For me, are the individual parts good investments in and of themselves? Most portfolios are trying to reduce the volatility of stocks by adding something that has a negative correlation. But in and of themselves those things are almost always horrible investments. As a generic statement, "Are bonds a better investment than stocks?" "Is gold a better investment than stocks?" If the answer is "no" or "no, but" then why would you want to own a sub par investment?
You do it to reduce sequence of return risk. Here's an example, constructed using Tyler's calculators over at portfoliocharts.com, using data from 1972 - 2015.
100% Total Stock Market (capitalization-weighted index covering the entire US stock market)
Real CAGR = 5.87%
Worst year = -37.1%
Longest draw-down = 13 years
40-yr safe withdrawal rate = 3.9%
Now let's diversify to remove large-cap bias and US bias. Divide evenly between US stocks and international stocks, with the US portion divided equally among large cap blend, mid cap blend, and small cap blend; and the international divided evenly between international developed and international emerging markets.
Real CAGR = 6.37%
Worst year = -40.6
Longest draw-down = 10 years
40-yr safe withdrawal rate = 5.4%
So the diversification increased the SWR a good bit, with only a little increase in downside volatility. But that -40.6% year was pretty scary, and 10 years is still a long time to see your net worth lower than it used to be (note that the draw-down from the heat map calculator is not applying any withdrawals, so if you were withdrawing assets to fund retirement, the draw down would be longer).
Let's add some bonds in an attempt to tame the beast. Put 30% of the portfolio in total bond market, with the rest divided among stocks in the same proportions as before.
Real CAGR = 5.74
Worst year = -26.9%
Longest draw-down = 9 years
40-year safe withdrawal rate = 5.2%
That tamed the volatility some, with a minimal hit to the safe withdrawal rate. And note that the SWR was substantially higher than if we had just dumped everything in a cap-weighted US stock index.
You could also consider adding a REIT and/or commodity index in place of some of the stock exposure. Because of the low correlation between these assets and stocks, the volatility would have been reduced substantially and the SWR would have increased relative to both of the stock-only allocations. Of course, no one knows if history will repeat itself in this regard. The more you get into slicing and dicing more narrowly defined asset classes, the more the exercise starts to resemble data mining. But the point is that a reasonable amount of diversification has the potential to lower volatility without materially impacting the SWR (or perhaps even increase it).