Skip to main content
NCBI home page
Canadian Journal of Public Health = Revue Canadienne de Santé Publique logoLink to Canadian Journal of Public Health = Revue Canadienne de Santé Publique
. 2024 Oct 15;115(6):892–902. doi: 10.17269/s41997-024-00952-z

General unemployment and serious workplace injury rates: Workers compensation claims analysis from the Canadian province of Saskatchewan, 2007–2018

Samuel Kwaku Essien 1,2,3,✉, Cindy Feng 4,5, Catherine Trask 6,7
PMCID: PMC11644137  PMID: 39407070

Abstract

Objectives

There is conflicting published evidence that unemployment impacts workplace safety. Some studies suggest that the workplace injury rate decreases during economic contractions, while others propose an increased rate of injuries during periods of economic contractions. This study investigated the association between unemployment rates and traumatic work-related non-fatal injury (WRNFI) in Saskatchewan, 2007–2018, in order to provide new insight into injury prevention.

Methods

Saskatchewan’s retrospective linked workplace claims data from 2007 to 2018 were grouped by year, season, and worker characteristics (e.g., age and sex). Total employment, total labour force, and the number of unemployed workers from the Statistics Canada Labour Force Survey were grouped by year, season, sex, and age. These data were linked to the worker’s compensation board injury claim data to determine the number of people at risk, serving as the denominator (offset term) for WRNFI rates, calculated as WRNFI cases per total employed workers. A negative binomial generalized additive model was used to examine the association between unemployment rates and WRNFI, adjusted for age, sex, industry types, and seasons.

Results

The WRNFI rate has declined since 2007. On average, workers aged 20–29 years had the highest WRNFI rate (541.6 ± 84.8/100,000). Men had 3.2 times higher WRNFI risk than women (RR = 3.2, 95% CI 3.12–3.22), with the highest WRNFI risk observed in the manufacturing (RR = 1.68, 95% CI 1.63–1.73) and construction (RR = 1.67, 95% CI 1.63–1.72) industries. WRNFI risk decreased non-linearly with an increasing unemployment rate, indicating a pro-cyclic pattern.

Conclusion

This analysis showed that WRNFI rates tracked unemployment rates. This suggests a need to increase prevention strategies and reduce disincentives for under-reporting during an economic downturn.

Supplementary Information

The online version contains supplementary material available at 10.17269/s41997-024-00952-z.

Keywords: Compensation, Injury prevention, Non-fatal, Unemployment, Occupational health

Introduction

Workplace injuries and illnesses are a common and costly burden to society. In Canada in the year 2020, there were 253,397 lost-time worker compensation claims, and these comprise only a portion of all workplace injury (Tucker & Keefe, 2021). Traumatic workplace injuries account for a large portion of all non-fatal claims; they are distinct from workplace illnesses (i.e., occupational disease), which tend to manifest long after the workplace exposures that contributed to them. Although both types of claims have substantial personal and societal impacts, traumatic injuries may show clearer links to immediate contributing factors within the work environment (Tedone, 2017) and, therefore, provide greater opportunity for prevention. This provides a strong motivation for investigating potential leading indicators of traumatic workplace injuries at the societal, organizational, or individual level.

Societal-level economic indicators have been suggested as one potential predictor of workplace injury. For example, unemployment rates have been found to be associated with workers’ compensation claims in two seemingly contradictory ways. Boone and van Ours describe a distinction between a ‘pro-cyclic’ versus ‘counter-cyclic’ relationships between general unemployment rates and occupational health outcomes (Boone & van Ours, 2006): a ‘pro-cyclic’ pattern shows economic growth coinciding with higher rates of workplace injury, while a ‘counter-cyclic’ pattern occurs when a recession and concomitant higher unemployment coincides with increased workplace risk. Pro-cyclic patterns seem to be the most commonly reported, with recessions and higher unemployment decreasing the rate of overall work-related injuries (IWH, 2009). This ‘boom time’ pattern is hypothesized to arise from increased exposure to hazards. It is thought that greater production rates incentivize more overtime work and fatigue, as well as older, less safe equipment being used to increase production at both established enterprises and smaller new firms (IWH, 2009). Increased demand for labour may incentivize hiring more young or inexperienced workers who could lack the training, experience, and self-efficacy to recognize, report, and mitigate hazards (IWH, 2009). Further, the higher unemployment rates that accompany an economic recession are usually more pronounced among the more hazardous industries like construction, which could reduce population-level work-related injury rates (IWH, 2009).

Empirically, evidence from Asfaw et al.’s and Jenke et al.’s published studies among workers in the United States and Western Australia, respectively, shows that work-related injury rates decrease with an increase in unemployment rates (Asfaw et al., 2011; Jenke et al., 2021). In both studies, a 0.2% decrease in the work-related injury rate was associated with a unit increase in the unemployment rate (Asfaw et al., 2011; Jenke et al., 2021). Asfaw et al. further argued that the decrease in risk of work-related injuries that are observed during economic downturns (as indicated by higher unemployment rates) could be attributable to slower work pace and less work (Asfaw et al., 2011), and in economic upturns, the accelerated work pace could lead to organizations’ safety culture being compromised, including limited safety training offered to workers, especially new recruits, as well as failure to follow certain safety procedures or rules (Asfaw et al., 2011). Moreover, de la Fuente et al. attributed the decrease in occupational injuries during an economic downturn to a ‘natural selection’ phenomenon in the labour market (de la Fuente et al., 2014). For example, the authors argue that seniority, experience, and type of contracts (temporary vs. unionized permanent full-time) are major determinants for being laid off during economic downturns; hence, unskilled and temporary contract workers who are at greater risk of occupational injuries are laid off first, leaving seasoned and permanent workers who have considerably fewer injuries (de la Fuente et al., 2014). Among Italian manufacturing workers, the pro-cyclic relationship between economic cycles, including the unemployment rate and occupational injuries, was attributed in part to changes in the workforce composition over the years (Farina et al., 2018). These changes were a shift in the age of the workforce to older workers, increased length of services and more foreign workers than native Italians (Farina et al., 2018).

