Printer Friendly



Physicians handling workers compensation cases serve in a dual capacity not required of physicians treating non-workers compensation cases. Not only are they are asked to treat the workers' injuries and disability symptoms, but they are also frequently asked to evaluate the extent of economic damage done to a worker via the assignment of an impairment rating to certain types of injuries. Although physicians are trained to treat traumatic injuries as well as the symptoms of cumulative trauma processes, they generally receive no training in disability evaluation. The authors examine physicians' ability to carry out their latter, "lost earnings capacity" assessment, role. Their multivariate regressions indicate that physicians' impairment ratings are poor guides to subsequent wage loss; impairment ratings explain no more than one-half of 1 percent of subsequent wage loss.


Workers' compensation insurance pays both healthcare benefits and disability benefits to workers injured on the job. Physicians handling workers' compensation cases serve in a dual capacity not required of physicians treating non-workers' compensation cases. Not only are they asked to treat the workers' injuries and disability symptoms, but they are also frequently asked to evaluate the extent of economic damage done to a worker via the assignment of an impairment rating to certain types of injuries. Although physicians are trained to treat traumatic injuries as well as the symptoms of cumulative trauma processes, they generally receive no training in disability evaluation. In many states, such as Minnesota, physicians use guidelines to determine an impairment rating that is frequently taken as a proxy for potential wage loss.

Generally the doctor giving the worker a rating is necessarily going beyond his medical training--i.e., the doctor is "out of school"--when making the assessment of damage sustained by the "out-of-work" injured employee. This article addresses whether impairment ratings assessed by physicians are a good guide of lost wages; namely, is a doctor's medical training helpful in assessing economic damage sustained by a claimant?

Following an injury on the job, all workers with lost-time claims (including the most severely injured) initially receive temporary total disability pay to partially compensate for lost wages after a short waiting period. These benefits generally continue until they return to work or, for the severely injured, until the worker's medical condition has stabilized at what is known as the point of maximal medical improvement (MMI). At the point of MMI, workers with residual medical impairments are divided into two groups: those whose injuries preclude any gainful employment are classified as permanent total claims, while those whose injuries allow them to return to work are classified as permanent partial disability claims. [1] Permanent total disabilities are relatively rare, accounting for only about .3 percent of all national worker compensation claims receiving some type of disability pay, but because of their severity, permanent total disabilities account for about 6 percent of total system costs. Permanent partial disabilities, on the other hand, account for 23 percent of national disability claims and 63 percent of total claim costs (Appel and Borba, 1988, p. 4).

Nineteen states, including Minnesota, from which the authors' sample was drawn, use an impairment rating approach to help determine benefits. [2] The impairment rating approach to compensating permanent partial injuries is based on the physical or mental impairment of the claimant at the point of maximum medical improvement. The impairment assessment is generally expressed as a percentage that indicates the extent to which the injury limits motion, sensory perceptions, or physiological function. The assessment is made without regard to the claimant's vocation, age, or skill level. Hence a concert pianist would receive the same workers compensation benefits as the college professor for loss of the left hand, even though the latter's earning capacity was not affected as much by the injury. This is because the award is made prospectively, before actual wage losses are observed.

The data used in this study come from workers compensation claimants in Minnesota, where physicians' assessments are based on the Minnesota Permanency Schedule. This schedule was intended to make more objective the evaluation of whole-body disability associated with particular medical conditions in order to decrease litigation and other friction costs associated with disagreement about disability ratings among the system participants. Under the 1984 Minnesota Permanency Schedule, any condition without a specified rating was rated like the most similarly rated condition (the process in effect during the authors' 1989 sample period).

One complicating aspect of Minnesota permanent partial disability (PPD) benefits was a two-tier reimbursement system in effect during the authors' sample period. The 1983 workers compensation legislation included a two-tier reimbursement formula to determine for PPD benefit amounts. This system stayed in effect until October 1995. Under this system, there are two types of reimbursement: impairment compensation (IC) and economic recovery compensation (ERC). ERC is used when an employer has not offered an injured employee a suitable job within 90 days of that employee having reached MMI. IC benefits are lower than ERC benefits to encourage employers to find suitable employment for their injured workers. This statute led to much dispute about the definition of the suitability of the job offered, both in terms of the work and the wages.

If an employee with a permanency rating was offered a suitable job within 90 days of MMI, the compensation for his or her permanency was determined by the schedule in Table 1, below. The amount was determined by multiplying the number of PPD points by the appropriate dollar amount on the IC table. The money was paid to the employee in a lump sum after 30 days of work were completed successfully. If the employer offered a suitable job, and the employee refused it, the IC amount was awarded for the permanency as determined by Table 1 but given to the worker at the same rate as his or her temporary total disability (TTD) benefits and not in a lump sum.

If no suitable job offer was made to the injured employee within 90 days of reaching MMI, a different method was used to calculate benefits associated with a permanent injury. In this case, economic recovery compensation (ERC) was paid in accordance with Table 2. To determine reimbursement amount under ERC, the percentage of disability was multiplied by the number of weeks of compensation per rating point found in Table 2, then multiplied by the employee's compensation rate at the time of the injury. ERC benefits vary with the injured employee's wages and are significantly higher than IC benefits.

ERC was paid at the same weekly rate as the employee's TTD benefits until the employee returned to work for 30 days. At this point, any remaining ERC benefits were paid in a lump sum to the employee.

