The Effects of Job Quality on the Health of Wage Workers: Congruence between the Hard and Soft Job Quality

Background This study analyzes the linear and non-linear effects of the hard and soft dimensions of job quality on the overall health of wage workers. It also examines the congruence or fit between the hard and soft job quality on the overall health of wage workers. Methods This study measured thirty indicators that constitute job quality and reduced the indicators into twelve sub-dimensions of job quality using reflective factor analysis. In addition, this study derived two dimensions of job quality from the twelve subdimensions, namely the hard and soft job quality using formative factor analysis. This paper applied the response surface analysis to analyze the congruence effect between the two dimensions of job quality. Results A logarithmic relationship was found between the dimension of hard job quality and the worker’s overall health. This study also verified that the congruence effect between the two dimensions of job quality does not exist, and the combined effect of job quality is lower when the two dimensions of job quality are at the same level than the effect when either level of job quality is high or low. Conclusions Although hard and soft job quality has independent positive effects on the overall health of wage workers, the two dimensions of job quality are not congruent or not in harmony with each other. This incongruence between hard and soft job quality, together with a higher impact of hard job quality, suggests that the role of soft job quality on overall health is relatively limited.


Introduction
Job quality has been studied as an important factor for job stress, physical, and mental health, thereby increasing life satisfaction or quality of life [1e4]. Policymakers and international organizations have increased awareness of job quality over the past two decades, and research to define and measure job quality has resulted in the development of several international frameworks on the quality of jobs [5e7]. These frameworks have expanded their role as a basis for establishing better labor relations and policies, proposing a wide range of information for industrial and international comparisons. The European Foundation for Improving Living and Working Conditions (EWCS), e.g., developed a framework for measuring job quality in 33 European countries based on the results of the 5th EWCS. This framework consists of seven dimensions: physical environment, labor intensity, quality of working hours, social environment, skills and discretion or autonomy, job prospects, and income. The effect of certain attributes of job quality on the health of wage workers, such as the physical environment of jobs and working hours, has been studied considerably [8e10]. However, few studies have analyzed the impact of job quality on health with a comprehensive index to examine the role of job quality as a whole.
Various attributes or indicators of job quality can be organized into several conceptual and abstract dimensions, and distinguishing job quality into two contrasting dimensions such as hard and soft can be a simple but effective method to analyze the impact of job quality [11]. This dimensional classification or mapping can be helpful when linking the job quality of employees with general management issues to compare and select more effective measures for not only the effectiveness and performance at the organizational level but also the health and well-being at the individual level [12,13]. However, few studies have tried to reduce various indicators of job quality into hard and soft dimensions and analyze the effect of each dimension on the health of wage workers. Adapting the job quality frameworks of previous studies [5,7,11e 13], this study defined that the hard dimension of job quality encompasses the elements in terms of physical and contractual working conditions including the physical working environment, working hours, and intensity of work. The soft dimension consists of elements related to social and cultural aspects in the working environment including social support and discretion in the workplace, job prospects, and earnings. Unlike previous research, however, this study examines the dimensionality of job quality using factor analysis, suggesting that numerous job quality indicators can be reduced to higher dimensions of the hard and soft job quality.
Meanwhile, both the hard and soft dimensions of job quality represent the contrasting aspects and conditions of a good job that is not mutually substitutable and congruent to each other. Congruence, in social science research, refers to the degree to which business elements such as job, people, structure, and culture are harmonized or consistent in an organization [14]. The entire organization as a system works efficiently and effectively when these elements of management are consistently matched and fit with each other to support and promote the performance of an organization. Applying this concept of consistency or optimal combination to the concept of job quality, it can be assumed to have a greater positive effect on the health of wage workers if the two dimensions of job quality are implemented together with higher congruence. However, studies on the existence and the effect of congruence between hard and soft dimensions of job quality are extremely rare, failing to provide opportunities to understand the interaction between two dimensions of job quality.
The purpose of this study is threefold. First, it is to investigate the indicators of the job quality of wage workers using raw data from the Institute for Occupational Safety and Health in Korea and to show whether these indicators can be statistically summarized into the hard and hard dimensions. Second, this study will analyze the linear and non-linear effects of the hard and soft dimensions of job quality on the health of wage workers. Third, it will examine the congruence or fit between hard and soft job quality on the overall health of wage workers. By analyzing the relationship between the hard and soft dimensions of job quality from the perspective of congruence, the interactive effect of hard and soft dimensions of job quality on the overall health of wage workers can be more accurately explained.

