It’s a valid theory and under some circumstances such micro-optimization might make a difference (making content fit into single packet rather than be split into two).
However you misinterpret the size information. 8.5KB is gzip-compressed version of your page. Since repeats of same string are highly compressible, your optimization won’t reduce this by 5KB.
How much would it actually? From quick test by about 100 bytes. 🙂 Which is completely not worth worrying about.