Do you have access to statistical software (R, SAS, Stata)? If you do, you would want to download the PUMS data the way you need them (ipums.org, in my opinion, provides the best customizable environment to do this: you can easily select just the variables, the samples and the geographies that you need), and estimate the median earnings using the survey commands in these packages. A few years ago I developed a Stata package for estimation of percentiles and their standard errors.