To work or not to work?

Thesis title: An analysis of the relationship between the marital status of an employee and the period of time taken off from work due to injuries sustained at the workplace

Abstract
The primary aim of this paper is to analyze the statistical relationship between the marital status of an employee and the period of time said employee takes off from work due to injuries sustained at the workplace. The analysis will be conducted on the basis of data consisting of 7,150 observations and 13 variables. This paper will aim to observe as many factors which have bearing on the duration of benefits, as is reasonably possible, with a specific focus on the role of the marital status of an employee. Such an endeavor will necessitate the observation of a variety of aspects consisting of emotional, physical and sexual factors. The overarching aim of our analysis is to draw the attention of employers towards the different factors which impact durat (the time duration of the provision of benefits) and pique the interest of other researchers to conduct further studies on the issue we have raised in our current undertaking.

Introduction
The primary assumption of this paper shall be that married individuals have greater tendency to prolong the time they take off due to injuries, as compared to unmarried individuals, because their spouses are also members of the active workforce. Accordingly, married individuals can afford to take more time off because they are not the sole breadwinners of their household. Such an assumption is further reinforced by the notion that this data was gathered after the women’s movement which advocated financial independence for females and thus, encouraged a greater amount of them to seek financially active roles, coupled with their homemaking tasks.
At the same time, a diametrically opposed state of affairs may also hold true. Such a notion is rooted in the assumption that as married individuals typically earn more than unmarried individuals, taking leave off work may entail greater opportunity costs for the former and, thus, encourage them to return to work sooner than the latter.
One may also make a case for a middle path situated between the two aforementioned extremes which would entail arguments based on the hypothesis that marital status does not have any influence on durat and that, therefore, any possible reasons for an individual’s tendency to take more time off work is relegated to reasons arising out of factors other than marital status. Such a line of reasoning would typically hold when the statistical relation between durat and married is deemed insignificant.
Therefore, having traced the possible trajectories this paper could pursue, the ultimate aim of this paper will be to determine the effect of marital status on durat. Accordingly, the question we have posed is: “What effect does marital status of an employee have on the time duration taken off work due to injuries sustained at the work place?” For this purpose, we have a sufficient number of observations i.e. 7150, even though the variables accounted for are quantitatively limited. 7150 individuals were surveyed to provide a multi-dimensional dataset which includes variables such as the real Dollar value of benefits, total medical costs, type of injuries, wage and demographic factors such as age, gender and marital status etc.
The paper will, initially, present a literature review in order to ensure that our current undertaking is accurately contextualized within the existing body of literature. Accordingly, the studies cited and explored will present a comprehensive overview of existing scholarly articles that substantiate and reinforce the overarching claim of this paper. Having established the scholarly milieu within which it is meant to be situated, the paper will then not only elaborate upon the methodology used to govern its empirical analysis but will also indicate how it has been safeguarded from bias and inconsistency. Accordingly, it will take note of the limitations of our study and will expound upon the impact those limitations may have on our findings. The last section will then state the conclusions that may be inferred from our research and analysis.
Regarding the consequence and inherent value of our current undertaking, let it be noted that this paper will provide invaluable insight into the psyche underlying an employee’s demand to take time off work and may, therefore, function as an integral element of a company’s employment policies and practices.

Literature Review
Having reviewed a wide range of scholarly articles, we have identified a conspicuous gap in existing scholarship because there are, currently, no empirical works that attempt to answer the question we have raised. Let it be noted, however, that a variety of papers do allude to the question at hand and that we have incorporated those in our current undertaking.
The study entitled Factors Influencing the Duration of Work-Related Disability: A Population-Based Study of Washington State Workers’ Compensation by Allen Cheadle et al. discusses a variety of factors that affect the duration of work related disability. Although the study does not explicitly discuss the effect of an employee’s marital status on duration out of work due to injury, it does acknowledge (in its conclusion) that marital status does, indeed, have a significant effect on duration out of work due to injury which is typically characterized by an order of small magnitude.
The study entitled Predictive factors of work disability in rheumatoid arthritis: a systematic literature review by E M de Croon et al. accounts for predictive factors of work disability and asserts that there is strong evidence to indicate that there is no association between marital status and work disability, let alone between the former and durat.
The study entitled A Theory of Allocations of Time in Markets for Labour and Marriage by Grossbard-Shechtman investigates the effects of marital status on the allocation of time by couples between market and non-market hours. Although their conclusions do support our intuitive claim that married individuals take more time off due to work related injuries as compared to unmarried individuals, it must be noted that the findings of the aforementioned study are based entirely on women. One may, thereby, conclude that the study indicates that marriage decreases the market work of women and increases their non-market i.e. domestic work.
The study entitled The effects divorce risk on the labour supply of married couples by Kerry L. Papps asserts that if there is an increase in the risk of divorce then spouses, due to financial, and other, reasons are likely to allocate more time to work. One may, thus, conversely infer that healthy marriages are not conducive to hard work and that the findings inferred by the aforementioned paper are, indirectly, aligned with the claim of this current undertaking.

