For several years now, I've had the opinion in the back of my head that "serious" strategy games incorporate several mechanisms that more casual games do away with. Morale is one of the primary ones. When the various dudes who do your fighting in these games value their own e-lives, it vastly complicates things.
When it comes to games, I prefer things on the complicated side. Games like Starcraft or Supreme Commander (or just about any other popular strategy game) allow you to treat your units as disposable cannon fodder. The continued existence of any single unit counts for practically nothing in the grand scheme of things, greatly simplifying the playing experience (this also allows the games to play faster, which is a benefit).
I identify two general approaches to making an individual unit's "life" matter more: Unit-side and Player-side. The latter is about making the Player care about that unit's ability to live or die. This usually involves mechanisms like gaining experience (and thus combat effectiveness) over time, but can be as simple as giving units individual names. Even when they differ only cosmetically from the rest of their ilk, this allows the player to invest some sort of emotion in them (but only works on players so inclined to do so).
On the other hand, you can make the unit care about it's OWN life: thus, morale. I think this has historically been one of the mechanisms designers are quickest to ditch. By necessity it implies developing some kind of AI for those instances in which morale "breaks" and the unit no longer obeys the players actions, but rather starts behaving according to some other directive (usually being, "run away!"). This can not only be frustrating for a player (it lessens their control over the game, always a risky thing especially with more casual players) but time-consuming for the developer to implement effectively. If a routing unit doesn't behave at least somewhat intelligently, it only aggravates the player further.
In my current project I'm developing the morale system right now. The actual numerical book-keeping is relatively easy, though there are quite a few factors to consider (should the unit gain morale by killing enemies?). The hard part is fleshing out the behavior a unit takes when it's morale is broken and it routes. Currently, I'm aiming to satisfy a simple three-step decision process:
1. If enemies are nearby, retreat along the path which will put maximum distance between you and them.
2. If no enemies are nearby, retreat towards the nearest board edge.
3. If multiple board edges are nearby, bias towards the one near your player's start zone.
So, first a unit runs away from the enemy. After that, it tries to get off the map, preferring to get off the map near where it started (retreating back the way it came). But how far away will it look for enemies? What if a unit breaks near to the enemy start zone and there is a clear path off that edge, but some enemies are somewhat nearby? Generally speaking, units fleeing in panic should behave "intelligently" for a given small set of information, but planning far ahead isn't a requirement.
This is one of those situations where you design with a few simple rules and hope that the emergent behavior seems to make sense...