Consider a claimant receiving $391 per week, the maximum temporary benefit in 1989, and who--at the point of MMI--receives a 5 percent impairment rating. If the worker received a suitable job offer from his or her employer, then his or her benefit would be calculated according to Table 1: five (the impairment rating) times $750, or $3,750. If the employee did not receive a suitable job offer within 90 days, then his or her award would be equal to five (the impairment rating) times six (from the schedule in Table 2) times $391, or $11,730. From an administrative point of view, the intent of the two-tier system was to encourage firms to re-employ their seriously injured workers. Because of obvious incentive problems, the effect was to engender a lot of litigation over whether a given job offer was "suitable."

The differential between IC and ERC benefits increases with the impairment rating, and this cost differential would be paid by fully experience-rated and self insuring firms. For these firms, the higher impairment rating provides greater re-employment probabilities and a lowered potential wage loss at the margin. This would weaken the link between higher permanency ratings and wage loss. However, there are three reasons the authors believe that the two-tier system will not distort whatever relationship may exist between wage loss and permanency ratings: (1) virtually all permanency ratings ended up being classified as IC ratings, (2) all firms in the residual market have limited experience rating, and so their insurance premiums do not fully reflect ERC and IC cost differentials among their claimants, [3] and (3) about two-thirds of the workers receiving permanency ratings have a rating of less than 7 percent, and only one-fifth of one percent have ratings greater than 25 percent (moving them into the next co st bracket under Tables 1 and 2). ERC ratings generally are only relevant for those with higher ratings. To control for differences in ERC and IC awards, in the empirical research reported below the authors originally included an IC dummy variable in the specification. In the simple specifications (without demographic, industry, occupation, and nature of injury control variables) the IC variable was negative and usually statistically significant, though none of the results changed: the impairment variables did not become more statistically significant, and the overall [R.sup.2] increased very little. In the specifications with the control variables, the IC dummy variable was statistically insignificant. [4]

In 1989, cases with stipulated benefits (contested claims in which a legal settlement between the worker and the insurer had been reached) accounted for 18.2 percent of all indemnity claims filed with the Department of Labor and Industry, while cases with recorded PPDs accounted for 23.6 percent of all cases with more than three days of lost time. Unfortunately, it is not clear how many of the stipulated claims implicitly contained permanency ratings in them. It is likewise unclear whether the distribution of PPDs in stipulated cases followed a similar pattern in terms of rating distributions. It is possible that stipulated claims with particularly high benefit payments had received a PPD rating but did not have the permanency rates written in their workers compensation records. The authors' sample excludes stipulated claims that do not explicitly record a permanency rate. However, because the authors knew which claims were stipulated and without a permanency rating, they could adjust the wage loss regression s using the Heckman sample selection technique in the specifications reported below. The sample selection variable [lambda] turned out to be statistically significant in most of the specifications, but these corrections made no difference to the results reported below and are not reported here. [5]


Research on the relationship between permanency ratings and wage loss is scant. Burton and Vroman (1979) examined the extent of the actual wage replacement under the workers' compensation system for those workers receiving permanent partial disability benefits. They used samples of injured workers from California, Wisconsin, and Florida, and compared the workers' post-injury earnings with those of a control group with similar demographic characteristics. To get wages following the injury, Burton and Vroman used covered wage data available from the Social Security Administration for the specific individuals injured in their sample. Comparing claimant wages to the wages of demographically similar individuals from the social security files, they found large differences across states in the fraction of wages lost (not replaced by benefits): from a high of 26.5 percent in contested cases in California, to a low of 6 percent in uncontested cases in Wisconsin and one-hall of 1 percent in Florida in uncontested cases .

Though Burton and Vroman considered the same types of injuries as those focused on in this analysis, they did not examine how well the impairment ratings tracked lost wages. A recent study by Peterson et al. (1998) on permanent workplace injuries in California has begun to fill this research void. Consistent with the research design used for this study, they obtained wage history data from the California unemployment agency and matched this with workers' compensation administrative records, tracking wages for claimants before and after the injury. They report actual replacement rates varied by the impairment rating (suggesting that impairment ratings are not good predictors of wage loss), and they find that permanency ratings do a particularly poor job of predicting wage loss at lower levels of disability. Some of these issues are examined below for the Minnesota sample.

Although Sinclair and Burton (1995) do not examine permanency ratings and wage loss, they do compare permanency ratings with quality-of-life measures generated from a survey of injured workers in Ontario, Canada. Their analysis was in preparation for a workers compensation statute that sought to compensate workers not only for work disability (and, hence, loss of earnings or earning potential), but also for noneconomic, quality-of-life losses. They asked 12,000 injured workers to benchmark 78 medical conditions covering a variety of impairments. The responses of the injured workers (and a noninjured control group) indicate that: (1) the correlation between the permanency rating and workers' quality of life lost ratings is very low and (2) permanency ratings in Ontario significantly underestimate the quality of life losses perceived by injured workers for the 78 injuries/diseases they examined. The estimates in the next section seek to examine whether permanency ratings also correlate poorly with actual wage l oss.


