization. To the extent that
       assessments or regulations imposed on one group are more costly than they are to
       another group, conflicts will result. To the extent that experience rated
       assessments allow low risk employers to avoid subsidizing high risk employers,
       ERA could alleviate this source of conflict. In addition, adopting more narrow
       definitions of each subgroup will also make the insured risks for these subgroups
       more homogeneous. A more detailed discussion of subgroup classifications and
       experience rating can be found in chapter 4 of our study.

          In turn, cross-subsidization can only exist if participants do not have access to
       alternatives. Compulsory participation in the current system eliminates these
       alternatives. Exit and voice are alternative ways of dealing with an organizational
       policy that a participant does not find appropriate. By eliminating exit,
       participants are left with voice and, as WCB history shows, voice can be
       exercised in very disruptive ways. As we noted earlier, allowing some choice does
       not necessarily compromise coverage since coverage can be made mandatory



Garvey and Giammarino                       45
       while the decision of who will provide the insurance can be left to the firm and/or
       the worker.

          Finally, it may be beneficial to allow for at least some elected representation on
       in the governance process. It is one vehicle through which the preferences of
       participants are expressed.




Garvey and Giammarino                       46
                     Table 1: Determinants of First-Paid Claims Rates
 (t-statistics in parentheses, * indicates significant at 5%, ** indicates significant at 10%)
                             Average First-Paid         Average First-Paid       Average First-Paid
                               Claims, 1988-95            Claims, 1988-95          Claims, 1983-86
 Average First-Paid                                             0.951
    Claims, 1983-86                                            (29.1)*
   Herfindahl Index                  -4.59                      0.234                     -5.07
                                   (1.83)**                    (0.380)                   (2.03)*
   Absolute Subsidy                  -2.00                      -1.09                    -0.958
                                     (1.00)                    (2.45)*                   (0.437)
  Fraction of Small                 12.2**                       3.51                      9.15
          Firms                      (1.70)                    (1.70)*                    (1.23)
Participation in ERA                  14.3                       9.91                      4.59
                                     (1.17)                    (3.70)*                   (0.377)
  Fraction Swinging                   205                       -13.6                      229
   from Demerit to                   (1.32)                    (0.527)                    (1.54)
          Merit
  Fraction Swinging                  -90.9                       33.7                      -131
     from Merit to                  (0.766)                     (1.57)                    (1.04)
        Demerit
Fraction over 900%                    -122                       5.88                      -135
below Max. Demerit                   (1.41)                    (0.355)                    (1.53)
 Fraction 400-899%                   -24.5                      -16.6                     -33.5
below Max. Demerit                  (0.521)                    (1.69)*                   (0.572)
 Fraction 200-399%                   -48.4                     -0.848                     -24.5
below Max. Demerit                  (0.824)                    (0.074)                   (0.483)
 Fraction 100-199%                    128                        25.1                      108
below Max. Demerit                  (2.28)*                    (2.64)*                  (1.90)**
Forestry Assessment                  -1.49                     -2.155                     0.703
           Plan                     (0.467)                    (4.30)*                   (0.220)
     Construction                     1.90                     -1.041                      3.09
   Assessment Plan                  (0.772)                   (1.89)**                    (1.32)
    Voluntary Plan                   -5.91                      0.328                     -6.56
                                    (3.85)*                    (0.884)                   (4.20)*
       Constant                      -4.90                      -8.96                      4.27
                                    (0.420)                     (3.43)                   (0.369)
      Adjusted R2                     0.27                       0.96                     0.308




Garvey and Giammarino                        47
                                 REFERENCES

Dickens, William and Lawrence Katz (1990) “Reflections on the Inter-Industry Wage
Structure”

Hunt, H. Allan (1992) Workers Compensation Board of BC: Assessment Department
Administrative Inventory W. E. Upjohn Institute for Employment Research

Hunt, H. Allan, Peter S. Barth, and Michael J. Leahy (1996) The Workers’ Compensation
System of BC: Still in Transition W. E. Upjohn Institute for Employment Research

Hyatt, Doug (1997) “Econometric Estimation of Some of the Effects of Experience
Rating: A Report to the Workers’ Compensation Board of British Columbia”, Royal
Commission on Workers Compensation in BC

Hyatt, Doug and Boris Kralj (1995) “The Impact of Workers’ Compensation Experience
Rating on Employer Appeals Activity” Industrial Relations 34, 95-106.

