Save For Web vs. Resizing vs. File Size vs. PNG

Question

In playing around with some settings while saving a .png file, I noticed some weird results. For this example, the original image is 300 x 300, but I want the final size to be 200 x 200. I get different results depending on the order I perform the following operations (I never thought the order mattered until now).
1. start with 300 x 300 image, choose save for web, select png-24, with transparency, white matte, convert to sRGB, change image size to 200 x 200, then save - the file size is 37kb
2. start with 300 x 300 image, use Photoshop's "image size" to change to 200 x 200, then save for web, etc - the file size is 111kb
So depending on which step I resize the image, the file size is significantly different - if I resize BEFORE using save for web, the file size is much larger. This is just weird to me, but I always resize AFTER choosing save for web, so that's why I've never caught this until now. In case you ask, while using Photoshop's "image size", all three options are checked at the bottom of that window, so nothing is getting re-sampled or anything like that.
The only thing I can think of is each of those methods treat pixel data differently when reducing the dimensions. When I overlay both exported .png files on top of each other, I see no difference in pixel quality and/or color shift - so why the big difference in file size? Unsurprisingly, if I just save the file straight to .png, the sharpness is much better, and the file size is 46kb. I did notice when saving for web, the colors become a little more saturated.
Are all these results typical? I've never really paid much attention to the results when exporting .png's. I always thought the results were lossless (in general).

Answer 1

Too many variables
What? There are 2 variables here (variables = scenarios = steps). Only two different operations.
1. resize the image, THEN use "save for web" = 91kb
2."save for web", THEN change image size = 157kb
Hopefully my logic translates here:
All other settings are the same. I even made sure the .psd was sRGB this time. The specific question was "why does the different workflow order produce such a big difference in file size"?
In creating a new test file (http://www.shan-dysigns.com/userForums/photoshop/savePNG.zip), I did notice an interesting thing: (I'm including my files in case you want to follow along or test for yourself)
If I merge all layers BEFORE performing the steps above, the file sizes were relatively closer to one another - 168kb and 157kb (respectively from the 1, 2 list above)
One thing to note is all the .psd layers are either shapes or text - this whole scenario may be totally different if each layer had rasterized content (actually, it does, by about -20kb).
The file size in step 1 above didn't jump until it had to consider rastered text/shapes into the calculation. So all this tells me the difference in file size has to do with how each process handles vector data. Maybe when you allow PS to resize the image first, the overall file size is smaller because the vector data gets recalculated???
I don't know how to properly interpret all these differences, but I do know there is something about PS's operation of resizing the canvas with vector data versus resizing the canvas with rasterized data - this has to be where the difference in file size lies.
Maybe I've just wasted a bunch of time on the obvious, but I think I learned something here. Now my head hurts and my eyes have popped out of my head.
So, Chris, I guess there ARE more variables in this situation. My curse is being nosey and wanting to know more technically about PS than most casual designers probably would care about.