In the absence of an injury, the worker is assumed to have a weekly wage of WG and work WK number of weeks per year. The physician estimates the fraction of the worker's earning capacity that has been lost because of the injury. This fraction is known as the impairment rating. Wages may be reduced either because the worker, while returning to work full time at WK weeks, cannot do his or her former job and hence will end up with the lower-paying job, or because the worker can return to his or her former job at wage WG but cannot work as much. The authors' "impairment rating" econometric model captures both types of assessments: the impairment rating is assumed to be based partly on how the physician believes that the injury will affect the worker's weekly wage and partly on how the physician believes the injury will affect the number of weeks that the injured claimant can work. For a given impairment rating IMPR, the weekly wage is assumed to be reduced by the fraction [phi] IMPR (where [phi] is a constant, so that the observed wage after the injury would be WG[1-[phi] *IMPR]), and the number of weeks worked would be reduced by the fraction IMPR (where [gamma] is another constant, so that the observed weeks worked after the injury would be WK[1 - [gamma]IMPR]). Since lost wages are the difference between pre-injury wages and post-injury wages, the percentage reduction [6] in lost wages for the "average worker" would be

ln(WG * WK) - ln(WG[1 - [phi]IMPR] * WK[1 - [gamma]IMPR]), (1)

where WG*WK is the average wages before the injury and

WG[1 - [phi]IMPR] * WK[1 - [phi]IMPR]

is the post-injury wages for the worker with an impairment rating of IMPR. Subtracting terms on the right side of Equation (1), and assuming that 1n(1 + X) is approximately X for values of X close to zero, [7] then the percent reduction in wages for the average worker becomes

-ln([1 - [gamma] IMPR][1 - [phi] IMPR]) = ln(1 - ([gamma] + [phi]) * IMPR + ([gamma] [phi]) * [IMPR.sup.2])

= ([gamma] + [phi]) * IMPR - ([gamma] [phi]) * [IMPR.sup.2]. (2)

Note that if the impairment rating is based on just lost weekly wages, is based just on lost weeks worked, or is based just on the total wages lost (so that there is no implicit separate evaluation for reduced hourly earning potential and reduced work stamina), then the second right-side term drops out of Equation (2) and the percent reduction in wages would be a linear function of the impairment rating. In the more general model, with both the reduction in the weekly wage rate and the weeks worked being evaluated separately, lost wages will be a quadratic function of wages. If lost wages are regressed on the impairment rating and its squared value (IMPR, [IMPR.sup.2]) then the model should provide a fairly good fit, with higher impairment ratings associated with greater lost earnings.

Although Equation (2) models the typical worker's wage after an injury, some workers will get more or fewer wages than the typical worker. Though we know the workers' pre-injury wage, not all workers are employed for the typical WK number of weeks. To complete the model, each individual worker is assumed to vary from the typical, fulltime schedule according to a random variable factor of proportionality, [micro]. Since this factor of proportionality captures differences in pre-injury variations in work weeks, it should be uncorrelated with the impairment rating, IMPR. Thus, Equation (1), adjusting for variations in work weeks across individuals, becomes

ln(WG * WK * [micro])-ln(WG[1 - [phi]IMPR] * WK[1 - [gamma]IMPR] * [micro]) (3)

Given the model of the impairment process in the last term, Equation (2) indicates that the right-side difference will be a quadratic function of the impairment rating if physicians are accurately pre-assessing wage loss through the impairment rating system.

The term WG [1- [phi]IMPR] * WK [1- [gamma]IMPR] * [micro] is the observed wage loss, WG is the observed pre-injury weekly wage, and WK for a full-time worker is approximately 50 weeks. The unobserved random variable, [micro], is the factor of proportionality that accounts for variation across individuals in weeks worked (as well as deviations from the observed pre-injury wage from a worker's usual pre-injury wage). Then the specification for wage loss becomes

ln(WG * WK)-ln(WG[1 - [phi]IMPR] * WK[1 - [gamma]IMPR] * [micro]) =

[[beta].sub.0] + [[beta].sub.1]IMPR + [[beta].sub.2][IMPR.sup.2] + ln([micro]). (4)

Since [micro] is log-normally distributed, then ln([micro]) will be normally distributed. [8] The specification in Equation (4) indicates that the log difference in pre- and post-injury wages should be modeled as a quadratic regression with IMPR and [IMPR.sup.2] as the independent variables. [9] If the impairment rating process is accurately assessing wage loss, then the following should be observed:

1. The overall fit of the regression equation should be significant, with the [R.sup.2] indicating how much of the variation in wage loss is explained by the impairment rating process, and

2. The IMPR and [IMFR.sup.2] coefficients should be statistically significant and quantitatively important.

This is the test the authors make in the next section.


Before examining the estimated regression function corresponding to Equation (4), wage losses and the wage loss ratios are presented in Table 3.

As indicated above, lost wages in Table 3 are the difference between the potential annual earnings of the injured workers defined as their pre-injury weekly wage multiplied by 50 (assuming they are paid for 50 weeks per year on average), and the actual wages they received after their injury. The actual wages were gathered from the quarterly wage data supplied to the Minnesota Department of Employment Security by the injured worker's respective employer. Hence, for the data in the first left column of Table 3, four quarters of workers' wages were collected beginning in the quarter after the reported injury. [10] For this study, the authors included data on all workers with impairment ratings in 1989, with their respective wages reported for the next four years (through 1993).

For example, suppose a worker reported an injury occurring on December 15, 1989, and he or she had a pre-injury wage of $400 per week. His or her potential wages would be calculated as 50*$400 = $20,000 for the next year [this is the WK*WG term in Equations (1) - (4)]. This is compared with the individual's actual wages, as reported by his or her employer(s), for the next four quarters, beginning in the first quarter of 1990. If the individual's actual wages for this period were $18,500, then we would record his or her lost wages as $1,500, and the "lost wages/potential wages" ratio would be 0.075. If he or she had no recorded earnings for the next four quarters, then lost earnings would be recorded as $20,000 with a "lost wages/potential wages" ratio of 1. Lost earnings were recorded for the first four quarters after the injury in the "One year after" column, for the first 8 quarters after the injury in the "Two years after" column, etc., so that the losses were cumulative. All data were adjusted for inflati on and were reported in 1982 through 1984 dollars.

One important pattern revealed in Table 3 is the relative constancy in lost earnings for those with impairment ratings under 25 percent. The absolute losses incurred each year are roughly the same as the first year's losses. In the 5-10 percent impairment group, for example, $3,944 are lost in wages for the first four quarters after the injury. An additional $3,537 is lost on average in the second year ($7,481 - $3,944); $3,780 in the third year ($11,261 - $7,481), and $3,735 in the fourth year ($14,996 - $11,261). The pattern holds for the other groups as well, and it is reflected by the relative stability of the fraction of lost wages relative to potential wages within an impairment rating group over time. This is expected since the impairment rating is made after MMI so that significant deterioration in a worker's health would not be expected. Only for those 12 cases in which the impairment rating is more than 25 percent is there a substantial increase in the fraction of lost wages over time (from 43 perce nt lost wages in the first year to a cumulative 54 percent by the end of the fourth year).

The other important pattern illustrated in Table 3 is that lost wages do not seem to increase significantly as the impairment rating increases. If the impairment ratings doctors gave were an accurate reflection of wage lost potential, then one might expect to see a doubling of lost wages when going from the 0-5 percent impairment group to the 5-10 percent impairment group and a tripling when going from the 0-5 percent impairment group to the 10-15 percent impairment group. Instead, wage losses increase only modestly with higher impairment ratings, and they even fall when going from the 5-10 percent group ($3,944) to the 10-15 percent group ($3,467). This strongly suggests that the impairment ratings are relatively poor predictors of lost wages.

However, the results in Table 3 do not prove that impairment ratings are poor predictors of subsequent wage loss for two reasons. First, the simple descriptive statistics do not control for other demographic, industry, and occupational effects. Second, since most of the awards are IC awards that depend on the permanency assessment but not on actual level of wages, then only one dimension of the "wage loss" is really being assessed in most cases: namely, expected weeks off the job, which is the [gamma]-type loss in Equations (1) through (4) above. In the econometric specifications in Table 4 and Appendix Tables 1 and 2, the authors addressed these problems by including other control variables in the regression specification, including pre-injury wage. [11]

Table 4 indicates that impairment ratings are, indeed, relatively poor predictors of lost wages for injured workers, even after all other factors are held constant. In Table 4, the Equation (4) specification results are reported in columns 1,3,5, and 7 where the difference between the log of potential wages and log of actual wages is regressed only on the impairment rating, the impairment rating squared, and the pre-injury wage (to account for the two-tiered compensation scheme). The [R.sup.2] statistics reported in the fourth from the last row indicate that the impairment ratings do not explain more than one-half of 1 percent of the actual variation in the lost earnings across injured workers, despite the presence of the pre-injury wage as an additional regressor.

The impairment ratings coefficients are individually insignificant in the "Losses First 4 Years After" regressions and only marginally significant for IMPR (but not IMPR [2]) in the other three regressions. Because of the large sample size, the impairment ratings are jointly statistically significant in all eight regressions in Table 4, as indicated by the F-test statistics reported in the next-to-last row. [12]

The results in Table 4 are the most favorable that the authors obtained. In Appendix Table 1, an alternative and equally reasonable specification of the lost earnings are estimated with the ratio of lost earnings to potential earnings as the dependent variable. The [R.sup.2] values for these models are always lower than even the low values in Table 4 and the estimated IMPR or [IMPR.sup.2] coefficients even less statistically significant.

What does explain wage loss? In columns 2, 4, 6, and 8 of Table 4 (and in Appendix Table 1), the basic specification in Equation (4) has been supplemented with the type of demographic data typically found in the worker's compensation files, including: basic demographic information, industry, occupation, and type of injury. The inclusion of these other factors improves explained variation in lost wages from less than one-hail of i percent to between 2.5 and 5 percent, but this still leaves most of the variation in lost wages unexplained. [13] Among the significant factors in explaining losses are age, gender, [14] the benefit/wage replacement ratio, and some occupational and industry dummy variables.


Using a detailed, objective standard for impairment ratings such as the American Medical Association's Guides to the Evaluation of Permanent Impairments is alleged to offer the advantage of reducing variance in permanency ratings across physicians by providing a uniform guide to rating permanent partial injuries. This seems plausible to the authors, as the use of a detailed standard reduces the influence of external factors such as geography and age. But even if detailed guides improved the reliability of the rating system in the sense that those using the rating scheme generally come up with the same permanency rate when faced with the same impairment, this does not mean that they provide an "accurate" assessment of workers' actual loss of functional use of their injured body parts nor of workers' potential loss in wage-earning capacity. That is, this study is not about the reliability of such schemes but about their validity. In particular, how well did the Minnesota permanency guidelines predict future wag e loss? Even if all physicians now come up with the same rating for the same impairment, does that rating indicate much about the actual wage loss? The answer, apparently, is no.

What this research shows is that the use of detailed guidelines in Minnesota has resulted in impairment ratings that explain less than one-half of 1 percent of the variation in lost wages, whether those wages are measured as (log) differences or the ratio of lost wages to potential wages. When the authors controlled for worker demographics, industry, worker occupation, and type of injury, the explanatory power of the impairment ratings was further reduced. And there is no reason to suppose that Minnesota physicians, accustomed for a long time to managed-care initiatives and guidelines, use guidelines any less well than do physicians in any of the other states. That is, there is no reason to suppose that the same problem with validity does not plague other states relying chiefly on impairment ratings in determining PPD benefits.

