We applied Bayesian methods to estimate excess mortality rates by selected causes of death for decedents with diabetes compared to those without diabetes in North Dakota and assessed changes in the excess rate between 1992-1998 and 1999-2003. We report the probability (Pr) of a rate decrease in the age-adjusted excess rate and considered the evidence strong if the probability was 0.90 or higher. Among men with diabetes, the evidence was strong for a probable decrease in excess rate for heart disease (8.7 per 1000 to 6.5), cerebrovascular disease (1.2 per 1000 to 0.75) and arterial disease (0.24 per 1000 to 0.08). Among women with diabetes, the evidence was strong for a probable decrease in excess rates for the overall (total) rate (17.8 per 1000 to 12.6), for heart disease (6.1 per 1000 to 4.4), IHD (4.4 per 1000 to 3.1), cerebrovascular disease (1.4 per 1000 to 0.5), arterial disease (0.17 per 1000 to 0.10) and cancer (2.1 per 1000 to 1.3) as underlying cause of death. The data reflect a high likelihood that cause-specific excess mortality is decreasing for men, and especially for women, with diabetes.