This is more of a philosophical or historical question, and I can be totally wrong in what I am about to write next.
It looks to me, that complex-analytic geometry has lost its relative positions since 50's, especially if we compare it to scheme theory. Are there internal mathematical reasons for why that happened?
As we all know, lot's of techniques which were later adapted by algebraic geometrers were originally developed in the complex analytic setting (sheaves, local algebra machinery, etc.). In the 50's, Serre, Cartan, Grothendieck and others seemed to have been developing scheme theory somewhat in parallel with complex-analytic spaces (results on coherent sheaves, base change and cohomology theorems, etc.). But already in the 60's it seems like complex analytic geometry started to lag behind – Grothendieck developed Hilbert scheme in 61, and it took already 5 years for Douady to build a complex-analytic analogue. Then in the 80's came Fulton's monograph on intersection theory, and as far as I can tell intersection theory monographs in the complex-analytic category are starting to appear only now. Finally, there is a lot of foundational algebraic work on stacks, and it seems that the amount of publications on complex-analytic stacks is far smaller.
And in general, my impression is that the amount of people doing foundational work in complex analytic geometry is minuscule compared to algebraic geometry (in the recent years I've only come across works of Mauro Porta about the derived complex-analytic set-up and analytic stacks), compared to how active it was before 1990's (there was a nice summary of results by Grauert and Remmert here going back to 1996, showing how active the field was).
Edit:
-
As others have pointed out in the comments, I should have been more specific about what I meant by complex-analytic geometry, since areas like Kahler geometry are very active right now. I basically meant that fundamental side of complex analytic geometry, which shared machinery with algebraic geometry in the 50-60's.
-
The point of asking this question at all is (besides curiosity) to learn whether there are some well-understood limitations to the subject. As I said in the main post, it appears that the number of people in the 50's working on scheme theory was comparable to that working on complex-analytic geometry, and I find the decrease in the number of people in the second cohort rather surprising as we keep getting new powerful techniques which originate from the complex-analytic treatment (like the creation of multiplier ideal sheaves among other things).
Best Answer
Though I am not an expert on this I think that the shift toward algebraic geometry is not entirely sociological. Consider the following statement which is true in both the category of schemes and analytic spaces :
The push-foward of a coherent sheaf by a proper map is coherent.
In algebraic geometry, this statement is rather a routine-like statement, once you have the tools crafted by Grothendieck. In the complex-analytic setting, this is a hard theorem (due to Grauert and Remmert) and no simple proof of it is known.
Another result one could be interested is Mori's bend-and-break lemma. It is probably one of the most important tool in modern birationnal geometry (and was celebrated as one of the most important results in algebraic geometry in the late 70's early 80's). The original proof goes through caracteristic $p$ techniques (namely the use of the Froebenius morphism to amplify a given vector bundle).
Siu and others have claimed to have a purely analytical proof of the bend-and-break lemma. But as far as I can tell (I am not an expert, but I know a bit of birational geometry) their analytical proofs are very hard to follow.
In my opinion, many people in the 70's and 80's left analytical geometry to embrace algebraic geometry because so many powerful and beautiful results had simple and crystalline (in the non-Grothendickian sense) proofs in the algebraic category while their analytic counterparts looked, at the time, either out of reach or extremely hard to prove.
A very down-to-earth baby example at the undergraduate level is the following:
In the algebraic category, there is a one-line proof using Euclidean division. In the analytic setting (that is for analytic functions over $\mathbb{C}$), you have to work a little to prove this.
Note however that the Empire may be striking back. Indeed, outside of Siu's and Demailly's circles, which are, in my opinion, not very active anymore, there is a new approach to the minimal model program using the Ricci-flow and the techniques Perelman introduced to prove the Poincaré conjecture. In a word, the new idea is to start with a (smooth) variety whose canonical bundle is not nef, run the Ricci flow on it and hope that it will converge to a minimal model. So, at least as far as birational geometry is concerned, it seems that analysis is back!