Internal Audit and Evaluation (1995) ERA Evaluation Study: Experience With
Experience Rating, Workers’ Compensation Board of British Columbia

Internal Audit and Evaluation (1995) ERA Evaluation Study: Phase I Statistical Profile,
Workers’ Compensation Board of British Columbia

Internal Audit and Evaluation (1996) ERA Evaluation Study: Phase II Statistical Profile-
Draft Report, Workers’ Compensation Board of British Columbia

Krueger, Alan and Lawrence Summers (1988) “Efficiency Wages and the Inter-Industry
Wage Structure” Econometrica 56, 259-93.

Lazear, Edward P. and Sherwin Rosen (1981) “Rank Order Tournaments as Optimal
Incentive Contracts” Journal of Political Economy

Ruser, James W. (1985) “Workers Compensation , Experience Rating, and Occupational
Injuries”, Rand Journal of Economics 16, 487-503.

Thomason, Terry and Richard P. Chaykowsi, eds (1995) Research in Canadian Workers
Compensation, IRC Press.

Vancouver Sun (1997) Guest Article June 30, D2.

Viscusi, W. Kip (1986) “The Impact of Occupational Safety and Health Regulation”
Rand Journal of Economics 17, 567-80.



Garvey and Giammarino                     E-1
Walker, Cathy (1986) “A Labour Perspective on Experience Rating”, presented to Third
Annual Corpus Workers’ Compensation Conference, Toronto, September 16, 1986.

Workers’ Compensation Board of BC (1996) “An Overview of the Historical
Development and Current Structure of the Workers’ Compensation Board of British
Columbia”, November

Baker, Laurence C. and Alan B. Krueger (1993) “Twenty-Four-Hour Coverage and
Workers’ Compensation Insurance”, Health Affairs 12, 271-81.

Butler, Richard J. (1996) “Lost Injury Days: Moral Hazard Differences Between Tort
and Workers’ Compensation” Journal of Risk and Insurance 63, 405-33.

Cantor, Shawn E and Price V Fishback (1995) “Nonfatal Accident Compensation and the
Common Law at the Turn of the Century”, Journal of Law, Economics and Organization
11, 406-33.

Epstein, Richard A. (1982) “The Historical Origins and Economic Structure of Workers’
Compensation Law”, Georgia Law Review 16, 775-819.

Experiences of Injured Workers Employed by Nonsubscribing Employers (1997)
Workers Compensation Board of Texas, February.

Hayek, Frederich (1945: “The Use of Knowledge in Society”, American Economic
Review 23, 25-42.

Kaestner, Robert (1996) “The Effects of Government-Mandated Benefits on Youth
Employment”, Industrial and Labor Relations Review 50, 122-42.

Public Policy Research Institute (1995) “Nonsubscribing employers in Texas”, Texas
A&M University

Rothschild, Michael and Joseph E. Stiglitz (1976) “Equilibrium in Competitive Insurance
Markets: An Essay on the Economics of Imperfect Information” Quarterly Journal of
Economics 90, 629-50.

System Performance Update: Employer Participation in the Texas Workers'
Compensation System (1995), Workers Compensation Board of Texas, April.
http://www.texas.gov/agency/478.html.

Veljanowski, C. (1980) “An Economic Analysis of Employer Liability”, manuscript,
London School of Economics

White, Halbert (1980) “A Heretoskedasticity-Consistent Covariance Matrix Estimator
and a Direct Test for Heteroskedasticity” Econometrica 48, 817-38.



Garvey and Giammarino                    E-2
                                         Technical Appendix
        Consider a representative member of an n+1-firm subgroup. Assume for now that they are
identical, and follow WCB convention by defining all variables as a fraction of each firm’s pay-
roll. The accident rate of i is denoted:
                                              a i = a max – ( s i + η + θ i )                                            (1)
        a max is the base accident rate, s i is i’s safety effort, η is a subclass-wide shock (mean
zero) and θ i is i’s specific shock, also mean zero and identically and independently distributed
with the shocks of the other n-1 firms in the subclass. The cost of safety is denoted.
                                                      C ( s i ) ;C′ ,C″ > 0                                              (2)
        Firm i pays the WCB an amount:
                                                      Pi = Bi + Y ( Ri )                                                 (3)
        Bi is the base rate, Y0 is the experience rating adjustment, and
                     Ri = ai –     ∑ aj ⁄ n       =   ∑ sj ⁄ n – si + θi – ∑ θ j ⁄ n ≡ Sj – s i – φ                      (4)
                                   j≠i                j≠i                      j≠i

        is the firm’s experience rating measure, defined relative to the other firms.1
       Finally, denote by G and g the cumulative and marginal density function of φ . Thus the
probability that firm i’s relative performance R i takes on any arbitrary value x is g ( S j – s i – x ) .
The BC WCB Experience Rated Assessment system can be summarized as

                                                            λL ,R i > L
                                             Y ( R i ) = λR i ,– L ≤ R i ≤ L                                             (6)
                                                          – λ L , Ri < –L

        The upper and lower limits L and -L are defined as the 100% demerit and merit levels,
respectively. If we write the firm’s profits net of worker accidents as π i , firm i chooses its safety
efort to maximize:
                                               πi – [ E { Pi ( si ) } + C ( si ) ]                                       (7)
        where
                                              L
     E { Pi ( si ) } = E { Bi ( si ) } + λ   ∫–L g ( sj – si – x ) dx + L { G ( sj – si + L ) – [ 1 – G ( sj – si – L ) ] }
                                                                                                                         (8)
        We now need to put some structure on the base rate. In principle, every subclass is
supposed to be self-financing, in which case B i = ∑ n      a ⁄ n. But we know that in fact there
                                                     j = 1 j
are cross-subsidies between subclasses. Summarize this by assuming that our subclass bears a
fraction z of its costs plus (1-z) of the deficit of the rest of the system which we denote by D, so
that Bi = z ∑ n a j / n + ( 1 – z )D .
                 j =1
        The first-order condition for i’s safety choice is now:
                             z ⁄ n + λ [ G ( S j – s i + L ) – G ( S j – s i – L ) ] – C′ = 0                                          (9)
        Clearly, smaller cross-subsidies (larger z) increase safety effort and reduce accidents. They
will also, however, reduce the marginal effect of introducing experience rating so long as the C’
function is convex.
        Holding constant the number of firms in a subclass, the square-bracket term is increased
and so is effort, the more similar are the firms (all else equal, this has the effect of increasing the
proportion of variance which is due to η rather than the idiosyncratic θ j ). Similarly, increasing
the number of identical firms reduces the variance of φ .
        Clearly, firms in a given subclass differ in more than just their idiosyncratic iid shocks.
The fact that there are some firms that stay at maximum demerit suggests that expected accident
rates have a firm-specific component. We can incorporate this by introducing an additional firm-
specific quality term so that
                                          a i = a max – ( s i + η + θ i + q i )                                                       (10)
        Since the WCB just places firms into subclasses, we can safety treat the q’s as exogenous,
and with a zero mean over all firms. This formulation has the nice property that if we go through
the model again it gives each firm its own set of effective limits, L j = L – q j . Intuitively, a high-
quality firm will hit its upper limit (100% merit) when its effort plus idiosyncratic shock add up to
a smaller number, and will only go into demerit if it has a very bad idiosyncratic shock. In general,
increased variability of the q j can only reduce effort since the term
[ G ( s j – s i + L i ) – G ( s j – s i – L i ) ] is maximized at q j = 0 if the distribution G is either bounded or
symmetric and unimodal.




    1. For notational convenience, we suppress the fact that accidents in year t affect premiums two and three
       years hence. When the discount rate is r, the strictly correct way to express the effect of actions at time t is
       in fact:
                                                 1 -                                              1 -
                                         ------------------ Y t + 2 ( R t ) + ------------------ Y t + 3 ( R t )
                                         ( 1 + r )2                                       ( 1 + r )3
                                         ------------------------------------------------------------------------------------------
                                                                                                                                  -    (5)
                                                                                     2