This suggests that, to the extent that they are meant to capture wage loss, impairment rating systems provide little useful information to workers compensation administrators and the workers compensation system. Critics of the guidelines abound, noting that they often reveal little about disability [15] and offer flawed promises of objectivity [16] (i.e., they are not valid indicators of wage loss). Our evidence from Minnesota supports those views.

The authors' evidence further indicates that the hybrid "loss of wage-earning capacity" approach does better at predicting wage loss, inasmuch as it adjusts the medical impairment ratings for such factors workers' demography and job circumstances. That is, regressors controlling for gender, age, industry, and job type increase the explanatory power of the model tenfold (however, this is not a particularly strong statement given the extraordinarily poor explanatory power of the impairment ratings). California employs such a system, and this might explain why the Rand study (Peterson et al., 1998) seemed to find slightly more explanatory power in its permanency ratings than the authors do with the Minnesota sample.

Clearly, if only "statistical reliability" matters, then impairment rating systems--such as those employed in Minnesota--may work fine. However, to the extent that impairment ratings are meant to reflect potential wage loss, pure impairment rating systems do not provide valid guidelines.

Yong-Seung Park is assistant professor of business administration at Kyung Hee University, Korea. Richard J. Butler is professor of economics at Brigham Young University in Provo, Utah. The authors are grateful for computing support and comments on earlier drafts of this article by the Minnesota Department of Labor and Industry, especially those of Kate Kimpan. Anonymous referees also made excellent suggestions on earlier drafts of the article.

(1.) Those claimants whose injuries are classified as permanent partial do not all return to work. Some of the permanent partial claimants collect Social Security Disability Insurance benefits.

(2.) For more information on how permanent partial injuries are compensated, including alternatives to the impairment rating system, see Burton (1997), Welsh (1994), and WCRI (1999).

(3.) In Minnesota, the residual market, formerly known as the Assigned Risk Pool, covers approximately 40 percent of the employers and accounts for about 20 percent of the employees. Experience rating for these firms limits premium deviations to no more than 10 percent, even if a firm's costs rise (or fall) by considerably more than 10 percent.

(4.) Results are reported below without the IC dummy. Results including the IC dummy are available on request.

(5.) These sample-selection adjusted results, including adjustments of the standard errors in the second stage for the selection process in the first stage, are given in the data appendix available on request. (See Heckman, 1979, and Greene, 1981, on adjustments for sample selection.)

(6.) The authors use the following properties of logarithms (In): that the percent difference between 2 numbers, X and Y, is in X -- InY; and that for values of X close to zero, ln(1 + X) = X.

(7.) As Table 1 indicates, most impairment ratings fall into this category, and virtually none are greater than 25 percent. Although the authors place no restrictions on the constants [gamma] and [phi], they generally expect their values to lie between zero and one. Hence, the approximation here should be fairly reasonable.

(8.) In formal tests for normality of the residuals for the specifications in Table 4, the authors used a Kolmogorov D-statistic. Not unsurprisingly, given their large sample, the D-statistic rejected the normality assumption.

(9.) In the empirical specification, the authors assume that the average worker is employed the equivalent of 50 weeks per year. The inclusion of the intercept allows the actual number of weeks to deviate from that assumed value, making the results robust relative to that assumption.

Note that the error term may also capture deviations in wages for those who appear to have no wages in the Department of Employment Security data set even though they left the state and are working elsewhere. For this sample of severely injured workers, given the generosity of the Minnesota workers compensation relative to the surrounding states (Minnesota pays significantly higher benefits than any of the contiguous states), and the very low unemployment rates during this period, the authors do not expect sample attrition to be a significant problem.

(10.) Since the number of days from the injury rate to the beginning of the first quarter of post-injury wages varied from zero to 93, the authors included a control variable measuring this difference in the model specification. It was generally insignificant and made no difference to the results reported below. Hence, the authors excluded this control variable from the results reported below.

(11.) Results without the pre-injury wage are similar in both the quantitative and statistical significance.

(12.) As referees pointed out, the results would be even stronger it the authors had additional wage data before the injury (quarterly data reported by the employers) and a longer time series after the injury. Their advice is excellent, and the authors hope that future researchers will gather such data for their analyses. In the present case, the 1989 claims data and the 1990 through 1993 wage data were all that the authors could get for this article.

(13.) Also note that the significance of the IMPR, [IMPR.sup.2] coefficients is not improved by the inclusion of other factors.

(14.) The authors partitioned the data by gender and ran all the models in Table 4 and Appendix Table 1 separately for males and females to see whether the fit or statistical significance improved. They did not. The coefficients' remains were less significant than those reported here, and the R-squares were no greater.

(15.) Norton Hadler (1992) argues that "Adhering to the [American Medical Association's] Guides in attempting to quantify impairment is in my opinion an unappealing, if not Orwellian, exercise, and not just for musculoskeletal diseases but for all diseases. Even the authors of the Guides discourage extrapolations from impairment to disability or handicap. Here I am in accord with this publication; the numbers generated by following the AMA Guides offer little, if any, insight into disability. It follows that this tedious, often expensive, stultifying examination is a sophism."

(16.) Ellen Smith Pryor (1990) observes: "The Guides is not the objective, medical evaluative system that it purports to be and that has been so appealing to legislators and other decision makers. Instead, like any impairment rating scheme, it rests in large part on important and difficult normative judgments. Yet the Guides obscures this from the reader; it is laden with hidden or poorly explained value judgments that frequently are gender based. The Guides flawed promises of objectivity are especially troubling because they appeal to the craving of legislators and other decision makers for certainty and clarity in the difficult arena of impairment and disability assessment."


Appel, David, and Philip S. Borba, 1988, Workers' Compensation Insurance Pricing (Boston: Kluwer Academic Publishers).