Materials and methods
This study uses the survey data from the 5th Korean Working Conditions Survey (KWCS) funded by the Occupational Safety and Health Research Institute (OSHRI) of South Korea. This survey sampled 50,205 workers in 2017 from a population consisting of wage workers, non-wage workers, and self-employed workers. Excluding self-employed and non-wage workers, this study used only the samples of wage workers and the final number of samples was 29,818 after excluding observations with missing variables defined in this study.
Following the framework that proposes indicators for measuring seven dimensions, this study established the hierarchical layers or structures based on the questionnaire items of the KWCS that were designed to share questionnaires with EWCS. In the Appendix I, the first layer consists of seven dimensions, the second layer consists of twelve sub-dimensions, the third layer consists of thirty indicators, and the fourth layer consists of eightyseven survey questionnaire items. The indicators that have more than one questionnaire item were represented with the ordinal reliability coefficients which all have 0.7 or greater value enough to justify reliability [15]. Appendix II showed the result of the factor structure of indicators and factor loadings with the reliability and validity statistics for sub-dimensions of job quality. The factor loadings of all indicators exceed 0.6, all the composite reliability coefficients for each sub-dimension exceed 0.7, and all the average variances extracted (AVEs) for each sub-dimension also exceed 0.5. These results confirmed that the indicators compose the constructs of each sub-dimension with good reliability and convergent validity. Appendix III exhibited the correlations among subdimensions versus the square roots of AVEs of sub-dimensions to examine the discriminant validity among sub-dimensions. The table indicated that the square roots of AVEs shown on the diagonal are greater than the off-diagonal elements in the corresponding rows and columns, demonstrating adequate discriminant validity for all sub-dimensions of job quality. It is noted that the subdimension of earnings is not included in Appendix II and Appendix III because it has only one indicator.
All indicators of sub-dimensions were normalized and averaged to build the sub-dimension scores and Table 1 summarizes the subdimension scores of job quality with means and standard deviations.
Instead of using a reflective model to analyze factors affecting the overall health of wage workers and job quality, this study used a formative model to examine the multi-dimensional structure of job quality [16]. This is because the sub-dimensions of job quality or overall health represent many dimensional areas, each subdimension collectively constitutes job quality, sub-dimensions such as physical environment cannot be substituted for other subdimensions of job quality, and each sub-dimension is not assumed to be correlated with other sub-dimensions [17]. Meanwhile, this study used the partial least squares (PLS) method to analyze the factor structure with a formative model, assessing the reliability and validity of sub-dimensions within the framework of the PLS method. Table 2 shows the minimum and maximum values, regression coefficients or indicator weights, variance inflation factors (VIF), and effect sizes measured by f-square coefficients for each subdimension of job quality. This study first investigated the factor structure with the assumption that the seven sub-dimensions of the physical environment, work demands and pace, emotional demands, work duration, working time arrangements, protection, and earnings consist of the hard dimension of job quality. In addition, the five sub-dimensions of social support, work discretion, training, career prospects, and job security were assumed to construct the soft dimension of job quality and test whether job quality can be constructed as two dimensions with twelve subdimensions that this study established based on the factor analysis results as explained above. The result of this initial structural assumption showed that the indicator weight of the sub-dimension of earnings assigned to the hard dimension of job quality was À0.052 with a negative value and effect size of 0.004, meaning that the earning sub-dimension cannot be the component of the hard job quality dimension. Thus, this sub-dimension of earnings was reassigned to the soft dimension of job quality and the result showed that the indicator weight changed to 0.271 and the effect size also increased to 0.161. These findings verified that the subdimension of earnings should be the component of soft job quality and the content and convergent validity of the dimensional structure and twelve components. Table 2 shows the final results that all sub-dimensional indicators have significant relationships with hard or soft dimensions of job quality at the p < 0.001 level, showing the reliability and validity of each sub-dimension of job quality [18]. The values of all VIF are less than 2.0, satisfying strict criteria of multicollinearity and confirming that there was no problem due to overlaps between items [19]. The effect sizes of all the sub-dimensions were greater than 0.24, confirming that all the sub-dimensions have enough practical significance [20]. The discriminant validity of hard and soft job quality was evaluated using the correlation coefficients between these variables and overall health. Table 3 showed that the correlation coefficients between those were 0.294, 0.178, and À0.078, respectively. The upper limit of the 95% confidence interval of the largest correlation coefficients was 0.304, much smaller than the general threshold of 0.8 [21]. This result verified that these variables are not highly correlated to each other and demonstrated the discriminant validity between overall health and hard and soft job quality dimensions. All sub-dimensions of hard and soft job quality were normalized and averaged to form variables representing two dimensions of job quality with values ranging from 0 to 1.
