Blame COBOL or PL/1 or a lazy developer for this! If you lay a bunch of blanks (i.e. eight characters with the hexadecimal value of '40') over a packed decimal variable that is the decimal value you get. It's still a valid decimal value so no error will be shown. I remember digging through tons of source code while analyzing the weird behaviour of account no. 40404040. (Yeah, good old 90s...)