Data Description
The paper will fulfill the primary purpose of determining the effect of marital status on durat by running a basic regression of durat on married as is shown below:
The dataset entitled Benefits is essentially, cross-sectional in nature and includes information on 7150 employees who were injured during the course of their jobs. Moreover, the dataset includes demographic information such as an individuals age, gender and marital status. Other variables of interest included are total medical costs, benefits received, wages prior to injuries, duration of the receipt of benefits and the industry to which the individual belongs.
This data was collected by econometricians looking to establish a relationship between the value of benefits received by injured workers and the corresponding time taken off work. In turn, this paper uses the aforementioned dataset in order to verify whether or not marital status has bearing on the duration of work related disability and, thereby, the two variables of primary interest are durat (duration of benefits) and married (marital status).
Table 1 in the appendix provides the names and descriptions of all the relevant variables in our analysis.
The variable ‘afchnge’ is a dummy variable which equals to a value of 1 if one considers the post-policy change era, whereas it equals a value of 0 if one takes into account the time before the policy change. This variable has been included in order to neutralize the effects that time may have on our analysis. Moreover, benefit, totmed, male, marriedmale, hosp, age, highearn and prewage have all been included in this paper’s analysis because these variables have significant bearing on the time employees take off from their jobs due to of injuries.
At this point, let it be noted that the data used in our current undertaking has a number of limitations. Firstly, the dearth of variables in the dataset provided affects our analysis because it fails to acknowledge key variables such as the number of dependents, spouse's wage, type of job and unemployment rate etc. Secondly the variable benefit is stated in real terms whereas totmed and prewage are mentioned in nominal terms. Therefore, the implication is that these specific variables have not been adjusted for inflation, which posits a threat to our analysis because it indicates that the comparison of these terms is largely inconsistent. Thirdly, the variable hosp is a dummy variable which merely informs us about whether or not an individual was hospitalized due to an injury. It must be pointed out that this variable would, indeed, have been of greater use and applicability if it stated the number of days spent in hospital which would then provide one with a fair indication of the severity of the injury sustained.
Model Specifications
The dependent variable in our model is durat i.e. the duration of benefits provided to workers to compensate for injuries resulting at the workplace. In order to determine the relationship between the marital status of an employee and the duration of benefits provided to him, we carry out the simple regression: ldurat = α0 + α1married + v
It is quite apparent that the estimation of the model yields a highly significant positive estimate for the coefficient of married (3.24) since the p-value is 0.0000, which is less than the significance level of 5%. Furthermore, the R-squared value is very small and equals to 0.0360. Indeed, the coefficient α1 makes intuitive sense because, having stated that we are of the opinion that married individuals are expected to claim longer duration of benefits by staying out of work for a longer time period. In fact, if an employee is married, the time duration taken off work increases by 3.24 weeks. Since we have not taken into account many other factors that effect the duration of benefits due to injury, the R-squared value of the model is very small and the regression is bound to provide biased estimates due to the omitted variable bias. Our simple model, moreover, does not fulfill the Gauss Markov assumption of E (u|x) = 0 since there are bound to be factors in the error term which are correlated with marriage, as is shown in the table, and do, indeed, effect durat. Furthermore, we can see that the data on durat is highly skewed, as is shown in figure 1.

| durat | married | male | age | hosp | benefit | totmed | prewage | highearn | afchnge | married | 0.0586 | 1 | | | durat | 1 | | | | | | | | | | male | 0.0067 | 0.0819 | 1 | | | | | | | | age | 0.0767 | 0.2851 | -0.1150 | 1 | | | | | | | hosp | 0.2653 | 0.0437 | 0.0516 | 0.0312 | 1 | | | | | | benefit | 0.1395 | 0.2020 | 0.2792 | 0.1500 | 0.0878 | 1 | | | | | totmed | 0.1206 | 0.0111 | -0.0231 | 0.0065 | 0.0731 | 0.0162 | 1 | | | | prewage | 0.1237 | 0.2627 | 0.3250 | 0.1881 | 0.0851 | 0.6868 | 0.0226 | 1 | | | highearn | 0.0961 | 0.2851 | 0.3462 | 0.1579 | 0.0835 | 0.5680 | 0.0229 | 0.8544 | 1 | | afchnge | 0.0332 | 0.0226 | 0.0300 | 0.0179 | -0.0133 | 0.2928 | -0.0055 | 0.0133 | 0.0173 | 1 |

