The earlier GTA games where quicker to develop because the computers did not have the CPU power or RAM to support large cities, many different NPC models, detailed cars, objects or enviroments. This means levels, assests and the engines where much quicker to develop.

if you double width+height of your map, you have to fill 4x the area with meaningfull content, belivable cities, scripted events, nature, streets, shops, maybe NPS who have a daily routine to walk from one place to another.

Plus, you have to create a large amount of different(!) buildings to place around your map, because you dont want to recognize the same house, shop, park or Parking lot multiple times in the same street. On top of that is car physics, traffic simulation, cop chase behavior and now apparently NPCs interact with each other.

None of this has been done before, and getting this right and bugfree is not trivial - see cyberpunk.