Okay, here's the part where you have to be comfortable with abstraction. If you can buy an "investment" now (let's not worry about what it actually is, yet) that will give you a series of payouts in the future, then it's straightforward to compute the "present value" of that future payoff (mathematically, it's just like mortgage payments -- you adjust the future payments backward based on an assumed inflation rate). As you get closer to the beginning of that series of payments, the value of the underlying investment necessarily increases, because the payments are coming sooner in the future (or, there's less of a discount due to inflation).

Stocks are a little more complex than this because of volatility / uncertainty. There's some chance that there won't ever be a payout, so you're (rationally) only willing to buy the stock if the payout is proportionally higher, and thus worth the risk. (That's what Black-Scholes is all about.)

Still, if you're willing to believe that it's rational to buy stock now because of a payout two decades hence, then you necessarily must believe that the underlying value of the stock increases over time, based on the increasing present value of those future dividends, as estimated by the players in the market. The only way that this breaks down is if you don't believe that the market is rational. Irrationality, of course, causes all kinds of chaos in the markets, but it also creates opportunities for rational investors who wait for irrational behavior to depress stock prices below their rational value.

And, back to Microsoft, the reason that Microsoft's yield is lower than, say, Ford, is because (rational) investors are betting that Microsoft will continue to grow, and thus their dividends will continue to grow. If Microsoft gets back on its previous exponential growth curve, then the current stock price (and dividend yield) look like a bargain.