I conduct validity tests for the RD design by examining two features of the underlying assumptions for the RD design. First, individuals do not have precise control over the forcing variable in the neighborhood of the cutoff point. According to this context, the running variable age is unlikely to be manipulated.
Second, I have chosen to test the validity of the RD design by identifying whether other variables correlate with the jump in the probability of retirement for females at age 50. The variables being tested include education level and marriage status. The continuity assumption of the RD design considers the influence of any other factors that might affect the outcome of interest trends at age 50.
Fig 9 and Fig 10 indicate that these variables do not significantly change around age 50. These findings are confirmed by the regressions that are reported in columns (2)-(3) in Table 7 and Table 8, which indicate that the coefficients of the education and marital statuses are insignificant.
Taken as a whole, both the education levels and marital statuses trend relatively smoothly around the age thresholds. As such, they are unlikely to confound my analysis of the impact of retirement on medical expenditures.