Burton, John F., Jr., 1983, Compensation for Permanent Partial Disabilities, in: John D. Worrall, ed., Safety and the Workforce: Incentives and Disincentives in Workers' Compensation (Ithaca, N.Y.: ILR Press).

Burton, John F., Jr., 1997, Permanent Partial Disability Benefits: A Reexamination, in: John F. Burton, Jr., and Timothy P. Schmidle, eds., 1998 Workers' Compensation Yearbook (Horsham, Pa.: LRP Publications).

Burton, John F., Jr., and Wayne Vroman, 1979, A Report on Permanent Partial Disabilities Under Workers' Compensation, in: Research Report of the Interdepartmental Workers' Compensation Task Force 6 (Washington, D.C.: U.S. Government Printing Office), 11-77.

Hadler, Norton M., 1992, Impairment Rating in Disability Determination for Low Back Pain: Placing the AMA Guides and the Quebec Institute Report into Perspective, in: John F. Burton, Jr., and Timothy P. Schmidle, eds., Workers' Compensation Desk Book (Horsham, Pa.: LRP Publications, 1992), 1129-1133.

Heckman, James J., 1979, Sample Selection Bias as a Specification Error, Econometrica, 47:153-161.

Green, William, 1981, Sample Selection Bias as a Specification Error: Comment, Econometrica, 49:795-798.

Peterson, Mark A., Robert T. Reville, Rachel Kaganoff Stern, and Peter S. Barth, 1998, "Compensating Permanent Workplace Injuries: A Study of the California System," Rand Institute for Civil Justice, report #MR-920-ICL.

Pryor, Ellen Smith, 1990, Flawed Promises: A Critical Evaluation of the American Medical Association's Guides to the Evaluation of Permanent Impairment, Harvard Law Review, 103:964-976.

Sinclair, Sandra, and John F. Burton, Jr., 1995, Development of a Schedule for Compensation of Noneconomic Loss: Quality-of-Life Values vs. Clinical Impairment Ratings, in: Terry Thomason and Richard P. Chaykowski, eds., Research in Canadian Workers' Compensation (Kingston, Ontario: IRC Press, Industrial Relations Centre).

Welch, Edward M., 1994, Employer's Guide to Worker's Compensation (Washington, D.C.: BNA Books).