However, a more recent study of traumatic workplace fatalities reported a ‘counter-cyclic’ pattern, where mortality risk significantly increased as the unemployment rate increased (Essien et al., 2022). Although this relationship was non-linear and showed a drop at the highest observed unemployment rate of 6.5% (Essien et al., 2022), it was an unusual finding in the context of existing injury research. It has been previously suggested that under-reporting plays a part in lower reported claims rates when unemployment is high (Boone & van Ours, 2006; IWH, 2009), and under-reporting could help account for previous findings predominately on non-fatal injury. The persistent effects of unfavourable economic conditions on under-reporting non-severe injuries in the work environment were further echoed by Leombruni et al. and Askenazy (Askenazy, 2006; Leombruni et al., 2019). Leombruni et al. argue that the weak bargaining position of low- and medium-skilled young new workers during economic downturns may contribute to their under-reporting of injuries (Leombruni et al., 2019). In addition, Askenazy asserts that workers in precarious positions due to unfavourable labour market conditions are reluctant to report accidents/injury claims (Askenazy, 2006). However, under-reporting phenomena would be expected to be much reduced for fatalities; traumatic workplace fatalities, in particular, are difficult to hide and, by their very nature, remove any incentive to ‘keep quiet’ to keep the job. Unfortunately, it is not yet known whether fatal and non-fatal injuries show the same pattern under the same fluctuating economic conditions. A better understanding of how patterns compare between fatal and non-fatal workplace injuries could clarify the degree to which under-reporting drives the relationship with economic growth. This, in turn, could give insight into how societal-level economic factors might be used to anticipate and prevent negative workplace health outcomes.

The paper investigates traumatic work-related non-fatal injury (WRNFI) claims in the Canadian province of Saskatchewan over the period 2007 to 2018. Conducting this study in the same jurisdiction and same time frame as the previous study of fatalities (Essien et al., 2022), which identified a ‘counter-cyclic’ pattern, allows for a more robust comparison and, thus, a better understanding of the mechanisms driving associations between the economy and workplace health outcomes. The specific aims of this paper are (1) to describe temporal trends in WRNFI rate overall and by age, sex, and industrial sectors; and (2) to investigate the association between economic indicators, i.e., unemployment rate, while adjusting for other factors (age, sex, season, and industrial sectors).

Methods

Study population and data sources

In Saskatchewan, about 80% of all workers are covered by the Saskatchewan Workers Compensation Board (SK-WCB) (IAIABC, 2022). This study used claims records of the Saskatchewan working population with a traumatic work-related non-fatal injury between 2007 and 2018. For comparability and consistency with previously published studies (Essien et al., 2022), the present study uses data covering the same time frame. The non-fatal claims records of workers aged 15 years and over were extracted from the Saskatchewan Workers Compensation Board (SK-WCB) data using the diagnosis summary description (Essien et al., 2022). The retrieved non-fatal claims cases were further grouped by year, age group, sex (male/female), seasons/quarters, and industry sector. The denominator of the WRNFI rate, i.e., the number of employed workers in the same categories of the numerator, is retrieved based on the Statistics Canada Labour Force Survey (LFS) (2007–2018) (Essien et al., 2022).

Primary exposure variable of interest

The unemployment rate was calculated by dividing the total quarterly unemployed workers by the total quarterly labour force in Saskatchewan in the stipulated study period, all multiplied by 100.

Other control variables

Workers’ age in years was categorized into six groups (15–19, 20–29, 30–39, 40–49, 50–59, and ≥ 60 years). As a follow-up to an earlier published study (Essien et al., 2022), our rationale is that using the same categorizations as the last paper allows for better comparability to test our central hypothesis about whether fatal and non-fatal claims have a different relationship with unemployment. Industries were classified into the following groups: business, construction, manufacturing, mining, professional (e.g., telecommunications, offices, and white-collar professionals), transportation and warehousing, and ‘other’ industries. The ‘other’ industries comprised public administration, accommodation, agriculture, education, finance, insurance, the healthcare sector, wholesale and chain stores, and all reported claims with unknown industry classification codes. To ensure comparability, the ‘other’ industries category is the same as in our previously published study on fatal claims. Four seasons/quarters of injury were included: quarter 1, January–March; quarter 2, April–June; quarter 3, July–September; and quarter 4, October–December. The data are aggregated by years, seasons, age groups, sex, and industrial sectors, which resulted in a sample size of 7392.

Data analysis

We calculated the traumatic non-fatal injury rates in a given worker group (age group, season, sex, and industrial sector) by dividing the number of non-fatal injury cases in each group by the total number of employed workers in that group and multiplying by 100,000. To model the potential non-linear effects of age and unemployment rate on the risk of non-fatal workplace injuries, we used a negative binomial generalized additive model (GAM) based on the mgcv package in R (Wood, 2017). GAMs are advantageous due to their flexibility in analyzing complex non-linear effects of continuous covariates on the response through a smoothing function (Wood, 2017).

