Because PNG file size is not determined by bit depth alone. PNG uses lossless compression, so the saved size depends heavily on how well the image data compresses, not just on the raw number of bits per pixel.

Reducing from 32-bit PNG usually means reducing the image data that must be stored, but the file is already compressed, so the size will not necessarily drop by exactly 4×. Images with large smooth areas may already compress very efficiently, while detailed images compress less efficiently.

Also, “32-bit” here refers to pixel data depth (often 24-bit color plus 8-bit alpha), not directly to final file size on disk. If the original included transparency, converting to 8-bit may also remove or change that, but compression behavior still determines the final result.

In short: bit depth affects the amount of image information, but PNG compression and image content determine the actual file size, so a quarter-size result should not be expected.