16 & 24 bit files automatically converted to 32 bit float

I regularly use audacity to convert 24 bit files from bandcamp artists to 16 bit and/or 48+khz files to 44.1. Audacity always open files as 32bit float, whether they are 16 or 24 bit, even after changing my preferences. Until now what i always did was to manually change the format in the drop down menu for the file to either 24 or 16 before resampling the tracks (if necessary) and exporting the file to 16 bit. I figured it would give me a more ‘accurate’ result than if audacity exports or edits a 24 bit file like it was a 32 bit file. Now after reading some other forum posts i am not so sure anymore that was the best way of action. (mainly referring to this post https://forum.audacityteam.org/t/audacity-keeps-converting-my-24-bit-samples-to-32-bit-float/39021/1)
Can anyone tell me if there is even a difference whether i change the format manually before doing anything else or just leave it at 32 bit, and if so which on produces the ‘better(more accurate) result’.

There’s no harm in leaving the floating-point default. The re-sampling (of the sample rate) is probably being done by temporarily converting to floating-point anyway.

The results should be audibly perfect in any case, as long as you stay at or above “CD quality”.

Another option to think about…
When you reduce the bit bit depth you’re supposed to dither. Dither is added noise which is preferable to the “natural” quantization noise.

If the original is 16-bits you can turn dither off because upsampling or dithering doesn’t remove existing quantization noise… But, it’s not a big deal at 16-bits or better because under normal conditions you can’t hear quantization noise or dither.

You can hear quantization noise at 8-bits if you want to do some experiments… Quantization noise “rides on top of” the signal and like any regular noise it’s most noticeable at low signal levels. But unlike regular noise, it goes-away completely with “digital silence”.

I figured it would give me a more ‘accurate’ result

Of course, when you downsample you are permanently throwing-away data, but that doesn’t always mean you are damaging the sound. And you probably have a good reason for doing it (such as making a CD?) so you may not have a choice. Increasing the sample-rate is mathematically imperfect and not perfectly reversible. Increasing the bit-depth can be mathematically perfect and reversible (if you don’t dither).