Background Research has generally found a significant inverse relationship in mortality risk across socioeconomic (SE) groups. This paper focuses on Spain, a country for which there continues to be very little evidence available concerning retirement pensioners. We draw on the Continuous Sample of Working Lives (CSWL) to investigate disparities in SE mortality among retired men aged 65 and above over the longest possible period covered by this data source: 2005-2018. We use the initial pension income (PI) level as our single indicator of the SE status of the retired population. Methods The mortality gradient by income is quantified in two ways: via an indicator referred to as 'relative mortality', and by estimating changes in total life expectancy (LE) by PI level at ages 65 and 75 over time. We show that, should the information provided by the relative mortality ratio not be completely clear, a second indicator needs to be introduced to give a broad picture of the true extent of inequality in mortality. Results The first indicator reveals that, for the period covered and for all age groups, the differences in death rates across PI levels widens over time. At older age groups, these differences across PI levels diminish. The second indicator shows that disparities in LE at ages 65 and 75 between pensioners in the lowest and highest income groups are relatively small, although slightly higher than previously reported for Spain. This gap in LE widens over time, from 1.49 to 2.54 years and from 0.71 to 1.40 years respectively for pensioners aged 65 and 75. These differences are statistically significant. Conclusions Along with other behavioral and structural aspects, a combination of factors such as the design of the pension system, the universality and quality of the health system, and high levels of family support could explain why LE inequalities for retired Spanish men are relatively small. To establish the reasons for this increased inequality in LE, more research needs to be carried out. An analysis of all Spanish social security records instead of just a sample would provide us with more information.