We examined associations between workforce demographics and job characteristics, grouped by industrial sector, and declines in workers’ compensation claim rates in Ontario, Canada, between 1990 and 2003. Gender, age, occupation, and job tenure were predictors for claim rates in 12 industrial sectors. The decline in claims was significantly associated with a decline in the proportion of employment in occupations with high physical demands. These findings should generate interest in economic incentives and regulatory policies designed to encourage investment in safer production processes.
Compensation claim rates for days of lost work as a result of work-related injury or illness have declined in North America and Europe.1–5 Previous research regarding this general decline has examined the influence of disproportionate employment growth in the service sector relative to the goods-producing sector on the hypothesis that service sector development has resulted in proportionally fewer hazardous jobs.5,6
Although important, these explanations do not account for the declines in claim rates experienced by high-hazard industries and observed by Conway and Svenson,5 who did not examine actual numbers of hazardous jobs within these industries. Structural changes within industries, such as technological advances in production methods (reducing hazard exposure), changes in government regulatory practices, and improvement in safety management at the workplace level (to increase economic competitiveness), might help account for declines in claim rates within specific industries.1
Changes in workforce demographic composition (e.g., age, gender) within industries also may influence claim trends. For example, the greater experience and skill that older workers possess compared with young workers would be expected to lead to lower injury risk as the workforce ages.7 However, work-force compositional changes within an industry also may be correlated with structural changes. Shifts in the age distribution in an industry may be associated with changes in work arrangements (e.g., more temporary work for youths).
We used workers’ compensation claims for days of lost work as a result of work-related injury or illness and employment patterns from Ontario during 1990 to 2003 to examine relations between claim rates and work-force composition and job characteristics within and across industrial sectors.
Workers’ compensation claims for occupational injury or illnesses leading to 1 or more days of lost work for calendar years 1990 to 2003 were obtained from the Ontario Work-place Safety and Insurance Board. Each record contained demographic information, hire date, occupation, and industry (according to the 1980 Canadian Standard Industrial Classification for Companies and Enterprises) for each worker with a claim.8
The unit of analysis was yearly claim rates for days of lost work per 100 full-time equivalents (2000 hours = 1 full-time equivalent) for each industrial sector by year. Annual estimates of the number of workers and work hours for each industry were derived from the Canadian Labour Force Survey public use files.9–11
The predictors of interest were industry-and year-specific percentages of workers stratified by age group (15–24 years, 25–49 years, or 50 years or older), gender, job tenure (less than 12 months on the job, 12–47 months, or 48 months or longer), and occupational physical demands. The physical demands of occupations were classified according to standard occupational codes into 3 categories: manual, mixed, and nonmanual.12
To predict claim rates, we performed a linear regression of the form
(1)
where xkit values are the set of focal predictors (e.g., percentage in manual occupation) for industry i and year t, μi represents unobserved influences related to each industry (i.e., dummy-coding industries), λt represents unobserved influences that vary across time (i.e., dummy-coding year), and eit is an error term. Including the industry dummy codes allowed us to examine within-industry fluctuations without between-industry variability confounding the other parameter estimates (e.g., 1 industry, in a given year, might have a higher percentage of male workers than do the other industries).7
We used SAS version 9.1 (SAS Institute Inc, Cary, NC) to perform an ordinary least squares regression analysis. We used an ordinary least squares model rather than a Poisson model because the former was less prone to misspecification bias of the parameters13 than was the latter and it met the ordinary least squares assumptions regarding normal distribution of the residuals and constant variance.
Workers’ compensation claim rates generally declined from 1990 to 2003; the exact percentage of decline varied by industry (Table 1). Table 2 shows the overall means and standard deviations of the predictors in the regression model.
The percentage of manual jobs in each industry was significantly associated with claim rates over time (Table 3). The coefficient indicates that a decrease of 1% in the proportion of manual jobs led to a decrease of almost 0.03 in the overall claim rate. Percentages of workers in the age, gender, and job tenure groups were not significantly related to temporal trends in industry claim rates.
