Portfolio rebalancing means that as some assets appreciate and others depreciate, you periodically adjust your asset allocation to stay in line with your original plan. For example, if your bond ETFs outperform your stock ETFs, you sell some of the bond ETFs and re-allocate the cash to purchasing more of the stock ETFs. That way, you maintain your target percentage asset allocation between stocks and bonds.
Why would you do this? First, portfolio rebalancing is a safety mechanism that protects your portfolio from becoming dominated by over-valued assets. Think of what happened to many US investors between 1995 and 2002. During the tech bubble, too many people allowed their portfolios to become dominated by large cap technology stocks and mutual funds as technology company valuations rose, and at the same time they invested insufficient funds in bonds. Rigorous asset allocation and portfolio rebalancing would have prevented this. As tech stocks and mutual funds become more and more over-valued, portfolio rebalancing would have led you to trim your large cap US stock ETF holdings and redeploy the proceeds into bonds. Many people would have avoided much of the pain of the NASDAQ and S&P 500 melt-down, had they rebalanced their assets in this way.
The second reason why you would want to use portfolio rebalancing is that it’s an automated mechanism for buying (relatively) low and selling (relatively) high. As those tech stock funds go up, you sell them; as bond fund prices fall, you buy them. Then, when tech stocks crash, you buy them again, and as bond funds rocket, you sell them. Using the prior example, when you trimmed your stock ETF position during the bubble, you would have purchased more bond ETFs which at the time were cheap and out of favor. They subsequently appreciated considerably. So rebalancing is an automated mechanism for making money.
However, portfolio rebalancing is not without risks. It relies on the assumption that different classes of assets trend upwards together in the long run: while in the short run different asset classes may diverge from their long-term growth rates, reversion to the mean ensures that ultimately they should return to their long-term trajectory.
But is this assumption correct? It may be true of US stocks, bonds and real estate. (Though only a couple of years ago many people claimed that stocks appreciated more over the long term than other asset classes.) But for a long time, it wasn't true for gold or Japanese stocks. If asset classes do not trend upwards together in the long run, then rebalancing pushes an investor to move funds from asset classes that appreciate to asset classes that go nowhere or depreciate.
Another possible objection to rebalancing is that the gain is simply too small to justify the trading costs, hassle and loss of tax efficiency. For this view, see Vanguard's Jack Bogle On Rebalancing: Don't.
So the recommendation to rebalance comes with a caveat: think carefully about the asset classes you are rebalancing. Some should clearly trend together in the long run (small cap and large cap US stocks), while others may not if fundamental structural change occurs. Overall, you should try to use portfolio rebalancing between the key asset classes represented in your portfolio, but ask yourself whether there is any reason why a particular asset class may be in long run decline relative to the others.