The motivation for this thesis is the investigation of the socioeconomic determinants of health in Taiwan. When considering the variations in socioeconomic status, income is an indicator much discussed in the literature because it is a complex issue and yet easily measured. Many different economists investigate income issue from the perspective of their own particular expertise. For health economists, the influence of income on health outcome is primary. Therefore, many health income hypotheses have been advanced and the debate is ongoing. Among these hypotheses, the absolute income hypothesis, the relative income hypothesis, and the income inequality hypothesis are discussed primarily. The argument of these hypotheses is straightforward. The debates of these different hypotheses are based on two main dimensions, economic development and data. The advocates of the absolute income hypothesis claim that absolute income affects health significantly. The advocates of the relative income hypothesis or the income inequality hypothesis, however, argue that the absolute income hypothesis holds principally before a society moves to an affluent stage. After economic transition, the relative income or income inequality hypothesis becomes more influential. Wilkinson and Pickett (2006) summarize the conclusions of 169 papers relevant to the relative income hypothesis and income inequality hypothesis and find a phenomenon that the studies using large area data are more likely supportive than those using small area data. They argue that income inequality in large area is a good measure of the scale of social stratification or the degree of social hierarchy rather than in a small area. Another argument to explain this phenomenon is aggregate bias proposed by Gravelle et al. (2002). The thesis tries to find the answers for the following questions. What income hypotheses hold for Taiwanese society? Do these hypotheses coexist or are they mutually exclusive? Is aggregate bias a negligible issue in Taiwan when aggregate data are used to infer individuals’ health income relationship? Chapter 3 combines aggregate data and individual data and creates a panel dataset to examine the absolute income hypothesis and the income inequality hypothesis. The motivation is to avoid aggregate bias. Chapter 4 employs nonparametric estimations to describe the relationship between health outcome and income and compares the results of parametric estimations and of nonparametric estimations. Chapter 5 utilizes the quasi-experimental methods to identify the absolute income effect on health outcome. The nonlinear relationship between a) self-assessed health, b) depression, c) life satisfaction and income is found in chapter 4. This finding implies that in the Taiwanese studies the aggregate bias needs to be considered when aggregate data are used to infer individual health income relationship and it is consistent with the motivation of the proposed approach of combining aggregate data and individual data in Chapter 3. The difference between parametric estimations and nonparametric estimations is not only shown in the figures but the model specification test also shows that the parametric linear, quadratic, and cubic forms in terms of income are a misspecification in the estimations of depression and life satisfaction. The absolute income hypothesis and the income inequality hypothesis are supportive in this thesis when long-run income and long-run Gini coefficient are the regressors under the assumption of health social gradient. This finding shows that health income hypotheses are not contradictory. Chapter 5 also provides evidence to support that long-term income has a significant effect on mental health. Thus, the absolute income hypothesis is also supported after taking causality into consideration.