View source

Okay. This one needs some explanation behind the scene. Let's imagine the case of a panorama in the psd file with no embedded layer, just the panorama.

PSD file format has a structure with layers and one background layer. The background layer, is the sum of all other layers ( you can sometimes see this message when saving a big PSD with photoshop : generating full composite, something like that. In fact, it does this background layer ).
The PSD file can have no layer and just a background, one layer and the background or many layer and the background. The background is always here.
You can see pure background PSD file when you open it : it shows a locked picture, you have to double click on it to be able to do something with it. The double click creates a real layer with the content of the background. So if you say it again, you'll get a PSD with one layer and one background. In this case, the background content is of course the same as the layer.

v1.3 : we create the PSD this way : one layer + one background layer ( the background layer is the same as the layer ). The transparency is then visible directly when opening the file. But in fact, the PSD file is just twice as big as needed, because background and layer is just the same data.
v1.4 : we now create the PSD this way : only the background layer. The transparency is then not visible directly when opening the file. But it is still there : just use menu, Select, Load Selection, choose alpha 1 and you have the transparency back. This way you have a PSD smaller with access to transparency.

If we now inspect the case with embedded layer. It's another story. With more than one layer, the background has not anymore the content of the first PSD layer, because there are many not. And if you generate a PSD with at least one embedded layer plus the blended layer, you see transparency everywhere.

Now what : Do you prefer to have a direct access to transparency or file twice smaller ?