Consistent with trends in North America and Europe,3 claim rates in Ontario declined 3.9% per year from 1990 to 2003. All industries showed substantial declines in claim rates, supporting the idea that overall declines are the result of more than just the shifting of the workforce into low-hazard service sector jobs.
Across 12 industrial sectors, the reduction in work-related compensation claims in Ontario during 1990 to 2003 more strongly correlated with the contemporary decrease in the proportion of workers in occupations with high physical demands than with changes in the demographic composition of the work-force. This pattern is consistent with the notion that macrolevel economic changes such as the export of hazardous jobs overseas or workplace technology investments (e.g., automation) influence claim rate trends in a downward direction. Our findings reinforce the importance of giving attention to economic incentives and regulatory policies that encourage investment in safer production processes.
Note. 2000 hours = 1 full-time equivalent. aFor all years and industry sectors. aSignificant focal predictor. Claim Rate, 1990 Claim Rate, 2003 Rate Change, % All sectors 3.96 1.96 −50.63 Industrial sector Agriculture 2.83 2.44 −13.82 Forestry, fishing, mining 3.29 1.58 −51.99 Manufacturing: nondurables 3.67 1.80 −51.14 Manufacturing: durables 6.22 2.51 −59.63 Construction 4.23 1.99 −52.89 Transport/communication 3.84 2.59 −32.49 Wholesale trade 3.81 1.60 −58.15 Retail trade 3.23 2.22 −31.12 Community services 3.33 2.10 −37.11 Personal services 2.88 1.50 −47.80 Business/miscellaneous services 0.79 0.29 −63.52 Public administration 3.32 1.57 −52.76 Mean % of Workforcea (SD) Tenure group < 12 mo 22.37 (8.75) 12–47 mo 26.74 (6.44) ≥ 48 mo 50.89 (13.70) Gender Women 39.34 (19.10) Men 60.66 (19.10) Occupation Manual 43.59 (23.82) Mixed 22.50 (12.25) Nonmanual 33.91 (14.61) Age group, y 15–24 18.63 (12.92) 25–49 64.63 (9.90) ≥ 50 16.74 (4.01) b P Intercept 2.111 .374 Year vs 1990 1991 0.117 .426 1992 −0.044 .773 1993 −0.366 .027 1994 −0.455 .011 1995 −0.684 .001 1996 −1.068 < .001 1997 −1.169 < .001 1998 −1.408 < .001 1999 −1.475 < .001 2000 −1.413 < .001 2001 −1.553 < .001 2002 −1.618 < .001 2003 −1.814 < .001 Industrial sector vs public administration Agriculture 1.119 .149 Forestry, fishing, mining 0.226 .764 Manufacturing: nondurables 0.773 .233 Manufacturing: durables 2.175 .003 Construction 0.974 .170 Transport/communication 1.972 .002 Wholesale trade 1.461 .002 Retail trade 2.123 < .001 Community services 1.669 .021 Personal services 1.715 .008 Business/miscellaneous services 1.755 < .001 Tenure, % < 12 mo 0.003 .807 12–47 mo −0.006 .656 ≥ 48 mo Referent Referent Gender, % Men 0.003 .817 Women Referent Referent Occupation, % In manual jobs 0.029 .035a In mixed jobs 0.017 .429 In nonmanual jobs Referent Referent Age, y, % 15–24 −0.044 .189 25–49 −0.014 .591 ≥ 50 Referent Referent
This study was supported in part by the Ontario Workplace Safety and Insurance Board (grant 02–007). P. Smith was supported by a strategic training research fellowship from the Canadian Institutes for Health Research Strategic Training Program in the Transdisciplinary Approach to the Health of Marginalized Populations.
We would also like to acknowledge Marjan Vidmar’s assistance with data retrieval.
Human Participant Protection The study protocol was approved by the University of Toronto Health Sciences research ethics committee.