In order to improve our current model we must account for the factors mentioned in the table above. The following model will provide us with the ceteris paribus effect of married on duration. ldurat = β0 + β1married + β2marrmate + β3male + β4age + β5hosp + β6ltotmed + β7lbenefit + β8lprewage + β9highearn + β10afchange + u
In order to ensure a better fit of the distribution of durat compared to the standard normal distribution, we have taken the logarithm of durat and have done the same for totmed, prewage and benefits in order to ensure that we are able to gauge a more accurate interpretation of the estimates. It must be noted that although the estimation of this model yields significant estimates for male, age hosp, ltotmed, lbenefit it also generates insignificant estimates for married, marrmale, lprewage, hoghearn and afchnge.
The coefficient of our variable of interest, married, is 0.0402807 which is entirely in line with our intuitive claim that a married employee will claim higher duration benefits because he can afford to take more time off work because he can rely on his spouse's income to maintain the expenses of the household. However, this estimate is insignificant at the 5 % sig. level since our p-value is 0.470. We have included an interaction term, marrmale, which is the product of married and male in order to check for a gender bias in our findings. The variable afchnge has a negative coefficient which is suggestive of the fact that after the policy change the durat decreases. However, it is insignificant at the 5% sig level since p-value is 0.373. Moreover, age and hosp have highly significant positive estimates (p-values of 0.000) which intuitively make sense since older people will stay out of work longer, in order to recover from their injuries, and people who are hospitalized will also stay out of work longer due to the relatively greater severity of their injury. Let it be noted, however, that males are expected to return to work earlier since as is envinced by the negative coefficient. This value is significant at the 5% level with a p value of 0.027.
The coefficients of ltotmed and lbenefit have highly significant positive coefficients which makes logical sense as one may infer that the higher the benefit and totmed, the correspondingly longer duration the worker will stay out of his job. The coefficient of lprewage has a negative coefficient which is counter intuitive but because it is insignificant we ignore its affects.
Conclusion
Aim of the paper. This is what our findings tell us. These are ways in which the thing can be improved. Future researchers shoul do what?
The overarching aim of this current undertaking was to analyse the relationship between the marital status of an employee and the period of time said employee takes off from work due to injuries sustained at the workplace. Having conducted the empirical analysis detailed above, our findings indicate that one may, indeed, conclude that marital status does not appear to have a significant effect on the time duration an employee takes off work due to an injury resulting at the workplace. However, it must be noted that such an endeavour may be improved greatly if the number of variables included in the data set is increased in order to enable researchers to control for a greater number of factors. Moreover, certain variables in the dataset were not clearly defined due to which we were rendered unable to gauge accurately the estimation of results. Therefore a key aspect to focus on in order to improve the quality of our model and its findings would be the standardization of all the variables provided which would, thereby, enable one to extract more accurate estimators of the models’ slope coefficients. The purpose of our study is to draw employers attention towards the different factors that affect the time taken off from job due to injury and to pique the interest of other researchers to conduct further studies on the issue we have raised.

Appendix
Table 1: Variable Name | Description | Benefit | real dollar value of benefit | Totmed | total medical costs in 1982$ | Age | age at time of injury | Durat | duration of benefits | prewage | previous weekly wage in 1982$ | Afchnge | equal to 1 if afchnge in benefits | highearn | equal to 1 if highe earner | Male | equal to 1 if male | Married | equal to 1 if married | Hosp | equal to 1 if hospital stay required | Ldurat | log of duration | Ltotmed | log of totmed | lprewage | log of prewage | Lbenefit | log of benefit | marrmale | married*male |

Table 2: ldurat | Coefficient(Standard errors) | t-statistics | married | 0.0402807 (0.0577443) | 0.70 | marrmale | -0.0078046(0.0673174) | -0.12 | male | -0.1149515 (0.0544086) | -2.11 | age | 0.0077237 (0.001116) | 6.92 | hosp | 0.2404434 (0.0376739) | 6.38 | Ltotmed | 0.3454934 (0.0096446) | 35.82 | Lbenefit | 0.5240025 (0.0525828) | 9.97 | lprewage | -0.1206881 (0.0661538) | -1.82 | Afchnge | -0.0253679 (0.0277736) | -0.91 | highearn | 0.0098904 (0.0582156) | 0.17 |

Table 3: Ldurat | Coefficient(Standard errors) | t-statistics | | married | 0.0402807 (0.0557709) | 0.72 | | marrmale | -0.0078046(0.0654462) | -0.12 | | Male | -0.1149515 (0.0521114) | -2.21 | | Age | 0.0077237 (0.0011501) | 6.72 | | Hosp | 0.2404434 (.0426624) | 5.64 | | Ltotmed | 0.3454934 (0.0149345) | 23.13 | | Lbenefit | 0.5240025 (0.051194) | 10.24 | | lprewage | -0.1206881 (0.064801) | -1.86 | | Afchnge | -0.0253679 (0.0284925) | -0.89 | | highearn | 0.0098904 (0.0578222) | 0.17 | |

Table 4: Variable | Obs | Mean | Std. Dev. | Min | Max | durat | 7150 | 9.922203 | 24.49754 | 0.25 | 182 |

Figure 1: