Workers Compensation

Employee Responsibilities and Rights Journal, Vol. 10, No. 3, 1997

Procedural Fairness, Return to Work, and the Decision to Dispute in Workers' Compensation

Kareo Roberts' and WiUard Yotuig>

This study investigates how injured workers evaluate the fairness of the workers' compensation claims process arid how that evaluation affects the decision to formally dispute their clairru Survey and administrative data are used to test a model where individuals are hypothesized to base their overall impression of the fairness of the process based on dimensions of procedural justice criteria. They are then hypothesized to decide whether to formally dispute the claim based on a combination of procedural fairness concerns and whether or not they retumed to work with the same employer. The implications of the results for structuring the workers' compensation claims process are then discussed.

KEY WORDS: procedural justice; fairness; workers' compensation; disability; dispute; retum to work.

INTRODUCTION

Work-related illness and injuries constitute a significant cost to employers. Ac- cording to the Bureau of Labor Statistics Survey of Occupation Illnesses and Inju- ries, there were 6.7 million work-related illnesses or injuries in 1993. In the same year, 8.5 out of every 100 fuU-time workers experienced lost work time due to oc- cupational injuries and illnesses. The U.S. Department of Health and Human Serv- ices estimated that employers spent approximately $53.1 billion on Workers' Compensation insurance in 1993 (Wyatt Coip-, 1994). Workers' compensation, the primary benefit program for work-related disabilities, was originally designed as an insurance s>'stem to replace tort remedies. The hope was that the workers* com- pensation claims process would be relatively uncontentious; however, this has not been the case. A significant number of claims are contested—over 20% in Michigan where the study described in this article was carried out. Contested cases are usually more costly because they last longer, involve attorneys' fees, and risk resulting in potentially higher awards (Boden & Victor, 1994). In addition, they almost always sour the employment relationship (Roberts & Gleason, 1994).

'School of Labor and Industrial ReUtions, Michigan SUte University, East Lansing, Michigan 4S824.

193

0892-7545/97/O90(M)l93$ll50l« C 1997 Plenum PuUtshing Cotpomkw

X94 Roberts and Young

While the high costs of diqjutes are commonly recognized, there is little research on the nature of disputes that has clear policy implications for employers. The re- search on disputes in workers' compensation is sparse in part because disputes are processes rather than events and as such are difficult to research. Disputes typically have an incubation period where they evolve from some experience that is perceived as injurious by at least one party. However, an injurious experience is not automat- ically a dispute: Once an experience is perceived as injurious, it must be transformed into a grievance where some party is blamed; some redress must be requested and then refused for a dispute to occur (Felsdner, Abel, & Sarat, 1980-81). Most of the research on disputes in workers' compensation begins at this final stage when a for- mal dispute has been articulated. Conflict is only identified as such when it reaches the point of becoming a formally contested case where third-party intervention, in the form of mediation or adjudication, is required. The difficulty with this approach from a policy perspective is that it provides little guidance for the actors who have the most control over the process, employers and insurers.

The research reported in this article attempts to fill this gap by examining the claims process from the perspective of the injured worker to determine what mo- tivates injured workers to formally dispute their claims. The framework used in this study is based on procedural justice theory, which is useful for evaluating the work- ers' compensation claims process in two ways. First, the literature on procedural justice divides people's perceptions of fairness into several dimensions that can be valuable in the formulation of policy. Second, it is frequently asserted in the pro- cedural justice literature that perceptions of fairness about the decision-making process can affect one's reactions to the decisions generated by that process (Thibaut & Walker, 1975; lyier, 1988). Specifically, for any given outcome, people are more likely to abide by a decision that they think was generated by a fair process (Folger & Greenberg, 1985; Eskew, 1993; Youngblood, Trevino, & Favia, 1992). While disputes around disability issues have been a relatively minor workplace con- cem in the past, the combined effects of the rising costs of workers' compensation and the passage of the Americans with DisabiLties Act have made them more im- portant. With ADA, employers are going to experience more disability-related con- flict as they confront issues of reasonable accommodation mandated by the law.

This research is designed to address three questions that have implications for disability policy. The first question is. When people describe the workers' compensa- tion claims process as fair, what do they mean? This can be restated as. Which di- mensions of fairness are most important to claimants? Second, how does the fairness of the claims process relate to the likelihood of an injured worker retuming to the same employer? Specifically, does the injured worker's evaluation of the faimess of the claims process affect his/her likelihood of continuing the preinjury employment relationship? And finally, how do faimess perceptions affect the decision to contest the claim, directly through the individual dimensions and/or the global perceptions of the faimess of the claims process, or indirectly through the retum to work outcome?

In the next section of the artide is a brief description of the workers' com- pensation claims process, followed by a review of the faimess framework used in this article. This is followed by an empirical analysis. The artide condudes with a discussion of the results and their implications for employers.

Pcrceptioni of Faitncss 195

The Setting: Tbe Workers' Compensation Claims Process

Injured workers are entitled by law to wage replacement benefits and full medi- cal care plus rehabilitation if the injury appears to be resulting in long-term or permanent disablement.^ In most workers' compensation cases, there are three pri- maiy actors who make decisions about an injured workers' claim: the employer, the claims examiner, and the evaluating physician who assesses worker readiness to retum to work. In a simple daim the employer, acting on behalf of the employee, files the daim by reporting the acddent or injury to the carrier. The claim is ad- ministered by a claims examiner from the employer's insurance company, who is meant to act on behalf of the employer. Administering a claim entails investigation, monitoring, benefits and medical bill payment, and approving medical and reha- bUitation treatments. For most daims, the insurer pays wage replacement and medi- cal benefits, and the employee returns to work as soon as suffidently recovered.

Either party can dispute some asp>ect of the claim. There is a standing Board of Magistrates, all attomeys by training, who hear the case. Formal rules of evidence apply in these hearings, and usually both sides have attomeys present. There is an Appeal Board which can review the hearing-level decisions.

