I would think that the valence electrons are the most sensitive to bonding, but I am looking at W and its the 4f not the 5d electrons which everyone studies.
Two reasons. First, XPS is not sensitive at the valence band. The cross-sections for excitation are too small, so peak intensities are small. The better tool for valence band spectroscopy is UPS or synchrotron radiation. They probe only to the first 10's of eVs (not the first 1000's of eVs as with XPS). Secondly, the intent of XPS is to look at oxidation or hybridization state changes to first order. In metals, shifts in the first core levels below the valence band give sufficient if not substantially better information compared to UPS in order to answer the key question.
The other practical reason that you see so much XPS and so little UPS by comparison is that XPS instruments exist all over the place whereas UPS systems are an "add-on component" (to XPS systems). But this is just saying the same thing ... In a market analysis, this is an example of product-pull where the popular desire for XPS analysis outweighs that for UPS analysis. Or, in simpler if not rather exaggerated terms, the abundance of materials engineers in the world who define the needs of the tools on the market are more interested in quantifying the levels of dirt on their samples than they are in exploring the fundamentals of precisely how the dirt is bonded at a molecular level, and UPS has absolutely no value for determining concentrations of dirt.
In any case, it is possible to derive information regarding the valence bands by careful observation of the B.E.s of the core shell electrons.
One of the biggest challenges, in my opinion, of using UPS is the broadness of the peaks, with FWHM in the order of 3 or more eV. This makes determination of a binding energy peak difficult. Meaning it will also be difficult to observe shifts of said peak.
If on the other hand, you observe the core shell electrons, you will notice that the peaks are narrow with FWHM of about 1 eV. This makes B.E. determination of peaks much easier. As a result, it also becomes significantly easier to observe shifts in this peak induced by changes in the valence band.
Say for example, carbon with C 1s = 284.7. This is a core shell electron, but its B.E. is affected by hybridization of the 2p orbitals as well as chemical bonding with other atoms. For example, a C=O bond will cause the C 1s peak to shift to a higher B.E. than a C-O bond.
The same goes for metal atoms, where the B.E. will shift depending on the bonding. Hence you observe people studying the 5d orbitals for W. A good database that I can recommend you is the NIST XPS database here: http://srdata.nist.gov/xps/selEnergyType.aspx
As a final point, there are times when the core shell electron does not shift. Case in point, the O 1s orbital. When bound to a transition metal atom, I have observed that neither the valence of the atom nor the atomic number of the TM atom appears to affect the B.E. It remains at 530.0 eV no matter what. Now I have not performed an exhaustive study of all the TM metal oxides, but from my own research and experiments that is my observation.