There is a potential issue with normalizing independently… Normalizing works on the peaks. Peaks don’t correspond that well with perceived loudness, and with vinyl transfers you can have a “click” in one channel that represents the loudest peak. That one “little” click becomes your normalization reference for that channel.
One way to do this (other than by-ear) is to check the RMS (or average) levels for each channel, and adjust-down the louder channel (by the dB difference) to match the RMS level of the quieter channel. The problem is… The [u]Wave Stats[/u] plug-in for Audacity can’t analyze the whole file. 
[u]SOX[/u] (FREE!!!) has a Stats “effect” that can read the RMS levels.
It’s best if you start by normalize both channels independently before balancing. That gives you a starting-point to make sure the overall file is normalized/maximized after channel-balancing. Alternatively, you can match the RMS levels first and then normalize non-independently to maintain the balance.
It’s also a good idea to make one big file for the whole album so that you don’t mess-up the original volume relationships between the files (quiet songs remain relatively quiet, and loud songs remain relativley loud).
Or if you want a “one-click” solution, [u]GoldWave[/u] ($50 USD) has a MaxMatch effect that maximizes (normalizes) and matches (balances) at the same time.