The range of benefits available under workers' compensation is fairly drcum- scdbed by statute and largely nonnegotiable. Injured workers receive partial wage replacement and the medical and rehabilitation costs associated with the disability are paid. There is no compensation for pain and suffering.^ Even for those cases that are settled by a lump-sum payment and the amount is negotiated, the settle- ment is meant to be based on statutorily set benefit levels. Most disagreements about claim outcomes thus tend not to be over amounts of money but rather over other aspects of the case.

TTiere are formally only two legitimate reasons in the Michigan workers' com- pensation system for an employer to deny a claim or to terminate benefits: If an injury or disability is not work-related or if a reasonable job offer is refused after the injured worker has reached maximum medical improvement (State of Michigan, 1988, p. 21). In reality, these criteria are often subjectively determined. Disputes over whether the disability is caused by work range from whether the person got hurt somewhere else but claimed it occurred at work (the "Monday moming" in- juries) or to more sophisticated problems such as whether work exacerbated an existing condition. In Michigan, the employer employs the worker "as is," and with a few exceptions is fully liable if the work contributed to an existing condition that results in disability.

The fundamental problem for evaluating the reasonableness of a job offer is that disability is an inherently ambiguous status, and assessing the extent of dis-

^ e description presented here is the workers' compensation ^ t e m in Michigan, wbere this study was conducted. Althou^ every state program is different, the basic daims process is fairly univer^.

'Some states award impairment benefits designed to compensate for a permanent impairment Depending on their structure, these benefits may resemble pain and suffering awards. Michigan law includes a schedule of payments for a vety limited set of losses. However, even for states that provide impairment benefits, these are usually based on limits set by statute, and big, hit-the-jackpot awards are virtually unheard of in workers' compensation.

19« Roberts and Yoang

r, the degree of recovery, and even the legitimacy of a disability can be difficult (Gleason & Roberts, 1993; Stone, 1984). The thrust of the job offer reasonabilily criterion is that the employee is physically able to do the job. Because it is some- times difficult to sort out when a good faith offer is made- and when a refusal is legitimate, workers' compensation daims often end with termination (26% of this sample did not retum to work with the employer where the injury occurred).

Ideally, returning to work is a strai^tforward decision making process about when and in what capacity; however, there is anecdotal evidence that this process is not perceived as tieutral by injured workers (Roberts & Gleason, 1994). For ex- ample, injured workers frequently complain that once they make a workers' com- pensation daim, they are labeled as troublemakers by employers and co-workers. It is also common for injured workers to note that employers, supervisors, and co- workers often do not want them back at work unless they are "100%." In addition, in cases of prolonged disability, injured workers report that their employers terminate them as a way to end benefits. Under these sorts of circumstances, inability to retum to work is likely to be perceived as the outcome of an unfair decision-making process. One recourse under workers' compensation is to formally contest the daim.

The Analytical Framework: Procedural Faimess

People appear to care about faimess and apply it as a yardstick in their evalu- ation of a variety of dedsion-making processes (Tyler, 1989). The research on fair- ness makes a distinction between two types of faimess: distributive and procedural. Distributive justice is the perceived faimess of the outcomes of a decision-making, resoun« allocation, or diqjute resolution process; and procedural justice is the per- ceived faimess of the process that generated that outcome (Barrett-Howard & Tyler, 1986; Folger & Konovsky, 1989; Folger, 1977; Greenberg, 1987; Sheppard & Minton, 1986). The study reported here addresses procedural justice concerns. The focus of the research on procedural justice has been on differentiating among the various dimensions individuals use to evaluate the faimess of a process, under what condi- tions these different dimensions appear to be important, and how perceived faimess affects various behaviors (Bies & Shapiro, 1988; Eskew, 1993; Folger & Konovsky, 1989; Greenberg, 1986, 1987; Leventhal, Kaniza, & Fiy, 1980; Lind & Ty\cT, 1988; Shapiro & Brett, 1993; T^ler, 1989; Youn^lood, Trevino, & Favia, 1992).

Dimensions of Procedural Faimess

Recent literature suggests that people use three types of criteria to evaluate the faimess of a process: the degree of opportunity offered to infiuence the out- come, the structural aspects of the dedsion-making process, and the quality of the interpersonal interactions (Bies & Shapiro, 1988; Eskew, 1993; Folger & Konovsky, 1989; Greenberg, 1990; Lind, Kanfer, & Earley, 1990; Lind & Tyler, 1988; Shapiro & Brett, 1993; Tyler, 1989; Tansky, 1993)." Procedural concerns of the first type

*rhe first two are often referred to as instrumental and noninstrumental criteria, respectively.

Pereeptioiis of Fainwst I "

are valued for the practical effects they may have on the outcomes of the decision- making process ( l i n d & Tyler, 1988). It is argued that when individuals believe that the opportunity to voice one's views can have a material effect on the outcome of a process, that process is more Ukely to be viewed as fair (Lind, Kanfer, & Earley, 1990; Shapiro & Brett, 1993).

The structural characteristics of the process are based on the criteria hypothe- sized by Leventhal, Karuza, and Fry (1980) and refer to the procedural mles as- sodated with decision making (Eskew, 1993). Structural characteristics include aspects such as the opportunity to present one's case, that the process is consistently administered, and that the dedsion maker has complete information while making the decision (Eskew, 1993; Shapiro & Brett, 1993). Unlike the first set of criteria, these characteristics are not valued for their potential effect on the outcome, but rather for what they signal about how the partidpant is valued in the group or relationship. In the work setting, a fair process would be seen as an indicator that the employee is valued by the employer.