The non-linearity effects of age and unemployment rates were determined through the estimated effective degrees of freedom (EDF), where EDF > 1 indicates a non-linear relationship and EDF > 2 represents a stronger non-linear relationship, while EDF = 1 denotes a linear relationship (Hunsicker et al., 2016; Zuur et al., 2009). The EDF basically measures the ‘oscillation’ of the smoothed-fitted line (Baayen & Linke, 2020). Furthermore, Economou and Theodossiou’s report on the effect of macroeconomic conditions on occupation health (Economou & Theodossiou, 2011) revealed that the effect of the unemployment rate on occupation health may take a long time to manifest (Economou & Theodossiou, 2011). To account for the delayed effect of the unemployment rate, Economou and Theodossiou recommended lagging the unemployment rate in an arbitrary manner (Economou & Theodossiou, 2011). The present study evaluated the potential delayed effect of the unemployment rate through four lags (lags 0, 1, 2, and 3 seasons), and the lag that produced the lowest Akaike information criterion (AIC) was chosen.

A two-stage model-building approach was employed in this study. First, a univariate regression was fitted, and each variable of interest with a P-value < 0.25 qualified for inclusion in the multivariable model. Second, a multivariable model was fitted and based on the manual backward selection method, significant variables in the multivariable model with a P-value < 0.05 were retained in the final model (Hosmer et al., 2013). Furthermore, the interaction between variables was examined to determine whether there are potentially different time trends for different gender groups, industrial sectors, age groups, or age-by-gender groups over time. To assess over-dispersion in the outcome variable, we performed both Poisson’s and negative binomial regressions. The over-dispersion was assessed by applying a scale parameter estimated based on the deviance statistic divided by the residual degree of freedom (Hilbe, 2014; Payne et al., 2017), which showed a substantial over-dispersion in the Poisson model. Model selection was based on the lagged unemployment rate, which produced the lowest AIC. The negative binomial regression model provided the best fit for the data and produced the lowest AIC value (AIC = 241,483) compared to the Poisson model (AIC = 551,330) (see Supplementary Table 1).

Results

Descriptive statistics

Over the 12 years (2007–2018), there were 330,118 (WRNFI) cases in Saskatchewan, of which 219,349 (66.4%) were among males and 110,769 (33.6%) among females. The highest number of WRNFI was among workers aged 20–29 years (26.6%). Workers aged 30–39 years made up the second largest WRNFI group (21.7%), followed by workers aged 40–49 years (21.2%). Figure 1 shows that the age 20–29 group has a much higher injury rate than the age 60 + years. However, the proportion of WRNFI cases reported in workers aged ≥ 60 years (6.0%) was similar to those observed in workers aged 15–19 years (5.8%). Overall, single industries with the most common WRNFI claimants were the construction sector (14.6%), followed by the manufacturing sector (13.3%) and then transportation and warehousing (11.2%). The combined industries group ‘other industries’ accounted for 47.7% of WRNFI in the province, with the most common WRNFI claimants found in the healthcare sector (13.1%), followed by wholesale and chain stores (9.0%) and public administration (7.8%). The WRNFI cases observed in quarter 3 (July–September) were slightly higher (26.4%) than those observed in quarter 1 (January–March) (24.7%), quarter 2 (April–June) (24.5%), and quarter 4 (October–December) (24.4%) (Table 1).

Fig. 1.

Fig. 1

Open in a new tab

a The yearly non-fatal injury rate and the unemployment rate over the years. b, c, and d The yearly non-fatal injury rate against years stratified by sex, age groups, and industrial sectors, respectively. Note on the difference vertical scale for d: in order to maintain comparability with prior analyses of fatalities (Essien et al., 2022), the same denominator was used for industries and the same age categorization as the prior analyses, resulting in a reduced scale

Table 1.

Workers’ characteristics and compensation claims, 2007–2018

Study variable Number (%)
Total traumatic work-related non-fatal injury 330,118 (100)
Age (years)
  15–19 19,258 (5.8)
  20–29 87,760 (26.6)
  30–39 71,724 (21.7)
  40–49 70,023 (21.2)
  50–59 61,666 (18.7)
  60 +  19,687 (6.0)
Sex
  Female 110,769 (33.6)
  Male 219,349 (66.4)
Industry group
  Business 16,531 (5.0)
  Construction 48,042 (14.6)
  Manufacturing 44,014 (13.3)
  Mining 13,777 (4.2)
  Professional services 13,527 (4.0)
  Transportation and warehousing 36,870 (11.2)
  Other industries 157,357 (47.7)
Working population by quarters
  Quarter 1 (Jan–March) 81,376 (24.7)
  Quarter 2 (April–June) 81,028 (24.5)
  Quarter 3 (July–Sept) 87,103 (26.4)
  Quarter 4 (Oct–Dec) 80,611 (24.4)
Number of injuries by year
  2007 30,417 (9.2)
  2008 30,319 (9.2)
  2009 26,725 (8.1)
  2010 26,860 (8.1)
  2011 28,537 (8.7)
  2012 29,255 (8.9)
  2013 29,137 (8.8)
  2014 28,177 (8.5)
  2015 25,992 (7.9)
  2016 24,994 (7.6)
  2017 25,505 (7.7)
  2018 24,200 (7.3)
Open in a new tab