WCRI, 1999, Permanent Partial Disability Benefits: Interstate Differences (Cambridge, Mass.: Workers Compensation Research Institute).
 Impairment Compensation (IC) Schedule
Percentage of Disabilitiy $ Amount per Point
 0-25 $750
 26-30 $800
 31-35 $850
 36-40 $900
 41-45 $950
 46-50 $1,000
 51-55 $1,200
 56-60 $1,400
 61-65 $1,600
 66-70 $1,800
 71-75 $2,000
 76-80 $2,400
 81-85 $2,800
 86-90 $3,200
 91-95 $3,600
 96-100 $4,000
 Economic Recovery Compensation
 (ERC) Schedule
Percentage of Disability Weeks of Compensation per Point
 0-25 6.0
 26-30 6.4
 31-35 6.8
 36-40 7.2
 41-49 7.6
 50 8.0
 51-55 8.8
 56-60 9.6
 61-65 10.4
 66-70 11.2
 71-100 12.0
 Losses by Permanent Partial
 Impairment Ratings
 Lost Wages
 Lost Wages/Potential Wages
 [Number of Claimants]
Impairment One Year Two Years Three Years Four Years
Rating After Injury After Injury After Injury After Injury
0-5 percent $3,085 $6,467 $10,001 $13,817
 0.16 0.17 0.17 0.18
 [3,513] [3,513] [3,513] [3,513]
5-10 percent $3,944 $7,481 $11,261 $14,996
 0.20 0.19 0.19 0.19
 [1,497] [1,497] [1,497] [1,497]
10-15 percent $3,467 $6,913 $10,827 $15,185
 0.17 0.17 0.18 0.19
 [723] [723] [723] [723]
15-20 percent $4,231 $7,421 $11,032 $15,764
 0.23 0.21 0.20 0.22
 [125] [125] [125] [125]
20-25 percent $4,535 $7,598 $10,738 $13,876
 0.25 0.21 0.20 0.19
 [43] [43] [43] [43]
25+ percent $8,970 $18,935 $31,011 $44,799
 0.43 0.46 0.50 0.54
 [12] [12] [12] [12]
 Impairment Ratings as Predictors of Wage Loss
 Dependent Variable: Log (Potential Wages)
 - Log (Post-Injury Wages) (Absolute Value
 of f-statistic in Parentheses)
Variables Losses First Losses First
 Year After 2 Years After
INTERCEPT 0.728 [***] -0.473 0.678 [***] -0.503
 (9.84) (1.32) (9.55) (1.46)
IMPR 0.047 [***] 0.021 0.041 [**] 0.030)
 (2.72) (0.94) (2.46) (1.38)
IMPR [**]2 0.000 0.002 -0.000 0.000
 (0.13) (1.53) (0.23) (0.45)
AWW 0.000 0.000 [**] 0.000 0.000 [**]
 (1.09) (2.51) (1.38) (2.54)
AGE -- 0.005 -- 0.011 [***]
 (1.40) (3.13)
MALE -- 0.293 [***] -- 0.280 [***]
 (2.70) (2.68)
BEN/WAGE -- 0.022 [***] -- 0.172 [***]
 (6.61) (5.42)
Industry Groupings
AGR -- -0.963 [***] -- -0.809 [***]
 (6.22) (5.45)
MIN -- -0.815 [***] -- -0.649 [***]
 (5.13) (4.25)
MANU -- -0.743 [***] -- -0.629 [***]
 (3.97) (3.50)
TRAN -- -0.571 [***] -- -0.470 [***]
 (3.50) (3.00)
TRAD -- -0.627 [**] -- -0.487 [*]
 (2.06) (1.67)
Variables Losses First Losses First
 3 Years After 4 Years After
INTERCEPT 0.674 [***] -0.604 [*] 0.702 [**] -0.753 [***]
 (9.78) (1.80) (10.36) (2.28)
IMPR 0.033 [**] 0.024 0.023 0.017
 (2.04) (1.56) (1.48) (0.84)
IMPR [**]2 0.000 0.000 0.000 0.000
 (0.04) (0.47) (0.41) (0.62)
AWW 0.000 0.000 [***] 0.000 0.000 [***]
 (1.62) (2.62) (1.72) (2.66)
AGE -- 0.017 [***] -- 0.021 [***]
 (4.95) (6.35)
MALE -- 0.245 [**] -- 0.246 [**]
 (2.41) (2.46)
BEN/WAGE -- 0.015 [***] -- 0.014 [***]
 (4.74) (4.62)
Industry Groupings
AGR -- -0.725 [***] -- -0.681 [***]
 (5.02) (4.78)
MIN -- -0.547 [***] -- -0.513 [***]
 (3.68) (3.50)
MANU -- -0.569 [***] -- -0.554 [***]
 (3.26) (3.21)
TRAN -- -0.405 [***] -- -0.413 [***]
 (2.66) (2.76)
TRAD -- -0.384 -- -0.366
 (1.35) (1.31)
FIN -- -0.363 [*] -- -0.204 -- -0.070 --
 (1.73) (1.02) (0.36)
SERV -- -0.395 [**] -- -0.261 -- -0.229 --
 (2.03) (1.40) (1.26)
ADMIN -- -1.240 [***] -- -1.009 [***] -- -0.925 [***] --
 (4.43) (3.76) (3.54)
Occupation of the Claimant
PROF/TECH -- 0.340 -- 0.191 -- 0.109 --
 (1.45) (0.85) (0.50)
MANAGER -- -0.041 -- -0.021 -- 0.001 --
 (0.22) (0.12) (0.00)
CLERICAL -- -0.237 -- 0.314 -- -0.334 [*] --
 (1.18) (1.63) (1.78)
CRAFT -- -0.322 [**] -- -0.393 [***] -- 0.373 [***] --
 (2.45) (3.12) (3.05)
OPERATOR -- -0.149 -- -0.207 -- -0.174 --
 (1.07) (1.55) (1.34)
TRANS -- -0.129 -- -0.266 -- -0.265 --
 (0.68) (1.46) (1.50)
SERVICE -- -0.041 -- -0.131 -- -0.153 --
 (0.24) (0.80) (0.95)
Type of the Injury
FRAC -- 0.153 -- 0.163 -- 0.193 --
 (1.06) (1.18) (1.44)
BKSTR -- -0.294 [**] -- -0.283 [**] -- -0.272 [**] --
 (2.17) (2.19) (2.16)
OTHSTR -- 0.057 -- 0.005 -- -0.004 --
 (0.51) (0.05) (0.03)
FIN -0.030
SERV -0.306 [*]
ADMIN -0.953 [***]
Occupation of the Claimant
MANAGER 0.078)
CRAFT -0.362 [***]
TRANS -0.275
SERVICE -0.091
Type of the Injury
FRAC 0.227 [*]
BKSTR -0.270 [**]
OTHSTR 0.017
CONTUSION -- -0.273 -- -0.260
 (1.32) (1.32)
CUT -- 0.019 -- 0.126
 (0.13) (0.91)
[R.sup.2] 0.0052 0.0300 0.0035 0.0252
F-value 10.282 [***] 6.640 [***] 9.923 [***] 5.539 [***]
F-IMPR, [IMPR.sup.2] 14.706 [***] 14.839 [***] 9.358 [***] 8.307 [**]