The interpersonal criteria (Bies & Moag, 1986; Shapiro & Brett, 1993; Tansky, 1993) refer to the quality of the interaction between the partidpants and the de- cision maker. More specifically, they refer to the demeanor of the decision maker, how that decision maker treats the participant, and the quality of the communica- tion between dedsion maker and partidpant (Bies & Moag, 1986). The decision- maker qualities that are evaluated include characteristics such as respectfulness, neutrality, consideration, and ethicality (Bies & Moag, 1986; Shapiro & Brett, 1993). The interpersonal criteria are similar to the structural characteristics in that they are often interpreted as a refiection of one's standing in a group or value in a relationship (Shapiro & Brett, 1993; Tyler, 1989).

These procedural faimess criteria can be applied to a variety of decision-mak- ing and resource allocation processes. While much of the research in this area has been done in laboratory settings, the field applications range trom legal decisions, public policy decisions, grievance mediation and arbitration dedsions, and admin- istrative decisions made in the workplace such as performance evaluation, pay, and termination dedsions (Folger & Konovsky, 1989; Greenberg, 1986; Konovsky & Cropanzano, 1991; Leung & U, 1990; Und, KuUk, Ambrose, & Park, 1993; Shapu^o & Brett, 1993; Tyler, 1990; Youngblood, Trevino, & Favia, 1992).

The literature on procedural justice suggests that the different types of criteria are not equally important in all situations; in other words, these criteria are not universally applied. In a lab study, Barrett-Howard and Tyler (1986) found that which criteria were valued depended on whether the setting was competitive or cooperative and whether the decision-making process was formal or informal. One, although not the only, likely predictor of which type of criteria model will be most important is whether the relationship where the d i l u t e is occurring is long or short- term (Lind & Tyler, 1988). In a short-term relationship, individuals are thought to be concemed with outcomes and so emphasize control dimensions of procedural faimess. In a longer-term relationship, preserving the relationship is also an objec- tive, so that those dimensions of procedural justice that refiect one's position in the relationship are valued. According to this criterion, therefore, a fair process is an end in itself.

198 Roberts and Young

Faimess, Outcomes, and Formal Dispute Behavior

One of the policy-relevant assumptions about procedural faimess is that the degree to which a process is perceived as fair affects partidpants' behaviors (Tyler, 1990). The perceived faimess of a process may determine whether or not the par- tidpants abide by the dedsion (Mondak, 1993). Several field studies explicitly ad- dress the question of the relationship between perceptions of faimess of a process and the dedsion to formally dispute its outcome.

Two studies of arbitration awards reported in a paper by Lind et al. (1993) tested the hypothesis that individuals develop a global impression of whether or not a process is fair and use that impression to determine whether or not that authority should be obeyed. Both studies examined litigants' reactions to arbitration awards and the subsequent decision to go to trial. Their results suggest that pro- cedural justice may mediate the effects of an unfavorable outcome and that a global impression of faimess rather than the individual dimensions determines the decision to accept the outcome of a process.

Youngblood, Trevino, and Favia (1992) examined how dismissed workers view their termination and how they respond to third-party dispute resolution of their ui^ust dismissal claim. The results show that procedural justice concems are cited most fiequently as the reason for challenging the termination. They find that, aside from wanting to be reinstated, a dominant reason terminated employees tum to a third party is that they are seeking a procedurally fair dedsion-making process, which they feel they were denied in the workplace.

Brockner et al. (1994) examined how procedural justice attributes can affect employee reactions to layoffe. Examining three different groups, layoff victims, sur- vivors, and lame ducks, they found that all viewed job loss as negative. However, procedural faimess perceptions affected how strenuously respondents reacted to the bad outcome. Similar results were found by Daly and Geyer (1994) in their examination of employee reactions to their employer relocating. In most cases, re- location is viewed as a negative outcome. Employee intentions to stay or leave the employer were infiuenced by the perceived procedural faimess of the relocation dedsion. Together this research suggests that the procedural faimess of a decision- making process that results in what are generally perceived as negative outcomes can affect how individuals respond to that outcome.

METHOD

Research Questions

The three questions addressed here are: Which dimensions of procedural fair- ness are most important in injured workers' overall evaluation of the faimess of the workers' compensation claims process? How does procedural faimess affect re- tum to the same employer? And, How does faimess affect the dedsion to dispute the claim, directly through the individual and global dimensions of faimess, or in-

PerctpUoiis of Fairness 199

o

I 2 e

"8

2 o

200 RiAerts and Young

directly through the retum to work outcome? To answer these questions, the daim is conceptualized here as three separate processes, as shown in Fig. 1.

During the first process, individuals are attempting to develop an overall im- pression of the faimess of the process in which they are partidpating. The empirical question about this process is. Which procedural jtistice asp>ects of the claims proc- ess are significant in injured workers' overall impression of how fairly they are treated? Theoretically, this is an open question. The research on faimess indicates that p)eople value those features of a process that may let them infiuence or control the outcome, as well as those that refiect their value or standing in their relation- ships with the other disputants and/or dedsion makers. A key feature of the work- ers' comp>ensation claims process is that it is an insurance process embedded in the employment relationship (Roberts & Gleason, 1994). Because the workers' com- pensation process takes place in a work context, it is likely that injured workers will value those aspects of the process that signify that the employee is valued by the employer, employees would then be expected to assign relatively high weights to their perceptions of the strurtural and interpersonal features of the process, those dimensions that have implications for the long-term employment relationship (Ty- ler, 1990). However, filing a workers' compensation claim also triggers an insurance process, which is typically exp)ected to be a short-term relationship. Therefore, it is possible that injured workers may put more weight on their opportunities to control or infiuence the outcome in their overall faimess evaluation.

The second process involves the relationship between faimess of the claims process and the likelihood the injured workers will retum to the same employer. In this process, retuming to the same employer is seen as depending on severity of injury (which may affect overall ability to work), financial hardship (which may motivate a retum to work to the same employer), and the faimess of the claims process. The expectation is that if the claims process is perceived as unfair, this will strain the employment relationship and reduce the likelihood of retum to work.

The third process involves the decision to formally dispute the daim. Here, individuals are seen as making this decision based on both the perceived faimess of the claims process as well as the outcomes of that process. TTie question at this point is: Is faimess a factor? If so, do the individual dimensions of procedural fair- ness or overall impression directly affect the decision to dispute; or, do faimess concems operate indirectly through the retum to work outcome?

One study, which examined the perceived faimess of a process and whether the outcome was formally accepted or disputed found strong support for the im- portance of overall faimess evaluations in the decision to accept or reject court- ordered arbitration awards relative to both subjective and objective measures of outcome (Lind et al., 1993). The study by Youngblood et al. (1992) suggests that individuals who feel they have been denied procedural justice during a termination process will seek that justice through an appeal to a third party. Applying these results to workers' compensation, the expectation would be that the overall faimess evaluation of the claims process will significantly affect the propensity to dispute, and that the effect of faimess perceptions on dispute behavior will be mediated by the retum to work outcome.

Perceptioas of Faimess ^ '

Data Sources

The sample was drawn from daims files kept by the Michigan Bureau of Work- ers' Disability Compensation claims files. A mail survey questionnaire was then ad- ministered to collect information about each individual's perceptions of the process. To construct the sample, 759 cases were randomly selected from a slice-in-time sampling frame of all cases with dates of injury between April 1, 1984 and March 15,1985.' The daim files induded information about the injury (type of injury and bcxly part), worker characteristics (age, gender, job title, marital status, number of dependents), employer characteristics (type of business, location, insurance status), benefits received (weekly amounts, duration), whether the daim was dispute4 and how it was resolved.

Questionnaires were mailed to 634 people* to collect information regarding the injured workers' perceptions of the disability claim process. Standard survey mail collection techniques were employed, induding reminder postcards, telephone calls to all nonrespondents, and a second mailing to all potential respondents who had said on the phone that they would retum the survey. The final sample size was 182, a 28.7% response rate.

Two-thirds (67.6%) of the sample was male. The mean age of the respondents was 43.1 years. Over one-third (35.2%) were employed in manufacturing, and one- quarter (25.8%) were in the service industry. Most, 58.2%, were in blue collar oc- cupations. Fewer than half, 41.8%, had education or training beyond high school. Mean tenure with their employer prior to the injuiy was 17.7 years. All were em- ployed in Michigan at the time of the injury. Comparing the sample to the total daims made in Michigan in 1984 (MIOSHA, 1985), there were slightly more males in total claims (71.2%); the modal age category was the same for both groups (25- 34), but our sample had fewer claimants under age 25 (6.6% compared to 17.9%). Slightly more of total claimants were employed in manufacturing (40.9%) and some- what fewer in services (14.9%). More of the total claimants were in blue collar occupations than in the sample (69.9%).

Measures

Five scales were formed from the data collected in the mail survey by utilizing exploratory factor analysis (prindpal components analysis with varimax rotation). Mirroring the work of Lind et al. (1993), none of the iten^ in the measures of the

This time period was selected to assure that most of the daims in the sample hud dosed ^ «h?J™e the data wire collected. The sample was constructed in three stages. First a manual roster held by the BWDC of cases where claimants received other employer-paid benefits was used to select cases from large employers. Second, these cases were found in a manual card file and the next case m the file ^ a date of injuiy in the sampling period was selected to construct the smaUer f»P'°y«.'?"«• ^ " ^ to assure having enough disputed cases, the contested case MIS system was marched and the fit« 20 cases with a date of injury in the observation period was selected for each month in that penod

W u ^ si:ver^ years had passed since the date of injuiy, a letter was sent to tbe 759 people for whom administrative daU were initially cdlected telling them to expect the survey. Of these letters, 125 were retumed for bad addresses.

202 Roberts and Young

dimensions of procedural justice contained the words "just" or "fair," and none sought a global evaluation of the claims process.

Opportunify to Influence Outcome is a three-item scale that measured the in- jured worker's perceived level of control over the case decisions make by the em- ployer, the examiner, and the evaluating doctor. The response format ranged from 1 to 3 with a smaller response indicating a greater perceived sense of control (Cron- bach's a = 0.86).

There were two measures of the structural characteristics of the process: Consistency is a two-item scale that measured the perceived consistency of the

treatment accorded the claimant by the employer and examiner with reference to other p)eople or comparable situations, specifically, the claimant's previous experi- ence with other insurance claims or other people's workers' comp)ensation claims. The response format ranged from 1 to 5, with a smaller response indicating a greater p)erceived level of consistency (correlation = 0.59).

Information accuracy is a four-item scale measuring the extent to which the injured worker thought that the employer, examiner, and evaluating physidan had sufficient information about the claim to make accurate dedsions. The response format ranged from 1 to 3, with a smaller response indicating a greater perceived sense that each actor had enough information to make accurate decisions (Cron- bach's a = 0.91).

Quality of Interaction is constmcted from nine items assessing the injured worker's perceptions regarding the ethicality and professionalism of the behavior of the employer, the claims examiner, and the evaluating physidan during the daims process. Three identical questions were asked about each of these three actors con- ceming how helpful they were, whether or not they listened, and the extent to which they appeared to trust the employee. The response format of the items was a 1 to 5 scale, with a smaller response indicating a greater perceived level of ethical appropriateness (Cronbach's a = 0.86).'

Global Faimess Impression is a six-item scale measuring the global impression of procedural faimess. Three of the items measure the injured worker's perceptions regarding the level of honesty displayed by the employer, the evaluating doctor, and the examiner, respectively (1 = always honest; 5 = always dishonest). The remaining three items measure how fairly the claimant thought he/she had been treated during her/his disability by the employer, the evaluating doctor, and the examiner, respectively (1 = very fairly; 4 = very unfairly). The scores were stand- ardized because the response ranges differed (Cronbach's a = 0.89).* Global fair- ness is endogenous to the model.

Two variables were constmcted from the data in the administrative claims re- cords and measure outcomes of the daims process.