Trend analysis

The overall WRNFI rate trends over the study period (2007–2018) are presented in Fig. 1. The average WRNFI rate from 2007 to 2018 was 420.5 ± 46.7 per 100,000. The WRNFI rate has been on a downward trend since 2007, with a sharp decline observed from 2008 to 2009 and periods of stability between 2011–2012 and 2015–2017. The highest annual WRNFI rate was recorded in 2007 (502.0 per 100,000), while the lowest occurred in 2018 (353.8 per 100,000). Figure 1(a) also shows that the timing of the rise and fall of the WRNFI rate corresponded with the timing of the upward and downward trends in the unemployment rate. The WRNFI rate went up in 2007 and declined thereafter through to 2010, whereas the unemployment rate declined between 2007 and 2008 but then rose between 2008 and 2010. However, a reverse trend was observed between 2014 and 2018, where the unemployment rate spiked sharply between 2014 and 2016 and remained stable between 2016 and 2018. In contrast, the WRNFI rate consistently declined between 2014 and 2018. Figure 1(b) reveals higher WRNFI rates in men than in women for each year over the study period. The average men and women WRNFI rates from 2007 to 2018 were 517.0 ± 71.9 and 305.7 ± 19.1 per 100,000.

Figure 1(c) depicts WRNFI rates by age distribution from 2007 to 2018. The results show that WRNFI varies by age group. Compared across age groups, workers aged 20–29 years had the highest WRNFI rates for each study year (average rate: 541.6 ± 84.8 per 100,000 over 12 years), whereas older workers (≥ 60 years) had the lowest rates of WRNFI (average rate: 230.3 ± 11.5 per 100,000 over 12 years). As noted in Fig. 1(d), although the ‘other’ industry group had the highest WRNFI rate for each study year investigated, our rationale of ensuring better comparability to test the central hypothesis about whether the present study’s non-fatal claims and our previously published study on fatal claims have a different relationship with unemployment through the use of the same categorizations precluded further split of the ‘other’ industry category. A closer look at the industry-specific WRNFI rate indicates that, on average, the highest WRNFI rate was observed in workers in the construction industry (61.0 ± 9.7 per 100,000), followed by the manufacturing industry (56.2 ± 14.9 per 100,000) and transportation and warehousing industry (46.9 ± 5.7 per 100,000). The professional sector reported the lowest WRNFI rate (17.2 ± 3.1 per 100,000).

Generalized additive model

Supplementary Table 1 compares the AIC scores of negative binomial and Poisson’s multivariable regression analysis. The model selection based on the AIC score shows that the negative binomial generalized additive model using the unemployment rate lagged by one quarter was the one that performed best in terms of the fit to the study data, which produced the lowest AIC value. Hence, all the subsequent analyses of the present study were performed in accordance with the chosen model, the negative binomial generalized model, with the unemployment rate lagged by one quarter.

Table 2 shows that men were at a 3.2 times higher risk of WRNFI than women (RR = 3.17, 95% CI 3.12–3.22). Also, in comparison to quarter 4 (October–December or autumn), the RR of WRNFI is significantly higher in quarter 3 (July–September or summer) (RR = 1.06, 95% CI 1.04–1.09). Moreover, the risk of WRNFI did not differ significantly between quarter 1 (January–March) and quarter 4 (October–December or the autumn) (RR = 0.98, 95% CI 0.96–1.01), but the risk of WRNFI was significantly lower in quarter 2 (April–June or the spring) than in quarter 4 (RR = 0.95, 95% CI 0.92–0.97). The results also revealed that the risk of WRNFI is significantly higher in three industry groups (construction, manufacturing, and transportation and warehousing) and significantly lower in the mining (RR = 0.51, 95% CI 0.49–0.52) and professional services (RR = 0.63, 95% CI 0.61–0.65) industries than in the business industry. The WRNFI risk was approximately twofold greater in the manufacturing (RR = 1.68, 95% CI 1.63–1.73) and construction (RR = 1.67, 95% CI 1.63–1.72) industries than in the business industry. The WRNFI risk observed in the transportation and warehousing industry was 1.4 times higher than that for the business industry workers (RR = 1.44, 95% CI 1.39–1.48).

Table 2.

Estimated effects of covariates in the negative binomial regression model with a log link function. The table includes estimated coefficients (Est.), standard errors (SE), relative risks (RR, exponentiated covariate effect), and their corresponding 95% confidence intervals (CI) for each covariate. In the GAM, smooth terms were applied to model the non-linear effects of continuous covariates. Effective degrees of freedom (EDF) are reported as measures of the flexibility of each smooth term

Covariate Estimated covariate effects
Estimate SE RR 95% CI
Gender
  Female (ref)
  Male 1.155 0.008 3.17 3.12–3.22
Quarters
  Quarter 4 (Oct–Dec) (ref)
  Quarter 1 (Jan–March)  − 0.016 0.013 0.98 0.96–1.01
  Quarter 2 (April–June)  − 0.055 0.012 0.95 0.92–0.97
  Quarter 3 (July–Sept) 0.064 0.011 1.06 1.04–1.09
Industry group
  Business (ref)
  Construction 0.515 0.015 1.67 1.63–1.72
  Manufacturing 0.520 0.014 1.68 1.63–1.73
  Mining  − 0.679 0.016 0.51 0.49–0.52
  Professional services  − 0.466 0.015 0.63 0.61–0.65
  Transportation and warehousing 0.362 0.015 1.44 1.39–1.48
  Other industries 2.284 0.014 9.82 9.54–10.10
Smooth terms EDF
  Age 8.947
  Year 8.634
  Unemployment rate 7.604
Open in a new tab

Figure 2 and Table 2 summarize the non-linear effect of age, year of injury, and unemployment rate on the risk of WRNFI. The results shown in Fig. 2(a) indicate a non-linear effect of age on the risk of WRNFI (EDF = 8.95), with the risk of WRNFI reaching the highest for workers aged 20–29 years followed by a gradually declining trend as age increases until aged 50–59 years, then rapidly declining. The risk of WRNFI fluctuated during the study year (Fig. 2(b)). Most notably, the risk decreased from 2007 to 2009, followed by a slightly increasing trend up until 2012 and then declined until 2015, then sharply increased, peaking in 2017, followed by a declining trend. Figure 2(c) reveals that the risk of WRNFI decreased non-linearly (EDF = 7.60) as the unemployment rate increased. A more rapid decline in the risk of WRNFI was observed between higher unemployment values of 6.0–6.5%.

Fig. 2.

Fig. 2

Open in a new tab

Panels (a), (b) and (c) are the non-linear effects of age, year, and unemployment rate (lagged by one quarter) on the logged work-related non-fatal injury risk. The estimated non-linear effect is represented in solid lines, and 95% confidence intervals of the non-linear effects are represented in dashed lines. The partial effect represents the isolated effects of covariates in each of the three figures—note: The plots are on the log scale because negative binomial regression models with a log link function were employed

Discussion

We investigated 12 years of traumatic WRNFI in the Canadian province of Saskatchewan. We found that the traumatic WRNFI rate has declined since 2007, with periods of stability observed between 2011–2012 and 2015–2017. After controlling for the effect of other characteristics in the present study, the risk of WRNFI decreased non-linearly with an increasing unemployment rate: a more rapid decline in the risk of WRNFI was observed at higher unemployment rates.

The negative relationship observed between WRNFI risk and the unemployment rate in the present study is supported by earlier published studies (Boone & van Ours, 2006; Dong & Jestrab, 2022; Farina et al., 2018; Wunderly, 2021) and reflects a pro-cyclic pattern (Boone & van Ours, 2006; Wunderly, 2021). Dong and Jestrab’s study on workers’ compensation claims in California and Farina et al.’s study on Italian manufacturing workers both found a pro-cyclic relationship between unemployment and injuries at the workplace, with a percentage point increase in unemployment resulting in a 0.2% and 3% decrease, respectively, in work-related injuries (Dong & Jestrab, 2022; Farina et al., 2018). Wunderly similarly found that a 1% decrease in the unemployment rate elevated the risk of work-related non-fatal injuries by 17.8% (Wunderly, 2021). If economic upswings lead to an increased demand for workers, resulting in more entry-level workers or less trained workers being employed and becoming more susceptible to injuries at the workplace, as has been suggested (Boone & van Ours, 2006; Wunderly, 2021), then one would expect both fatal and non-fatal injuries to be affected in a similar way. However, the present study findings contradict the counter-cyclic relationship previously found between traumatic work-related fatalities (WRF) when analyzing the same dataset (Essien et al., 2022). In the previous study of this same dataset, higher unemployment led to an elevated risk of traumatic WRF (Essien et al., 2022). Pro-cyclical patterns have been suggested to arise from fewer work-related injuries getting reported during economic downswings due to workers’ fear of being fired (Boone & van Ours, 2006; Wunderly, 2021). This provides one possible explanation for the difference between fatal and non-fatal outcomes (WRNFI and WRF), in that they could be partly attributed to the potential under-reporting of non-severe, non-fatal injuries at the workplace (Kreshpaj et al., 2022). Under-reporting is an important and acknowledged phenomenon; Kreshpaj et al. found 22.6% and 15.0% under-reporting rates for non-fatal occupational injuries among precarious and non-precarious Swedish workers, respectively (Kreshpaj et al., 2022). Evidence also shows that about 40–50% of non-fatal occupational injuries were not reported in Canada (Thompson, 2007), and 60–67% were not reported in the American state of Michigan (Rosenman et al., 2006). Several factors have been proposed in the literature to be responsible for the under-reporting of injuries at the workplace (Pransky et al., 1999; Probst et al., 2013; Taylor Moore et al., 2013). This includes perceived job insecurity (Probst et al., 2013), perception of the injuries as small/less harmful and workers’ misconception that injuries were an unavoidable part of their job (Taylor Moore et al., 2013). The role of job insecurity seems particularly salient in times of higher unemployment.

Several authors have pointed out that WRNFI affects industry groups differently (Lucas et al., 2020; Win et al., 2021), with the construction industry disproportionately affected by WRNFI (50.4–56.4%) (Lucas et al., 2020; Win et al., 2021). Despite the same yearly and monthly employment estimates used in the present study as industry denominators due to suppressed employment numbers in the labour force survey (LFS) and comparability with prior fatalities (Essien et al., 2022) analyses, which generally resulted in reduced WRNFI rates in the industrial sectors, the observed higher WRNFI rates in workers in the construction industry (61.0 ± 9.7 per 100,000) was consistent with previously published studies (Lucas et al., 2020; Win et al., 2021). However, compared with the reference category ‘business industry’, the risk of WRNFI did not differ between construction (RR = 1.67, 95% CI 1.63–1.72) and manufacturing (RR = 1.68, 95% CI 1.63–1.73) industries. The present study’s higher WRNFI rate observed in the construction sector is identified in the literature (Brooks & Davis, 1996). Brooks and Davis found that the WRI (work-related injuries) rate was higher among teens who work in the construction industry (3.2/100 FTE, 95% CI 2.4–4.8), followed by manufacturing (3.0/100 FTE, 95% CI 2.3–4.5) (Brooks & Davis, 1996). Likewise, in the Canadian province of British Columbia, the WRI rate in the construction industry was higher than the average provincial WRI rate (3.3 vs. 2.1 per 100 workers) (WSBC, 2022). The high rate of WRNFI consistently observed in the construction industry is attributable to the inherently risky nature of jobs in the industry (AIHA, 2020; Love et al., 2010), including physically demanding work (AIHA, 2020), significant work stress (Love et al., 2010), and awkward postures (AIHA, 2020). The interaction between industrial sectors and the unemployment rate was tested in the present study, but no statistically significant interaction was found.

