There are several aspects of this phenomenon, and I don't think it could be summed up easily even if it were better understood.
Tellingly, placebo has most effect on non-specific symptoms. The easiest category is that which relates to perception, such as pain and nausea. These are entirely mental representations of sensorineural input; constructions that exist because we somehow benefit from the averse reaction to whatever we think might have caused it (i.e. we feel pain because it motivates us to run when animals are eating us, and we are nauseated so that we may refrain from eating the typically harmful stuff we associate with the onset of nausea). These are brain utilities and we shouldn't be surprised that we can ignore listening to them just as we can filter out things that happen right in front of our eyes when we're focusing on something else (*).
It's a bit trickier when it comes to things that can be measured objectively, but even in that group, there are symptoms that stand out as extra easy to explain away. Fever falls into this category, as well as, say, a swelling. When a placebo treatment brings your fever down measurably, we know that something more is going on than simply a selective perception. But it doesn't have to be that much more. Fevers and swellings are prompted by infectious disease, but the foreign antigen isn't the direct cause of these symptoms, but rather, it is your body's way of dealing with the issue. You want that fever (well...), because that's your body adjusting itself to optimally cure you. So it's really not that big a deal that you can reduce a fever by means of placebo; sure, if you can trick yourself that you're not infected (or at least, that it's being taken care of), it makes sense that your body's response would be more relaxed, but you'd probably be worse off for it.
It must be acknowledged, however, that there are effects that seem to be purely driven by placebo, where it is not immediately obvious what the mechanisms behind it are. How do we explain entirely physical manifestations, such as how, say, warts disappear apparently entirely due to placebo? Or have a look at this neat effect, where the nocebo effect (which surely must be but a different application of the same phenomenon) provokes a rash:
Another interesting one that demonstrates both the placebo and nocebo effect in one study is a study done by Japanese researches on 57 high school aged boys. They selected these boys based on the fact that when exposed to Lacquer trees, these boys would get a severe rash, much like is common with poison ivy. They then blindfolded the boys and proceeded to brush one of their arms with the Lacquer tree leaves and the other with an innocuous leaf that would have no effect. They told them however that the arm brushed with the laquer tree was brushed by the harmless leaf and the one brushed by the harmless leaf they were told was brushed by the Lacquer tree leaf. What followed was a rash developed on most of the boys arms that were brushed by the harmless leaf and the other arm that was actually brushed by the Lacquer leaf that should have caused a rash was completely fine in nearly every case.
(source)
What's going on here is probably that our bodies are being conditioned to release certain endorphins and chemical substances, leading up to this result. I don't think anybody can account for exactly what is going on in every single case of placebo effect, but I don't doubt these effects are all psychogenic, although their manifestations might be very indirect outcomes, far removed from their provoking agent.