^One set of items in the measure of quality of interaction is the extent to which each of the actors listened to the claimant. Theoretically, these times could belong with the decision control construct as an indicator of voice. However, the factor loadings for these items were 0.79 or better on the quality of interaction factor compared to no greater thao 0.49 on the control factor.

^Third-party honest)' u> often cited as a dimension of quality of interaction, and that relationship was tested using factor analysis. However, the items related to honesty had factor loading of 0.79 or better on the overaU faimes.s scale compared to between 0.61 and 0.66 on the quality of interaction scale.

Perceptions ot Fairness 263

a

C or

mm

Ji

IS

2 1 - 1 1

VI VI

204 Roberts and Young

Returned to work is a dichotomous variable that measures whether or not the claimant retumed to the same employer foUowing the disability period (1 = yes; 0 = no). This was indicated on the case dosure form or if the case was still open by reading case materials. Retumed to work is endogenous to the model.

Dispute is a dichotomous variable that measures whether or not the case was disputed (1 = yes; 0 = no). Dispute was endogenous to the model.

Two additional variables were induded to control for other factors likely to affect retum to work as well as propensity to dispute:

HospitaUzation is a dichotomous measure of whether the claimant was hospi- talized during the disability period (1 = yes; 0 = no). The motivation for induding hospitalization was that if an injury was severe enough to warrant in-patient care, the greater the employee's concem about recovery and future employability and therefore lower likelihood of retuming to work and greater propensity to dispute an unfavorable outcome (Roberts, 1992).

Replacement Rate is a measure of the financial impact of the disability on the claimant. It is equal to the ratio of total benefits, excluding those received in the final phase of the claims process, to total lost earnings due to the disabUity. The smaller the replacement rate, the greater the finandal burden to the daimant dur- ing the disability period.

The mean, standard deviation, and correlational and reliability values are shown in Table 1. A recursive path analysis model was used to examine the hy- potheses discussed above and illustrated in Fig. 1. The strength of each link in the model was captured by regressing each variable on its antecedents. The path co- efficients are equivalent to standardized regression coefficients. The PATH routine in the PACKAGE computer program was used to generate the estimated stmctural coefficients (Hunter, Gerbing, Cohen, & Nicol, 1980).'

RESULTS

Fit of the Model

The reproduced and residual correlations are shown in Table II. The differ- ences between the empirical correlation matrix and the reproduced correlation ma-

*The underlying assumptions of path analysis are the same as those required for ordinary least squares (Asher, 1983). Our model risks violating these assumptions because the dependent variables are dichotomous. Ordinarily, when tbe dependent variable is dichotomous, nonlinear logit or probit estimation is used in order to constrain the predicted values to fall within the range of zero to one. Linear estimation with weighted least squares can be used under certain circumstances, however (Aldrich & Nelson, 1984). One simple test of the acceptability of using linear estimation is to estimate the model using OLS and examine the extent to which the predicted values fell outside of the zero- one range. Only one of the predicted values fell outside of the acceptable range for probability of dispute and three for retum to work, indicating that linear estimation would not generate inefiident estimates with these data. Also, the suges of the model were estimated separately using both ordinarily least squares and wei^ted least squares and the results were compared. While the level of significance increased using wei^ted least squares as would be expected, the change in significance level never exceeded 0.014 and was usually zero, it was concluded that the linear estimation used in the path model will result in unbiased and efiident path estimates.

Fcnxpthms orFaimcss 205

i I •a

i .a

S

8888 ooo'd

ggg I 2S2i ' o o c ^

S S 8 S | I 2 | g

g g I § c 8 2 § a od ddoc^oo

B is ! i •I Mil ill € i s i £ - 2 Sg-a g.8S I =5'§-5.2.1: j

^

206 Roberts and Yoong

trix are not ext«ssive. Before examining the individual error terms, however, the overall fit of the model to the data was testetl using Specht's Q x^ test (Fullagar & Barling, 1989).'" The computed 1} statistic equaleti 3.70, which failetl to exceed the critical value necessaiy to reject the mtxlel {y} bound was 18.475 {df = 1, p < 0.01)). To test the fit of each path, a comparison was made between the 5% a level error variance index (which is computed with the following fomiiila: {1.96 * j2]i/2 • Var (e)} and the error matrix. The absolute differences between the em- pirical and the reproduced correlations in the difference (error) matrix were com- pared to the error variance index; any difference larger than the index indicates an improperly specified path. The largest absolute tlifference between the empirical correlations and the reproduced correlations is 0.15. This is less than the error vari- ance threshold limit of 0.169, indicating that none of the paths was misspecified.

Path Coeffidents

Figure 1 shows the path coefficients. All of the variables in this model were standardized, so that the path coeffidents can be interpreted as partial p coefficients (Asher, 1983)." As is evident, 11 of the 15 paths are significant {1 ax p < 0.01, 2 at p < 0.05, and 2 at p < 0.10) for the sample size {N = 140).

Three overall questions were explored. The first was which aspects of prtKe- dural faimess would be important in injured workers' overall impression of the fair- ness of the workers' compensation claims process.'^ All of the measures of procedural faimess were statistically significant at the 0.01 level or better. Of the four dimensions, quality of interaction was by far the most important determinant of the overall faimess impression (B = 0.47), close to twice as important as op- portunity to influence the outcome, the dimension with the next largest coefficient. Consistency was next in importance, followed by accuracy of information. Taken together, these results suggest that injured workers evaluate the faimess of the claims process using all three types of procedural criteria—opportunity to influence the outcome, structural characteristics, and quality of interaction. However, the quality of interaction dimension is far more heavily weighted than the others.

The second question was how faimess of the claims process affected the retum to work outcome. Two exogenous variables that could also affect retum to work

'*rhe formulas for Specht's jt̂ sutistic are; Q = Sum Squared Errors/2[Var(c)]^ where: Var(e) = 1 - r^l[{N - l)^] and rmean = Sum(ABS(ripai(i«i))/# Specified Paths and r^^a&a is the correlatioas for all of the spedfied paths.

"initially, a full model was estimated that induded linkages between the three outcome variables and the global faimess measures. The path coefficients fbr these linkages were assumed to be equal to zero, an assumption that was supported by the empirical test.

'^Each actor's contribution to overall faimess perceptions was evaluated using analysis of variance. Com- bined, the evaluation of the three actors' overall faimess explained 85.4% of the total variation of global faimess (f = 45.49, p = 0.0001). Of the explained variation, AA2% was explained by employer faimess (F = 60J1, p = 0.0001), 33.1% by examiner faimess {F = 45.05, p = 0.0001), and 22.7% by physician faimess (F = 30.88, p = 0.0001). It should be noted that different actots may be more prominent than others in any individual case.

Perceptions of Fairness 207

were included in the model: severity of injury, as measured by whether or not the individual was hospitalized; and the replacement rate, an inverse measure of finan- cial hardship. As expected, severity of injury reduced the likelihood of retum to work. Also as expected, the greater the financial hardship from the reduction in income due to the disability, the greater the likelihood of retum to work. The nega- tive coefficient for global faimess was as expected. It should be interpreted as the fairer the claims process, the more likely a positive retum to work outcome. (This is because the lower the faimess score, the higher the faimess evaluation.)

The third question examined the detemiinants of the decision to dispute the claims outcome. The issue was. How did faimess affect the dispute outcome, through the individual dimensions of faimess, through the global perception of fair- ness, or through the retum to work outcome? The measures of severity of injury and inverse financial hardship were also included in this segment of the model. The coefficient for both of these variables were statistically significant, although the sign for replacement rate was opposite from what was expected, jwrhaps indi- cating that dispute behavior depentls in part on adequacy of financial resources.

The coefficient for retum to work is large, negative, and statistically significant at the 0.01 level, indicating that a negative retum to work outcome increases the likelihot)d of dispute. Although the global faimess of the claims process is important to the retum to work outcome, it does not directly affect the likelihood of dispute. Comparing the insignificant model coefficient with the significant first-order corre- lation in Table I suggests that faimess does affect the likelihood of dispute but its effect is mediated through the retum to work outcome. None of the coefficients for the individual dimensions were statistically significant except that for informa- tion accuracy, which was significant at the 0.1 level. The positive coefficient should be interpreted as indicating that the lower the perceived accuracy of the information with which claims decisions were made, the more likely the claim will be formally disputed.

DISCUSSION

The results of this study suggest that injured workers do base their overall faimess evaluation of the workers' compensation claims process on all three types of criteria, but that the quality of the interaction with decision makers is substan- tially more important than the other types of criteria. This result is consistent with several studies on the universality of procedural justice criteria, where different cri- teria are weighted differently depending on the nature of the outcome at issue and various characteristics of the dispute and dispute resolution process. Shapiro and Brett (1993) found that third-party faimess (the measure that most closely resem- bles quality of interaction in this study) accounted for between 10% and 27% of procedural justice evaluations of dispute resolution processes. These findings are also consistent with those of Tyler (1989), who found that decision-maker neutrality, whether the decision-maker appeared to trust the respondent, and whether the in- dividual's rights and status were respected, were the primary determinants of fair- ness evaluations.

208 Roberts and Youoc

Opportunity to influence the outcome was the next most important variable, but its contribution was one-half that of quality of interaction. Respondents in hi- erarchical organizations where authority fltws down from the top may not expect to have control. Also, most workers, who have not had much experience making a workers' compensation claim, often find the process confusing and intimidating (Roberts & Gleason, 1994).

For consistenqf to be a factor in a faimess evaluation, one must have some basis for comparison. Unless one has had a previous claim or knows someone who has, one is not likely to have well-fomied expectations about workers' compensation (Gleason & Roberts, 1993). Although the experience is analogous to filing a prop- erty claim (for example, auto insurance), there are enough differences that workers' compensation could be perceived as a completely new experience, thus explaining the small coefficient for consistency.

The infonnation accuracy tlimension has the smallest coefficient in the overall fairness regression but is the only form of faimess to be significant in the decision to formally contest the claim. The reason for this result may lie in the nature of the workers' cohipensation formal dispute resolution process. As described earlier, once a case becomes disputed, it dtwely resembles court proceedings where rules of evidence apply. Of the types of faimess dimensions presented in this article, that the decision was not based on correct information is the most commonly accepted reason for formally reviewing a decision.

The finding that the three different types of procedural faimess criteria do affect overall faimess perceptions of the claims process has policy implications for employers. One relationship that is increasingly documented in the faimess litera- ture has been on the relationship between procedural justice criteria and organiza- tional citizenship. Organizational citizenship refers to behaviors that are discretionaiy and not part of explicit job descriptions or recognized by formal re- ward systems, but that promote organizational functioning (Eskew, 1993). This re- search has shown that procedural faimess can affect how people respond to what may be very negative outcomes. In Bies, Martin, and Brockner's (1993) study of behavior after notification of lay-off, an indisputably bad outcome, the faimess of the lay-off procedure significantly affected the quality of organizational citizenship.

The findings here suggest that the quality of interaction between the primary actors and the injured employee, specificaUy along listening, tmsting, and helping dimensions, are important in developing a sense of overall procedural faimess. A policy where the employee is listened to, helped, and treated as a trusted employee will enhance perceptions of faimess and, by extension, organizational commitment. In other words, if the individual is treated as a valued employee by the organization and its agents, that will foster good will toward the organization during the claims prtx^ess.

A second result from this article is that after controlling for extemal conditions related to a workers' compensation claim (severity of injury and financial hardship), faimess of the claims process can affect the retum to work outcome. Who initiates the termination cannot be determined from the data used here. Anecdotal evidence suggests that tennination after filing a workers' compensation claim is more likely to be employer initiated (Roberu & Gleason, 1994). The relationship between pro-

Perceptions of Fairness 209

cedural faimess concems and retum to work is consistent with the results from Youngblood, et al. (1992), who found that procedural faimess concems was the most frequently cited reason for employees fihng an unjust dismissal claim. They also found that interaction justice concems was an important motivator, a result that is similar to the finding here that quality of interaction is an important deter- minant of overall faimess perceptions.

The significant first-order correlation between faimess and dispute, coupled with the insignificant coefficient for faimess in predicting dispute but significant coefficient in predicting retum to work, suggests that overall procedural justice af- fects dispute behavior through the retum to work outcome. This outcome is con- sistent with Daly and Geyer (1994), who found that the effects of certain aspects of process on final behaviors are mediated through perceptions of outcome faimess. In their study of employee reaction to employer decision to relocate, they found that the effect of how decisions were justified on employee intention to stay with their employer was mediated through how fair employees thought it was to relocate the firm.

An additional explanation may be that individuals want to participate in pro- cedurally fair decision-making processes. If the retum to work outcome is seen as the culmination of an unfair process, disputing the outcome may represent an effort to initiate a decision-making process that has the procedural justice characteristics absent from the claims process. Youngblood et al. (1992) find that employees mak- ing unjust dismissal claims cite procedural justice objectives in their reasons for going to a third party, such as wanting to tell the employer how they feel.

Finally, whether the process resulted in the employee being unable to retum to the same employer is the most important factor in initiating a formal dispute. Finding ways to get an employee back on the job is a basic principle of disability management (Tate, Habeck, & Galvin, 1986), and reasonable accommodation is legally mandated under ADA. The results of this article suggest that employers should make every effort to retum an employee with a disability to work. If, how- ever, this is not possible (it imposes an "undue hardship" to use the language of ADA), the degree to which the employee perceives the process as fair according to the infonnation accuracy' criterion can affect the decision to enter a formal dis- pute. In practical terms, this means that from the start of the claim employers should engage in careful record keeping and data gathering, should ensure that the decision-making process is open and clear to the employee with a disability, and should make certain the claims decisions are based on accurate and full information.

Study Limitations

There are several limitations to this study. First, the sample size is relatively small. This may introduce some instability in the path coefficients when this model is tested with other samples. Second, while the sample is representative of the work- ers' compensation population in Michigan, there may be important differences be- tween the respondents and the general work force that would affect these results if the model were applied to a different or more general dispute setting.

210 Roberts mad Young

Thirti, the data are retrospective in that respondents were asked to evaluate experiences that had occurred five years earlier. This was done in order to be certain that the claims of all of the respondents had been completed, so that all respondents were evaluating a completed process. Unfortunately, the cases that become formally tlisputed often t£ike as long as five years to resolution.

Another potential problem with this study is method variance, an artifact of measurement that biases results when relations are explored among constructs measured in the same way (Spector, 1987). Common method variance has been described as the overlap in variance that can be attributed to measurement method rather than to the "true" relationship between variables (Campbell & Fiske, 1959). Two techniques were used in this study to counteract pt»sible common method variance: tlifferent scale formats (Watson, 1988) and reverse-coded items (Gordon & Ladtl, 1990). In addition, the dispute decision and outcome variables were de- rived from a separate tlata source from that used for the faimess evaluations. One piece of evidence that common methods bias may not be a problem with this study is the distinctly different results for information accuracy from those for the other procedural justice scales.

ACKNOWLEDGMENTS

The authors would like to thank Donna Winthrop for her research contribution and Joel Cutcher-Gershenfeld, Dan Ugen, Ed Montemayor, David Wazeter, Susan Zonia, and the participants at the Institute for Public Policy and Social Research for their comments. "This research was supported by the Fund for Research on Dispute Resolution. The opinions expressed herein do not necessarily reflect the position of the Fund.

REFERENCES

Aldiich, J.H., & Nelson, F.D. (1984). Linear Probability. Logii. and Pmbii Models. Beverly Hills, CA: Sage Publications.

Asher, H.B. (1983). Causal Modeling, Second edition. Beverly Hills, CA: Sage Publications. Barrett-Howard, E., & TVler, T. (1986). Procedural justioe as. a criterion in allocation decisions. Joumal

of Personality and Social Psychology, 50, 296-304. Bies, R., & Moag, J. (1986). Interactional justice: Communication criteria of faimess. Research on

Negotiation in Orgartizatiorts, I, 43-55. Bies, R., & Shapiro, D. (1988). Voice and justification: Their influence on procedural faimess judgments.

Academy of Managemenl Joumal, 31, 676-685. Bies, R., Martin, C , & Brockner, J. (1993). Just laid o£f, but stiU a "good citizen?" Only if the process

is fair. Employee Responsibiliiies and kights Journal, 6, 227-238. Boden, L., & V i a o r , R. (1994). Models for reducing workers' compensation litigation. Joumal of Risk

and Insurance, 61, 458-475. Brockner, J., Konovsky, M., Cooper-Schneider, R.. Folger, R., Martin, C , & Bies, R. (1994). Interactive

effects of procedural justice and outcome negativity on victims and survivors of job loss. Academy of Managemeru Joumal, 37, 397-409.

Cami*ell. D., & Fiske, D. (1959). Convergent and discriminant validation by multitrait-multimethod matrix. Psychological Bulieiin, 56, 81-105.

Perceptions of Faimess 211

Daly, J. P., & Geyer, P.D. (1994). The role of faimess in implementing large-scale change: Employee evaluations of process and outcome in seven facility relocations. Journal of Organiiational Behavior, 15, 623-638.

Eskew, D. E. (1993). The role of organizational justice in organizational citizenship behavior. Employee Responsibilities and Rights Journal. 6, 185-194.

Felstiner, W., Abel, R., & Sarat, A. (1980-81). The emergence and tratisformation of disputes; Naming, blaming, and claiming. Law & Society Review. 15, 631-654.

Folger, R., & Greenberg, J. (1985). Procedural justice: An interpretive analysis of personnel systems. Research in Personnel and Human Resources Management. 3, 141-183.

Folger, R., & Konovsky, M. (1989). Effects of procedural justice and distributive justice on reactions to pay raise decisions. Acaderrjy of Management Journal. 32, 115-130.

Folger, R. (1977). Distributive and procedural justice: Combined impact of voice and improvement on experienced inequity. Journal of Personality and Social Psychology, 35, 108-119.

Fullagar, C , & Barling, J. (1989). A longitudinal test of a model of antecedents and consequences of union loyalty. Journal of Applied Psychology, 74. 213-228.

Gleason, S., & Roberts, K. (1993). Worker perceptions of procedural justice in workers' compensation claims: Do unions make a difference? Journal of Labor Research. 14. 45-58.

Gordon, M., & Ladd, R. (1990). Dual allegiance: Renewal, reconsideration, and recantation. Personnel Psychology, 43, 37-69.

Greenberg, J. (1986). Determinants of perceived faimess of performance evaluations. Journal of Applied Psychology. 71, 340-342.

Greenberg, J. (1987). Reactions to procedural injustice in payment distributions: Do the means justify the ends? Journal of Applied Psychology, 72, 55-61.

Greenberg, J. (1990). Organizational justice: Yesterday, today, and tomorrow. Journal of Management. 16, 399-432.

Hunter, J., Gerbing, D.W., Cohen, S.H. & Nicol, T.M. (1980). Package 1980: A System of FORTRAN Routines for the Analysis of Correlational Data. User's manual prepared by Academic Computing Services, Baylor University, Waco, TX.

Konovsky, M., & Cropanzano, R. (1991). Perceived faimes.i of employee drug testing as a prwlictor of employee attitudes and job performance. Joumal of Applied Psychology, 76, 698-707.

Leung, iC, & Li, W.-K. (1990). Psychological mechanisms of process-control effects. Joumal of Applied P^hohgy. 75, 613-620.

Leventhal, G.S. (1980). What should be done with equity theory? In Gergen, K J . , Greenberg, M.S., & Car5on, R.C. (eds.). Social Exchange: Advances in Theory arid Research, 27-55. New York: Plenum.

Leventhal, G.S, Karuza, J., & Fry, W J . (1980). Beyond fairness: A theory of allocation preferences. In Mikula, G. ( e d ) . Justice and Social Interaction. 167-218. New York: Springer Verlag.

Und, E. A, Kanfer, R., & Earley, P.C. (1990). Voice, conUol, and procedural justice: Instrumental and noninstrumental concems in faimess judgments. Joumal of Personality and Social Psychology; 59, 952-959.

Und, E. A.. Kulik, C. Ambrose, M.. & De Vera Park. M. (1993). Individual and corporate dispute resolution: Using procedural faimess as a dedsion heuristic. Administrative Science Quarterly, 38, 224-251.

Und. E. A., & Tyler, T. (1988). The Social Psychology of Procedural Justice. New York: Plenum. M l d s H A , Michigan D e p a r t m e n t of Labor (1985). Compensahle Occupational Injury and Illness

Report—Michigan, 1984. Lansing, MI. Mondak. J. (1993). Institutional legitimacy and procedural justice: Reexamining the question of causality.

Law & Society Review, 27. 599-608. Roberts, K. (1992). Predicting disputes in workers' compensation. Joumal of I&k and Insurance, 59.

252-261. . , . , Roberts K. & Gleason. S. (1994). Procedural justice in the workers' compensation claims process. In

Roj'ot, J., & Wheeler, H. ( e d s ) . Employee Rights and Industrial Justice, 77-78. Deventer. The Netherlands: Kluwer Law and Taxation Publishers.

Shapiro D & Brett, J. (1993). Comparing three processes underlying judgments of procedural jaitice: A field study of mediation and arbitration. Journal of Personality and Social Psychology; 65,

Sheppard B & Minton, J. (1986). Research on procedural justice: Implications for industrial relations. Proceedings of the Thirty Ninth Annual Meeting, 368-374. Industrial Relations Research Association Series, December. . , K« ..i.

Spector, P. (1987). Method variance as an artifact in self-reported affea and perceptioas at work: Myth or significant problem? Joumal of Applied Psychology, 72, 438-443.

212 RobcfU and Yoang

State of Michigan (1988). Worker's Disabilify Compensation Act of 1969 and Administrative Rules. Lansing, Michigan: Department of Labor.

Stone, D. (1984). The Disabled State. Philadelphia, PA: Temple University Press. Tansky, J. (1993). Justice and organizational behavior: What is the Relationship? Employee

Responsibilities and Rights Joumal, 6, 195-207. Tate, D., Habeck, R., & Gahdn, D. (1986). Disability management: Origins, concepts and principles for

practice. Joumal of Applied Rehabilitation Counseling, 17, 5-12. Thibaut, J., & Walker, L. (1975). Procedural Justice- A Psychological Analysis. Hillsdale, NJ: Erlbaum. TVIer, T. (1988). What is procedural justice: Criteria used by citizens to assess the faimess of legal

procedures. Law d Society Review, 22, 301-355. Tyler, T. (1989). The psychology of procedural jtistice: A test of the group-value model. Joumal of

Personality and Social Psychology, 57, 830-838. Tyler, T. (1990). Why People Obey the Lew. New Haven: Yale University Press. Tyler, T., & Und, E.A. (1992). A relational model of authority in groups. Advances in Experimental

Social Psycholoiy, 25, il5-l9Z Watson, D. (1988). The vicissitudes of mood measurement: Effects of varying descnpton, time frames,

and response formats on measures of positive and negative affecL JourrmI of Personality and Social Psychology, 55, 128-141.

Wyatt Corporation (1994). Targeting workers' compensation. Wyatt Communication, 12, 3-13. Youngblood, S., Trevino, L., & Favia, M. (1992). Reactions to unjust dismissal and third-party dispute

resolution: A justice framework. Employee Responsibilities and Rights Joumal, 5, 283-307.