A notable difference in sex and age in the present study, with men’s 3.2 times higher risk of traumatic WRNFI than women’s and WRNFI mostly occurring among workers aged 20–29 years, followed by workers 30–39 years and declining as age increased, is identified in the literature (DeLeire & Levy, 2001; Lindqvist et al., 1999). Lindqvist et al. found that among Swedish workers, men suffered more occupational-related injuries than women (46 and 12 per 1000, respectively) (Lindqvist et al., 1999), which reflects the potential affinity for men toward riskier jobs as compared with women (DeLeire & Levy, 2001). Likewise, Viscusi et al. found that depending on the type of industry, the WRNFI risk rises among workers aged 20–24 years or 25–34 years and declines as age increases, with the lowest risk at 55–62 years (Viscusi & Aldy, 2007). The decrease in WRNFI risk for older workers is partly attributable to their experience in mitigating hazards, and they are more likely to be assigned safer job tasks in most industries (Viscusi & Aldy, 2007).

Traumatic work-related non-fatal injury is concomitant with seasonal variation in the present study. The risk of WRNFI was significantly higher in summer (July–September) than in autumn (October–December). Although the classification of weather seasons in Canada may differ from other jurisdictions, findings from these regions are congruent with the elevated risk of work-related injuries in summer compared to autumn (Liao, 2012; Pierce, 2013).

Strengths and limitations

A strength of the current study is that it provides interesting insight into the rarely studied phenomenon of economic conditions and allows for comparison to previous work in similar conditions. In addition, the 12 years of data provided the present study with the opportunity to assess the impact of periods of economic up- and downturns on WRNFI, as well as the lagging of unemployment rates adjusted for the putative delayed effect of the unemployment changes.

This paper has other methodological aspects to consider. The limitations typical of calculating claims rates using LFS data include sampling errors (Sharpe & Hardt, 2006). Thus, since only a sample but not the entire population was used, estimates produced from the sample may not be representative of the entire population (Sharpe & Hardt, 2006). Aggregated analysis can pose the risk of ecological bias, as individual risk perceptions of unemployment may vary across localities, sectors, employers, and job roles. Similarly, both risk management processes and union-mediated labour agreements may differ between the private and public sectors; this distinction is not captured in the current dataset. Further, there was no information collected on safety culture, prevention strategies, or risk management programs in place at the employer level. These factors could act as mediators linking economic conditions to injury risk but were not explored in this study, presenting an opportunity for future research, e.g., employing hybrid designs that combine group and individual-level data to mitigate the risk of ecological bias (Smoot & Haneuse, 2015).

Also, the present study did not stratify the traumatic work-related non-fatal injuries into severe and non-severe WRNFI. Non-severe injuries may be more prone to under-reporting, which could not be evaluated using the current dataset. Due to suppressed employment numbers in the LFS and to maintain comparability with prior fatalities (Essien et al., 2022) analyses, the same yearly and monthly employment estimates were used as industry denominators. The difference between fatal injuries in our previous study (Essien et al., 2022) and WRNFI in the present study can be best explained if data on production incentives or other issues in various industry sectors in Saskatchewan were available. However, the lack of data precluded the present study from postulating that under-reporting of WRNFI may be the sole cause. This could be an interesting avenue for future research.

Conclusion

After controlling for several variables, this study identified a non-linear pro-cyclic relationship between unemployment rates and WRNFI. That is, periods of economic growth and lower unemployment are related to higher rates of injury. This is consistent with most literature on the subject but contradicts a previous analysis of fatal injuries made using the same dataset. We propose that the apparent difference in the effect of unemployment rates on workplace safety is driven by under-reporting, which can be intuitively seen to impact reported non-fatal injuries far more than reported fatalities. The larger conclusion for prevention efforts is not that recessions and economic downturns make workplaces safer, but rather that non-fatal injuries are less likely to be reported under those conditions. This would seem to suggest a need for increasing prevention strategies that reduce disincentives for under-reporting and that reduce exposure to hazards during an economic downturn rather than during periods of growth. Still, there remain few studies investigating the potential impacts of economic conditions on workplace safety, and more research is needed to effectively target prevention efforts.

Contributions to knowledge

What does this study add to existing knowledge?

  • This study provides interesting insight into the rarely studied phenomenon of economic conditions and allows for comparison to previous work in similar conditions.

  • The study also used 12 years of workplace claims data, which is uncommon in Canadian occupational health literature, and this provided the opportunity to assess the impact of periods of economic up- and downturns on WRNFI, as well as the lagging of unemployment rates adjusted for the putative delayed effect of the unemployment changes.

What are the key implications for public health interventions, practice, or policy?

  • The difference between fatal and non-fatal outcomes could be partly attributed to the potential under-reporting of non-severe, non-fatal injuries at the workplace. This suggests a need to increase prevention strategies and reduce disincentives for under-reporting during an economic downturn.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 14 KB) (13.9KB, docx)

Acknowledgements

We are grateful for the supportive and insightful discussions on this topic with our colleague, Dr. Sean Tucker, and the data partnership of the Workers Compensation Board of Saskatchewan.

Author contributions

SKE, CT, and CF conducted the literature reviews. SKE, CT, and CF conceptualized and designed the study. CT and CF secured the study data, SKE performed the analyses, and SKE, CT, and CF interpreted the results and drafted the manuscript. All authors read and approved the final manuscript.

Funding

The authors gratefully acknowledge the financial support of the Workers Compensation Board of Saskatchewan. This research was supported in part by the Canada Research Chairs program (#228136).

Availability of data and material

The datasets used are available from the corresponding author upon reasonable request and with permission from the Workers Compensation Board of Saskatchewan.

Code availability

R codes for statistical analysis are available from the corresponding author upon request.

Declarations

Ethics approval

The study secured ethics approval from the University of Saskatchewan Biomedical Ethics Committee (U of S # 16–148).

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Conflict of interest

The authors declare no competing interests.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. American Industrial Hygiene Association (AIHA). (2020). Focus four for health–an initiative to address four major construction health hazards. https://aiha-assets.sfo2.digitaloceanspaces.com/AIHA/resources/FocusFour-forHealth_GuidanceDocument.pdf. Accessed 31 July 2023.
  2. Asfaw, A., Pana-Cryan, R., & Rosa, R. (2011). The business cycle and the incidence of workplace injuries: Evidence from the USA. Journal of Safety Research,42(1), 1–8. [DOI] [PubMed] [Google Scholar]
  3. Askenazy, P. (2006). Some determinants of reporting workplace accidents in France: The role of labour contract. Centre pour la recherche économique et ses applications (CEPREMAP), Paper, 606.
  4. Baayen, R. H., & Linke, M. (2020). An introduction to the generalized additive model (pp. 563–591). Springer. [Google Scholar]
  5. Boone, J., & van Ours, J. C. (2006). Are recessions good for workplace safety? Journal of Health Economics,25(6), 1069–1093. [DOI] [PubMed] [Google Scholar]
  6. Brooks, D. R., & Davis, L. K. (1996). Work-related injuries to Massachusetts teens, 1987–1990. American Journal of Industrial Medicine,29(2), 153–160. [DOI] [PubMed] [Google Scholar]
  7. de la Fuente, V. S., López, M. A. C., González, I. F., Alcántara, O. J. G., & Ritzel, D. O. (2014). The impact of the economic crisis on occupational injuries. Journal of Safety Research,48, 77–85. [DOI] [PubMed] [Google Scholar]
  8. DeLeire, T., & Levy, H. G. (2001). Gender, occupation choice and the risk of death at work. Working Paper No. 8574. National Bureau of Economic Research. https://www.nber.org/system/files/working_papers/w8574/w8574.pdf. Accessed 5 Aug 2023.
  9. Dong, X., & Jestrab, R. (2022). Unemployment and workplace injuries: Evidence from California. Applied Economics Letters,29(21), 2040–2043. [Google Scholar]
  10. Economou, A., & Theodossiou, I. (2011). The effect of macroeconomic conditions on occupational health and safety: The European experience. https://aura.abdn.ac.uk/bitstream/handle/2164/2180/Work_Injuries_UD.pdf?sequence=1&isAllowed=y. Accessed 1 July 2024.
  11. Essien, S. K., Trask, C., & Feng, C. (2022). Higher unemployment and higher work-related traumatic fatality: Trends and associations from the Canadian province of Saskatchewan, 2007–2018. Scandinavian Journal of Work, Environment & Health,48(4), 273. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Farina, E., Giraudo, M., Costa, G., & Bena, A. (2018). Injury rates and economic cycles in the Italian manufacturing sector. Occupational Medicine,68(7), 459–463. [DOI] [PubMed] [Google Scholar]
  13. Hilbe, J. M. (2014). Modeling count data. Cambridge University Press. [Google Scholar]
  14. Hosmer, D. W., Jr., Lemeshow, S., & Sturdivant, R. X. (2013). Applied logistic regression. John Wiley & Sons. [Google Scholar]
  15. Hunsicker, M. E., Kappel, C. V., Selkoe, K. A., Halpern, B. S., Scarborough, C., Mease, L., & Amrhein, A. (2016). Characterizing driver–response relationships in marine pelagic ecosystems for improved ocean management. Ecological Applications,26(3), 651–663. [DOI] [PubMed] [Google Scholar]
  16. Institute for Work & Health (IWH). (2009). Issue briefing. Workers’ compensation and the business cycle. Toronto. https://www.iwh.on.ca/sites/iwh/files/iwh/reports/iwh_issue_briefing_business_cycles_2009.pdf. Accessed 3 Aug 2023.
  17. International Association of Industrial Accident Boards and Commissions (IAIABC). (2022). Saskatchewan Workers’ Compensation Board. https://www.iaiabc.org/saskatchewan-workers-compensation-board. Accessed 13 June 2024.
  18. Jenke, T., Oosthuizen, J., & Cattani, M. (2021). An investigation of the influence of economic cycles on safety performance in Western Australia. Safety Science,138, 105230. [Google Scholar]
  19. Kreshpaj, B., Bodin, T., Wegman, D. H., Matilla-Santander, N., Burstrom, B., Kjellberg, K., & Orellana, C. (2022). Under-reporting of nonfatal occupational injuries among precarious and non-precarious workers in Sweden. Occupational and Environmental Medicine,79(1), 3–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Leombruni, R., Razzolini, T., & Serti, F. (2019). Macroeconomic conditions at entry and injury risk in the workplace. The Scandinavian Journal of Economics,121(2), 783–807. [Google Scholar]
