I finished reading "War: What is it good for?" by Ian Morris. Like "Guns, Germs, and Steel," which it frequently cites, this books takes a cross-cutting approach to history to advance it's thesis, which is that there is both unproductive war, which does the society-destroying stuff we're familiar with, and productive war, which builds ever-larger states; and that since building ever-larger states tends to improve the material well-building and (more solidly) reduce the chance of violent death of their inhabitants, the productive sort of war is ultimately good, since it leads to these relatively safer societies.
I'm not morally repulsed by this sort of discussion, and I agree that as he observes, conquest or force-based coercion has historically been the only way humans have found to build larger polities (the EU is an early experiment on trying another way...).
The biggest problem is that his neat division into productive wars (the unification wars in China, the Punic Wars, etc.) and unproductive wars (Peloponnesian wars, Mongol Wars) is only the sort of classification that's possible after the fact. His criterion basically relies on, "did the winner in the war make a larger stable empire out of it?", which is only a judgment you can make 50+ years later. Virtually all the combatants in virtually all wars claim, and probably believe, that they're going to make a larger, more stable, and richer society -- they just need to get around to winning a bunch of wars first. Thus, for anyone staring into the face of the abyss of a looming war, his thesis holds no guidance, since you can't tell whether this particular war will be productive or not.
Secondly, and more affectingly, he doesn't speak to the fate of the people involved. Sure, the Roman Empire as a whole was more economically efficient than most of its neighbors. That doesn't mean the coming of the Roman Empire was anything to be welcomed by those who started outside its borders. Morris talks about the Roman Empire often in his book (as well as the Han Chinese empire) since we have relatively large amounts of data on them, and it's fair enough to observe that the average citizen of the empire had a lower chance of violent death and had much higher average material wealth than the average citizen of a neighboring kingdom. But the choice of that the average citizen of a neighboring kingdom faced wasn't to become an average citizen of the Roman Empire -- it was to continue being an average citizen or become a Roman slave, who was hardly likely to end up in the middle of the Roman income brackets.
Neither of these objections renders his core observation false. The data he has amassed seems to show that the building of large empires likely decreased the change of violent death and increased average material wealth for those within its borders. But the biggest reason to study history is that "those who do not study history are doomed to repeat it," and his particular thesis has no predictive or imperative power; if the Romans are trying to kill you and enslave your family, you're going to oppose them whether their war is productive or not.
In addition, the book doesn't consider what would have happened had empires chosen either an EU-like federation strategy, or simply chosen to live with the national borders as they were and focused on trade instead. Both of those have certainly been successful strategies and managed to reduce violent death rates and increase material wealth in some examples (seventeenth century Holland, once they decided to abandon the race for empire to the Britsh, is a good example).
That discussion about his main thesis aside, the best reason to read this book is because, like "Guns, Germs, and Steel," it brings a lot of historical data from different threads, ages, and cultures to bear on this hypothesis. I learned a lot more about the various cycles of civilization in India and the Middle East, and his take on the period 400-1200 A.D. ("the steppes riders terrorize the world") is an interesting holistic look at the setbacks civilization encountered in that era.