Meanwhile, overall health as the dependent variable was measured using five sub-dimensions or indicators including the level of risk to your health or safety because of work, the degree of negative effects of work on your health, subjective assessment of overall health, the number of longstanding health problems, and the number of physical health problems. The first four subdimensions were measured using responses to each corresponding questionnaire item of KWCS. The sub-dimension of the number of physical health problems was built using ten binary questionnaire items asking about physical health problems including backache, muscular pains, headaches, and so on. The ordinal reliability coefficient of this indicator was 0.899, showing strong reliability of measurements. The ten items were normalized and  averaged to build the sub-dimension or indicator of the physical health problems. This study applied the formative factor analysis method to construct the overall health index using the five subdimensions or indicators explained above because the causality goes from indicators to the factor. Table 2 shows that all subdimensions have significant relationships with the overall health at the p < 0.001 level, all variance inflation factors are less than 2, and all the effect sizes were greater than 0.2. These statistics confirm the structure that the five sub-dimensions constitute the overall health and showed the reliability and validity of the overall health. All the sub-dimensions of overall health were normalized and averaged to have a range of [0, 1] representing the overall health index. This study used five control variables that could influence the relationship between job quality and the overall health of wage workers. First, the socioeconomic status of workers was used to control the effects of the social and economic background of households on the relationship between job quality and overall health. This study used four sub-dimensions or indicators to measure socioeconomic status including household income to make ends meet, the number of years of education completed, occupational prestige, and relative level of income. A questionnaire item on a 6-point scale was used that asked the degree of balance between expenditure and income. The number of years of education completed was measured using the survey items asking the education levels ranging from elementary to graduate schools. The International Socioeconomic Index was used as a single index representing occupational prestige for each job category based on the standard occupational classification [22]. The relative income gap was calculated as the difference between individual income and the average income of reference groups. The number of reference groups was 118, which is constructed using the major categories of the standard occupational classification divided by gender and age. This study applied the formative factor analysis method to construct the socioeconomic status index using the indicators explained above because the causality goes from indicators to the factor or socioeconomic status. Table 2 shows that all indicators have significant relationships with the factor at the p < 0.001 level, all variance inflation factors are less than 2, and all the effect sizes were greater than 0.07. These statistics confirmed the structure that the four indicators constitute socioeconomic status and verified the reliability and validity of the socioeconomic status index. All the indicators were normalized and averaged to have a range of [0, 1] to build the index of socioeconomic status. Second, the age of workers was used to control for the effect of a generational gap because the evaluation of subjective health, as well as working conditions, can differ from generation to generation. Third, the size of the workplace that workers employed was categorized into six groups and controlled for the size effect of the quality of jobs. Fourth, gender was used as a control variable since women and men can have different preferences for job quality and health. Fifth, marital status was used to control for the different motives and objectives for evaluating job quality and health status.

Descriptive statistics
The mean and standard deviation of each variable used in this study, as well as their correlation coefficients and significance, are displayed in Table 3. The correlation coefficients between almost all variables were significant, showing that the statistical analysis models of this study are expected to produce potentially significant results and the intended role of control variables can also be expected. Table 4 exhibits the averages of hard and soft job quality, earnings, and overall health by the occupational classification and the type of employment with inequality index including p90/p10 ratio and Gini coefficients. It is noted that this study used the sampling weight of the survey based on a stratified two-stage cluster sampling to generate population estimates, and the number of samples and population estimates by the occupational classification and the type of employment were reported in Table 4. The level of hard job quality was the lowest in craft and equipment workers, and the soft job quality was the lowest in elementary and agricultural workers. The quality of income, which is a subdimension of soft job quality, is the lowest in elementary workers and service workers, and the overall health is the lowest in agricultural workers and elementary workers. In addition, there were significant differences in job quality and overall health between regular and daily workers classified by the type of employment status. Meanwhile, the degree of inequality for quality of income measured by p90/p10 ratio and Gini index was the highest, followed by soft job quality, hard job quality, and overall health. In summary, job quality and overall health level differ significantly depending on the job classification and employment status, and the overall health increases as the hard and soft job quality increases.
The regression analysis with the least square method was used for all analysis models, and the significance tests of the regression coefficient used robust standard errors that alleviate bias due to heterogeneity between observations. The sampling weight of the survey based on a stratified two-stage cluster sampling was used to estimate unbiased and accurate population parameters including regression weights. Meanwhile, the variables of the hard and soft job quality were mean-centered for the convenience of interpretation of the analysis results. The intercept of each dependent variable, thus, means the average of the estimates of each dependent variable when the job quality is at the mean level.

Results of multiple regression analysis
Model H1 of Table 5 is the result of regression analysis of the relationship between the overall health and the two dimensions of job quality. The effects of hard and soft job quality were all significant at the p < 0.001 level, verifying that the level of the overall health of wage workers was increased as the hard and soft quality of jobs was increased. The standardized beta coefficients of hard job quality and soft job quality were 0.316 and 0.079, respectively, confirming that the effect of hard job quality was about 4 times stronger than the soft job quality on the overall health of workers. Previous studies, not just limited to the response surface analysis (RSA), often have assumed and analyzed a quadratic or curvilinear relationship between working conditions and outcomes [23,24]. In model H2 of Table 4, the squared term of the hard job quality representing a quadratic relation was significant at the level of p < 0.01 in the negative (À) direction. This finding means that the overall health level of workers gradually increases as the hard job quality increases, but the increasing rate decreases following a logarithmic relationship. The square term of the soft job quality was also significant at the p < 0.001 level in the negative (À) direction, showing a logarithmic relationship similar to that of hard job quality.

Results of RSA
This paper used the RSA to analyze the main effect of hard and soft job quality and the congruence effect between the two dimensions of job quality [25]. Based on polynomial regression analysis, RSA visualizes the entire spectrum between variables in three-dimensional space using all possible combinations between job quality dimensions [26]. In addition, RSA provides a statistical means of verifying the existence and direction of agreement or fit between the two variables. This study uses the notations, estimation equations, and acceptance criteria for congruence of RSA suggested by Humberg et al. [25] and Edwards [26].
The response surface plot in Fig. 1 and the contour plot in Fig. 2 show the interactive effect of hard and soft job quality on the overall health of wage workers. The three-dimensional response surface plot in Fig. 1 shows that the level of hard and soft job quality and the positive impact of job quality on health increases as the response surface goes to the left and right corners of the plot. It means that the more the level of hard and soft job quality increases, the higher the positive impact of job quality on health. The twodimensional contour map in Fig. 2 shows patterns of relationship between job quality and overall health with the contour lines that connect points where the combination of hard and soft job quality has the same value for overall health. A scatter plot representing the sample value of hard and soft job quality was overlapped on the contour map to indicate the real ranges of job quality dimensions for convenience of interpretation.    Table 6 summarizes the results of the analysis with corresponding test statistics, confidence intervals, and standard errors to verify congruence according to the recommendations of Humberg et al. (2019). The consistency effect can be determined only when the test results of p10, p11, a4, and a3 are 'effective' by meeting the acceptance criteria set for each test statistic in Table 6. In addition, the test results of a2 and a1 are used to verify the strict condition of congruence that both slope and curvature were not significant. The null hypothesis that there is a congruence effect or fit between two dimensions of job quality was not supported by synthesizing test results and statistics for congruence.
Meanwhile, Figs. 1 and 2 show that the health level is higher in the back corner of the plot where both hard and soft job quality is higher than in the front corner of the plot. For example, Fig. 2 shows that the level of health is higher in the case where the combination of the two job quality dimensions is [þ0.2, þ0.2] than in the case of [À0.2, À0.2]. In other words, the ridge of the plot, where the highest agreement is reached between hard and soft job quality, is rising from the front corner to the rear corner. This result means that the level of health increases as the levels of two dimensions of job quality coincide with each other and increase. In the results of the response surface analysis in Table 5, a1 representing the slope of the ridge shows a significant (p < 0.001) and positive (þ) value. This study, thus, provides statistical evidence that combinations of higher levels of job quality had a greater impact on health than combinations of lower levels.

Discussion
This study analyzed the effect of hard and soft job quality on the overall health of wage workers and the congruence effect between hard and soft job quality. The results of analyzing survey data of the 2017 KWCS are as follows. First, both hard and soft job quality has a positive and linear impact on the wage worker's overall health. The hard job quality consisting of the physical working environment, labor intensity, and working hours lower the threat to workers' health and safety, thereby reducing health problems. In addition, the soft job quality consisting of skills and autonomy, social working environment, job prospects, and wages also have positive effects on the overall health of wage workers. Meanwhile, the standardized beta coefficient of hard and soft job quality was 0.316 and 0.079, meaning that the effects of hard job quality on health are 4 times greater than those of soft job quality. Second, a logarithmic relationship was found between the dimension of hard job quality and overall health in which the increasing rate of overall health decreases as the level of hard job quality increases. Third, the congruence effect between the two dimensions of job quality does not exist. In other words, the cases where the level of hard and soft job quality matches had smaller effects on health than the cases where one of the job quality dimensions was higher or lower than the other. Fourth, the positive effect of job quality on the health of workers was higher  where both levels of job quality were higher than where both levels of job quality were lower. The theoretical implications of this study on job quality are as follows. First, the quality of jobs can be distinguished as two contrasting dimensions of hard and soft job quality. This study established the hierarchical structures by reducing the eighty-seven survey questionnaire items of the KWCS into thirty indicators and consecutively factored the indicators into the twelve subdimensions of job quality. This paper finally identified hard and soft dimensions of job quality using formative factor analysis of the twelve sub-dimensions. In addition, hard job quality consisting of the physical environment, work demands, and working hours has a much greater effect on the overall health of wage workers than the effect of soft job quality including social support, discretion, and training elements. It suggests that focusing on the sub-dimensions of hard job quality can be relatively more efficient measures to improve the overall health of workers.
Second, the logarithmic relationship between the dimension of hard job quality and the overall health of wage workers means that improvement efforts and investments beyond a certain level of hard job quality are less efficient than those below the level of hard job quality. It means that health problems can be solved to some extent if hard job quality increases, but the role of hard job quality has some limitations similar to the role of hygiene factors, making it difficult to expect linear and proportional effects on health.
Third, the effort to increase the level of hard and soft job quality at the same time may be inefficient because the hard and soft job quality are not congruent or fit each other. This study showed that the combined effect of job quality is lower when the two dimensions of job quality are at the same level than the effect when either level of job quality is high or low. This incongruence between hard and soft job quality, together with a higher impact of hard job quality, suggests that the role of soft job quality on overall health is relatively limited. Fourth, the health level of workers is higher when both dimensions of hard and soft job quality are higher than the health level when both dimensions of job quality are lower. These findings imply that rather than just raising the level of hard work quality for health, the level of soft work quality must be raised to some extent to further increase the level of health. In sum, this study found that hard job quality is more effective than soft job quality on the overall health of wage workers in South Korea, and no congruence or fit was found between hard and job quality dimensions. It is noted that the elements of the hard job quality dimension, such as the physical work environment, work demands and pace, and work hours, have been largely controlled by government laws and regulations over decades, but hard or soft job quality has not been publicly measured and no comprehensive policy and action plan in terms of job quality have developed and implemented. More public actions on the working environment and conditions, therefore, are needed not only for the overall health but also for the life satisfaction and well-being of wage workers in this country.
Future research directions are suggested. Further analysis based on longitudinal data on job quality can provide a more causal explanation and in-depth analysis of the degree to which the relationship between job quality and the health of workers. In addition, a future study would benefit from analyzing the structural relationship between job quality and health. Job quality can have an indirect effect on work-life balance and quality of life by mediating health and job engagement. Structural analysis that explains the path or mechanisms from job quality to quality of life can increase the contribution of the study of job quality to the health of workers.

Conflicts of interest
The author has no conflicts of interest to declare.
Appendix I. Dimensions, Sub-dimensions, Indicators, and survey items to construct job quality index with descriptive and reliability statistics