The existence of unfair differences or disparities in access to and quality of health care is well known. However, the nature of disparities at different stages of the health seeking pathway and interventions to reduce them are less clear. Applying the tools of statistics and quasi experimental design-- interrupted time series, propensity score matching, hierarchical models---we can analyze how care is accessed in low, middle and high income countries and assess for disparities. The results are sometimes surprising and underscore the need to generate context specific evidence to ensure targeting of programs. My first paper evaluates the impact of a controversial policy, mandating of health insurance, on reducing disparities in health care access and affordability. Using longitudinal survey data from five states in USA (2002-2009), I show that living in MA, where health insurance is mandated, results in a higher probability of being insured and having a personal doctor and lower probability in forgoing care due to costs as compared to similar border states. The beneficial effect of the mandate is greatest in traditionally "disadvantaged" groups defined by race, income, education or employment status. My second paper examines gender disparities in access to medicines in sub Saharan Africa--Uganda, Kenya, Nigeria, Ghana, Gambia. Using medicines specific survey data, I construct a novel seven stage access to medicines pathway and assess gender disparities along it applying the Institute of Medicine framework. Contrary to prevailing belief, I find few gender differences in unadjusted outcomes which cease to be significant on controlling for health status and country characteristics. My third paper assesses disparities by educational attainment in process and outcomes of care. I use unique data extracted from an electronic medical record of diabetic patients in Mexico City. Using a matching algorithm, I control for only differences in health need and find few significant differences in processes and outcomes of care. The unmatched traditional regression based risk adjustments tend to overestimate the significance and magnitude of the association. The three papers demonstrate the need to use more sophisticated statistical tools to appropriately measure disparities and ensure the effectiveness of health programs.