Sample Size 5,909 5,600 5,909 5,600
CONTUSION -- -0.190 -- -0.147
 (0.99) (0.77)
CUT -- 0.178 -- 0.218
 (1.32) (1.64)
[R.sup.2] 0.0031 0.0243 0.0026 0.0261
F-value 6.161 [***] 5.331 [***] 5.169 [***] 5.739 [***]
F-IMPR, [IMPR.sup.2] 7.866 [**] 6.488 [**] 6.223 [**] 5.200 [*]
Sample Size 5,909 5,600 5,909 5,600
Note: (***.)significant at the 1 percent level;
(**.)significant at the 5 percent level;
(*.)significant at the 10 percent level
 Impairment Ratings as Predictors of Wage Loss
 Dependent Variable: Post-Injury
 Wage Loss/"Potential Wages" (Absolute
 Value of t-statistic in Parentheses)
Variables Losses First Year After
INTERCEPT 0.097 [***] 0.231 [***]
 (7.73) (3.49)
IMPR 0.006 [*] 0.001
 (1.89) (0.21)
IMPR [**]2 0.000 0.000 [**]
 (0.84) (2.17)
AWW 0.000 0.000 [**]
 (1.45) (2.00)
AGE -- 0.000
SEX -- -0.030
BEN/WAGE -- 0.000
Industry Groupings
AGR -- -0.226 [***]
MIN -- -0.175 [***]
MANU -- -0.222 [***]
TRAN -- 0.096 [***]
TRAD -- -0.194 [***]
Variables Losses First 2 Years After
INTERCEPT 0.105 0.264 [***]
 (8.16) (4.27)
IMPR 0.001 -0.002
 (0.40) (0.58)
IMPR [**]2 0.000 0.000 [**]
 (1.54) (2.19)
AWW 0.000 [*] 0.000 [*]
 (1.83) (1.80)
AGE -- 0.002 [***]
SEX -- -0.032 [*]
BEN/WAGE -- -0.000
Industry Groupings
AGR -- -0.203 [***]
MIN -- -0.146 [***]
MANU -- -0.214 [***]
TRAN -- -0.087 [***]
TRAD -- -0.180 [***]
Variables Losses First 3 Years After
INTERCEPT 0.112 [***] 0.239 [***]
 (8.55) (3.80)
IMPR -0.001 -0.004
 (0.19) (1.03)
IMPR [**]2 0.000 [*] 0.000 [**]
 (1.77) (2.20)
AWW 0.000 [*] 0.000
 (1.94) (1.63)
AGE -- 0.004 [***]
SEX -- -0.022
BEN/WAGE -- -0.001 [**]
Industry Groupings
AGR -- -0.190 [***]
MIN -- -0.138 [***]
MANU -- -0.215 [***]
TRAN -- -0.087 [***]
TRAD -- -0.191 [***]
Variables Losses First 4 Years After
INTERCEPT 0.118 [***] 0.218 [***]
 (8.78) (3.39)
IMPR -0.001 -0.005
 (0.35) (1.19)
IMPR [**]2 0.000 [*] 0.000
 (1.76) (2.10)
AWW 0.000 [*] 0.000 [**]
 (1.96) (1.49)
AGE -- 0.006 [***]
SEX -- -0.022
BEN/WAGE -- -0.001 [**]
Industry Groupings
AGR -- -0.191 [***]
MIN -- -0.141 [***]
MANU -- -0.225 [***]
TRAN -- -0.100 [***]
TRAD -- -0.206 [***]
FIN -- -0.112 [***] -- -0.105 [***] -- -0.104 [***] --
 (3.19) (2.91) (2.81)
SERV -- -0.144 [***] -- -0.149 [***] -- -0.144 [***] --
 (4.40) (4.45) (4.21)
ADMIN -- -0.035 [***] -- -0.354 [***] -- -0.349 [***] --
 (7.37) (7.34) (7.10)
Occupation of the Claimant
PROF/TECH -- -0.086 [**] -- -0.081 [**] -- -0.102 [**] --
 (2.19) (2.01) (2.48)
MANAGER -- -0.124 [***] -- -0.109 [***] -- -0.108 [***] --
 (4.04) (3.45) (3.36)
CLERICAL -- -0.057 [*] -- -0.058 [*] -- -0.083 [**] --
 (1.70) (1.68) (2.34)
CRAFT -- -0.059 [***] -- -0.068 [***] -- -0.075 [***] --
 (2.67) (2.98) (3.25)
OPERATOR -- 0.000 -- -0.011 -- -0.019 --
 (0.00) (0.48) (0.77)
TRANS -- 0.018 -- 0.008 -- 0.012 --
 (0.57) (0.24) (0.36)
SERVICE -- -0.041 -- -0.041 -- -0.050 [*] --
 (1.42) (1.38) (1.66)
Type of the Injury
FRAC -- 0.031 -- 0.019 -- 0.012 --
 (1.27) (0.76) (0.48)
BKSTR -- -0.015 -- -0.015 -- -0.013 --
 (0.67) (0.64) (0.55)
FIN -0.127 [***]
SERV -0.150 [***]
ADMIN -0.361 [***]
Occupation of the Claimant
PROF/TECH -0.124 [***]
MANAGER -0.104 [***]
CLERICAL -0.104 [***]
CRAFT -0.075 [***]
TRANS 0.014
SERVICE -0.047
Type of the Injury
FRAC 0.003
BKSTR -0.011
OTHSTR -- 0.003 -- -0.000
 (0.17) (0.02)
CONTUSION -- 0.057 -- -0.061 [*]
 (1.64) (1.72)
CUT -- -0.022 -- 0.006
 (0.90) (0.22)
[R.sup.2] 0.0047 0.0330 0.0028 0.0274
F-value 9.202 [***] 7.306 [***] 5.497 [***] 6.045 [***]
F-IMPR, [IMPR.sup.2] 12.771 [***] 14.300 [***] 6.496 [**] 7.227 [**]
Sample Size 5,909 5,600 5,909 5,600
OTHSTR -- 0.005 -- 0.010
 (0.25) (0.50)
CONTUSION -- -0.060 [*] -- -0.057
 (1.66) (1.53)
CUT -- 0.010 -- 0.012
 (0.39) (0.46)
[R.sup.2] 0.0028 0.0312 0.0020 0.0390
F-value 4.429 [***] 6.896 [***] 3.935 [**] 8.703 [***]
F-IMPR, [IMPR.sup.2] 4.745 [*] 4.707 [*] 3.912 3.498
Sample Size 5,909 5,600 5,909 5,600
Note:(***.)significant at the 1 percent level;
(**.)significant at the 5 percent level;
(*.)significant at the 10 percent level
COPYRIGHT 2000 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2000 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Park, Yang-Seung; Butler, Richard J.
Publication:Journal of Risk and Insurance
Article Type:Brief Article
Geographic Code:1USA
Date:Sep 1, 2000
Previous Article:Nonprofit Compensation and Benefits Practices, by Carol L. Barbeito and Jack P. Bowman (John Wiley & Sons Inc. 1998).

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters