There is a large body of workplace literature across disciplines that indicate a variety of job quality indicators and workplace characteristics (i.e. job security, pay, worker autonomy, interesting work, etc.) that may have an impact health. The foundation for most research in this field is Karasek’s (1979) job strain model. In this model, two key work characteristics—(1) one’s job demands and (2) the level of one’s decision latitude—combine to determine the risk of psychological strain and physical illness. If one has high job demands and low decision latitude, the likelihood of illness is high. On the other hand, if either job demands are low or decision latitude is high, there would be less risk for psychological strain and physical illness. However, despite the wide use of Karasek’s model in the literature (as well as variations such as the demand-control-support model), empirical evidence for this model is mixed (Totterdell, Wood, and Wall, 2006). Several studies explore worker health difference by gender and occupational type (see Bekker, Croon, and Bressers, 2005; Jamison, Wallace, and Jamison, 2004; Bogart, Vranceanu, and Walt, 2004; Morrison and Payne, 2003; Tsutsumi et al., 2003), while examining job characteristics as mediators of the SES-health relationship (Warren, et al., 2004).

Jamison, Wallace, and Jamison (2004) investigated the impact of contemporary work characteristics on worker stress and overall health and found that occupational status is negatively associated with ill health. Additionally, Bogart, Vranceanu, and Walt (2004) examined the independent and joint effects of occupational status and perceived demands, control, and social support at work on ambulatory systolic blood pressure and heart rate in women and found that occupational status and job characteristics accounted for 18% and 22% of the inter-individual variability in ambulatory systolic blood pressure. Finally, they found that women with both low status occupations and stressful job circumstances are at a disproportionately high risk of cardiovascular illness.

In addition to the research of work conditions and psychological and physical health, another important area of this literature addresses repetitive strain injuries in relation to worker physical health. Cole, Ibrahim, and Shannon (2005) examined the predictors of work-related repetitive strain injuries using data from four waves of the Canadian National Population Survey (CNPS) and found that female gender, some college or university education, job insecurity, high physical exertion levels, and high levels of psychological demands were all positively associated with work-related repetitive strain injuries.

In addition to the direct impacts of work characteristics on worker health, other studies look at the impact work characteristics have on worker health behaviors, which then in turn impact worker health. Tsutsumi et al. (2003) examined the association between job characteristics and health behaviors in Japanese rural workers and found that high psychological demands were associated with heavy smoking and alcohol consumption, as well as high work-related physical activity. Furthermore, low job control was associated with lower consumption of vegetables, less smoking, and a lower level of work-related physical activity. Additionally, job strain was associated with lower vegetable consumption, less smoking, and greater alcohol consumption.

Another important aspect of the workplace and health research is a look at potential gender differences in the impacts of particular work characteristics on health. The research on gender differences has produce inconsistent and sometimes contradicting results. Some researchers have found key gender differences in the impact of key variables on ill health (Jamison, Wallace, and Jamison, 2004; Tsutsumi, et al., 2003). For example, Bogart, Vranceanu, and Walt (2004) found that women with both low status occupations and stressful job circumstances are at a disproportionately high risk of cardiovascular illness. Additionally, Cole, Ibrahim, and Shannon (2005) found that female gender was positively associated with work-related repetitive strain injuries.

However, other researchers have had less clear results on the differences between health outcomes by gender. For example, Morrison and Payne (2003) examine the effect of gender on the relationship between work characteristics, job satisfaction, and psychological distress and found that job characteristics have predictable effects on men’s health, but not on women’s health. Furthermore, Bekker, Croon, and Bressers (2005) investigated the role of childcare involvement, job characteristics, gender, and worker attitudes on emotional exhaustion and sickness absence from work and found that women did not have higher levels of absenteeism and that men had higher levels of emotional exhaustion, linked to having significantly more working hours. Finally, the researchers found that a combination of work load and care load predicted emotional exhaustion and sickness absence from work.

Additionally, much research has found a persistent link between socioeconomic status and individual health (see Feinstein, 1993; Williams, 1990; Palmer, 1989; Feldman et al, 1989; Kitagawa and Hauser, 1973). Williams (1990) argues that one’s socioeconomic interacts with other factors, such as demographic factors, biomedical factors, and psychosocial factors, which in turn impact health outcomes. Furthermore, in a review of key literature, Feinstein (1993) found that psychosocial factors such as one’s educational attainment, access to health care, health practices, social ties, and stress all play an important role in this SES-health link. Additionally, Warren, et al. (2004) examined the mechanisms that link SES and health, focusing in on job characteristics as mediators for this relationship and found that job characteristics explain between 27-44 percent of the association between education and health.

Kitagawa and Hauser (1973) find that socioeconomic status had been in the past and continues to be a major determinant of mortality in the U.S. Furthermore, it is consistently acknowledged that females live longer than males (see Erhardt and Berlin, 1974). Additionally, Link and Phelan (1995) explore the social fundamental causes of disease, acknowledging the role that stigmatization, social support, gender, race, and SES all play. They find that SES is important as it grants access to resources, including money, knowledge, power, prestige, and social support.

Once more, Antonovsky (1967), in a review of several early historical epidemiological studies, including an analysis of historical pre-World War II data, found that there were class and occupational differences in both life expectancy rates and also causes of mortality, with a relatively small gap between the high and middle classes as compared to the much larger gap between them and the lowest class. Additionally, he found that individuals in higher-status occupations experienced lower death rates.

Furthermore, similar to the SES-health link explored above, despite much evidence suggesting a similar persistent relationship between SES and mortality, some recent researchers have found that the link is much weaker among women (Mcdonough et al., 1999). However, Mcdonough et al. (1999) find that, in direct contrast to men, increasing spousal income decreases women’s odds of dying, but that there are little-to-no gender differences in the effect of an individual’s own socioeconomic status and their own mortality risk.

The total number of cases included in this analysis is 36,272 (including individual females born prior to 1900’s with a work classification, or a housewife, cause of death, and socioeconomic status link to their death certificate.

To test the three hypotheses stated in the previous section, I use STATA 9 for all analysis, including descriptive statistics, cross-tabulation, survival curves, and a Cox proportional hazards model.

As can be seen in Table 1 below, I used 10 variables in the primary analysis.

Figure 1

Table 1: Descriptive Statistics

Death Year: This is the year of the individual’s death.

Years Lived Past 50: This is an integer variable and another measure of longevity, which looks at the number of years each individual lived past the age of 50.

Man/Admin/Serv: This is a variable derived from the 1980 U.S. Census Occupational classifications that were applied to occupational descriptions of the respondents in the UPDB. This designation includes all management, administrative, and service industry occupations (occupation codes 0-469).

Prod/Labor: This is another variable derived from the 1980 U.S. Census Occupational classifications that were applied to occupational descriptions of the respondents in the UPDB. This designation includes all production and labor occupations (occupation codes 473-889).

# Children: This is an integer variable of the total number of children the individual had.

LDS: This is a dummy variable for whether or not the individual was a member of the LDS faith.

Mom Death Age: This variable looks at the death age of the individual’s mother.

Dad Death Age: This variable looks at the death age of the individual’s father.

Birth Year: This variable simply indicates the individual’s year of birth.

Husband SES: This variable (NanPowers Score) is a measure of occupational prestige and socio-economic status.

Additional Variables in Specific Analysis not Included in Descriptive Statistics

Cause of Death Dummy: This is a measure of natural versus external causes of death for each individual—”external causes” and other labor-related ailments (cause of death codes 1, 4, 8, 13, 17, 0), “natural causes” (cause of death codes 2, 3 5, 6, 7, 9, 10, 11, 12, 14, 15).

Homemaker: Homemaker (occupational code 914) was excluded in the Cox Proportional Hazards Model as the comparison group for the other management and labor occupation variables.

First, I used simple cross-tabulation to look at the first hypothesis regarding occupational type and cause of death. To restate, the first hypothesis specifically states:

Hypothesis 1: Individuals with more production/labor intensive jobs will have causes of death more closely associated with “external causes” and other labor-related ailments, while individuals with managerial, administrative, and service occupations will have causes of death more closely associated with more natural causes, such as cancer, diabetes, and heart disease.

Table 2 below shows the results of the cross-tabulation of an individual’s cause of death by their occupational type (with occupations grouped into one of three categories—(1) management/administrative/service industry occupations, (2) production/labor occupations, and (3) homemaker).

Figure 2

Table 2: Cross-tabulation of Cause of Death Dummy* by Occupational Type**

As can be seen, the significant Pearson chi-squared and likelihood ratio tests (p<.05) indicates that there is a significant difference between the different occupations and the associated causes of death. While most cases were homemakers (who primarily died of natural causes—only 12.7% due to external causes), there was a slight increase in the prevalence of “external causes” of death when the individual leaves the home to work in a management/administrative/service occupation (14.7% of deaths due to external causes), and another small jump when the individual engages in a production/labor occupation (15.3% of deaths due to external causes). This tentatively supports hypothesis 1.

Next, I generated a survival curve to look at the first second hypothesis regarding occupational type and life expectancy. To restate, the second hypothesis specifically states:

Hypothesis 2: Individuals with more production/labor intensive jobs will experience shorter life spans, while individuals with managerial, administrative, and service occupations will experience longer life spans. Additionally, housewives will experience longer life spans than women who work outside the home.

Figure 1 below shows 3 survival curves for each of the three occupational categories outlined previously. One can see that individuals with more production/labor intensive jobs actually experience longer life spans than women in management, administration, or service occupations, while homemakers experience the shortest lifespan (Log-Rank Test chi2=474.09, P<.001). Relate results in the Cox Proportional Hazards model will be discussed further in that section. This finding is at odds with the original hypothesis that those in production/labor occupations would experience a greater risk of mortality than either management/administrative/service jobs or homemakers.

Figure 3

Figure 1: Survival Curves, by Occupational Category*

Next, I generated a survival curve to look at the first second hypothesis regarding occupational and socioeconomic status and life expectancy. To restate, the third hypothesis specifically states:

Hypothesis 3: Individuals in occupations with higher Nan-Powers score (index of occupation and socioeconomic status) and occupational prestige will experience longer life spans.

Figures 2 and 3 below show 4 survival curves for four categories of occupational prestige and socio-economic status (low, middle-low, middle-high, and high), for both the women and her husband. Whether looking at a woman’s socioeconomic status or that of her husband, both Figure 2 and Figure 3 show that individuals in occupations with higher Nan-Powers scores are more likely to reach the late 80’s and early 90’s than others (Log-Rank Test chi2=209.24, P<.001). Additionally, there is a clear divide between those with the lowest NanPowers scores and the top three categories, with survival rates of those in the lowest groups plummeting much more quickly after the age of 50. Relate results in the Cox Proportional Hazards model will be discussed further in that section. These results support the initial hypothesis that the greater one’s occupational and socio-economic status, the greater one’s life expectancy.

Figure 4

Figure 2: Survival Curves, by Individual’s SES Score Categories*

Figure 5

Figure 3: Survival Curves, by Husband SES Score Categories

Finally, I generated a Cox Proportional Hazards Model to further look at the second and third hypotheses, controlling for other factors. Table 3 shows the model specifications and results.

Figure 6

Table 3: Cox Proportional Hazards Model

In relation to the second hypothesis, the Cox Proportional Hazards Model in Table 3 (net all control variables) shows that, in relation to the omitted homemaker dummy, individuals in management, administrative, and service occupations have a 25.6% lesser risk of mortality than homemakers (P<.001), while individuals in production and labor occupations have a 42.6% lesser risk of mortality (P<.001). This finding is at odds with the original hypothesis that those in production/labor occupations would experience a greater risk of mortality than either management/administrative/service jobs or homemakers.

In relation to the third hypothesis, the Cox Proportional Hazards Model in Table 3 (net all control variables), shows that an individuals’ socio-economic status has a small (.23%, P<.001) protective effect on mortality. This supports the initial hypothesis.

Though not included in the analysis presented here due to space restrictions, there is a clear shift in the prevalence of homemaker status as one’s birth year approaches the 20^th century, with a much larger percentage of women finding occupations outside of the home (increasing from a mere 4.09% for those born prior to 1990, to 11.82% for those born after 1890). Additionally, the Cox Proportional Hazards Model in Table 3 shows that Birth year has a protective effect, with individuals born in later years living longer (1%, P<.001, while the number of each additional child increases the risk of mortality slightly (0.58%). LDS membership has the biggest protective effect on the risk of mortality, at 6.4%.

Future research should examine the surprising finding in relation to the second hypothesis. Additionally, future research can look at these same research questions for both men and women born in the 19^th and 20^th centuries to look at the influence of occupation and SES on mortality, as well as any possible historical shifts in these relationships.