  21. Liao, C. W. (2012). Pattern analysis of seasonal variation in occupational accidents in the construction industry. Procedia Engineering,29, 3240–3244. [Google Scholar]
  22. Lindqvist, K., Schelp, L., & Timpka, T. (1999). Gender aspects of work-related injuries in a Swedish municipality. Safety Science,31(3), 183–196. [Google Scholar]
  23. Love, P. E., Edwards, D. J., & Irani, Z. (2010). Work stress, support, and mental health in construction. Journal of Construction Engineering and Management,136(6), 650–658. [Google Scholar]
  24. Lucas, D. L., Lee, J. R., Moller, K. M., O’Connor, M. B., Syron, L. N., & Watson, J. R. (2020). Using workers’ compensation claims data to describe nonfatal injuries among workers in Alaska. Safety and Health at Work,11(2), 165–172. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Payne, E. H., Hardin, J. W., Egede, L. E., Ramakrishnan, V., Selassie, A., & Gebregziabher, M. (2017). Approaches for dealing with various sources of overdispersion in modeling count data: Scale adjustment versus modeling. Statistical Methods in Medical Research,26(4), 1802–1823. [DOI] [PubMed] [Google Scholar]
  26. Pierce, B. (2013). The seasonal timing of work-related injuries. U.S Bureau of Labor Statistics. https://www.bls.gov/osmr/research-papers/2013/pdf/st130230.pdf. Accessed 5 Aug 2023.
  27. Pransky, G., Snyder, T., Dembe, A., & Himmelstein, J. (1999). Under-reporting of work-related disorders in the workplace: A case study and review of the literature. Ergonomics,42(1), 171–182. [DOI] [PubMed] [Google Scholar]
  28. Probst, T. M., Barbaranelli, C., & Petitta, L. (2013). The relationship between job insecurity and accident under-reporting: A test in two countries. Work & Stress,27(4), 383–402. [Google Scholar]
  29. Rosenman, K. D., Kalush, A., Reilly, M. J., Gardiner, J. C., Reeves, M., & Luo, Z. (2006). How much work-related injury and illness is missed by the current national surveillance system? Journal of Occupational and Environmental Medicine, 48, 357-365. [DOI] [PubMed]
  30. Sharpe, A., & Hardt, J. (2006). Five deaths a day: Workplace fatalities in Canada. Centre for the Study of Living Standards. Ottawa. http://csls.ca/reports/csls2006-04.pdf. Accessed 7 Aug 2023.
  31. Smoot, E., & Haneuse, S. (2015). On the analysis of hybrid designs that combine group-and individual-level data. Biometrics,71(1), 227–236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Taylor Moore, J., Cigularov, K. P., Sampson, J. M., Rosecrance, J. C., & Chen, P. Y. (2013). Construction workers’ reasons for not reporting work-related injuries: An exploratory study. International Journal of Occupational Safety and Ergonomics,19(1), 97–105. [DOI] [PubMed] [Google Scholar]
  33. Tedone, T. S. (2017). Counting injuries and illnesses in the workplace: An international review. Monthly Labor Review, 1–24. https://www.bls.gov/opub/mlr/2017/article/pdf/counting-injuries-and-illnesses-in-the-workplace.pdf. Accessed 20 May 2023.
  34. Thompson, A. (2007). The consequences of underreporting workers’ compensation claims. CMAJ,176(3), 343–344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Tucker, S., & Keefe, A. (2021). 2021 Report on work fatality and injury rates in Canada. University of Regina. https://www.uregina.ca/business/faculty-staff/faculty/file_download/2022-Report-on-Workplace-Fatalities-and-Injuries-April-28-FINAL.pdf. Accessed 26 July 2023.
  36. Viscusi, W. K., & Aldy, J. E. (2007). Labor market estimates of the senior discount for the value of statistical life. Journal of Environmental Economics and Management,53(3), 377–392. [Google Scholar]
  37. Win, K. N., Trivedi, A., Lai, A., Hasylin, H., & Abdul-Mumin, K. (2021). Nonfatal occupational accidents in Brunei Darussalam. Industrial Health,59(3), 193–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Wood, S. N. (2017). Generalized additive models: An introduction with R. Chapman & Hall/CRC Press. [Google Scholar]
  39. Work Safe BC (WSBC). (2022). WorkSafeBC’s high-risk strategy for construction taking a risk-based approach to reduce injuries. https://www.worksafebc.com/en/about-us/news-events/news-releases/2022/April/worksafebc-high-risk-strategy-construction-taking-risk-based-approach-reduce-injuries. Accessed 30 July 2023.
  40. Wunderly S. (2021). The relationship between the unemployment rate and workplace injuries. https://samuelwunderly.netlify.app/files/Unemployment_Rate_and_Workplace_Injuries.pdf. Accessed 3 Aug 2023.
  41. Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R (Vol. 574, p. 574). Springer.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary file1 (DOCX 14 KB) (13.9KB, docx)

Data Availability Statement

The datasets used are available from the corresponding author upon reasonable request and with permission from the Workers Compensation Board of Saskatchewan.

R codes for statistical analysis are available from the corresponding author upon request.

Articles from Canadian Journal of Public Health = Revue Canadienne de Santé Publique are provided here courtesy of Springer

RESOURCES