1: \begin{abstract}                                                         

2: We use the recent estimates of NNLO splitting functions,  

3: made by van Neerven and Vogt, to  

4: perform exploratory fits to deep inelastic and related  

5: hard scattering data.  We investigate the  

6: hierarchy of parton distributions obtained at LO, NLO and NNLO, and,  

7: more important, the  

8: stability of the resulting predictions for physical observables.   

9: We use the longitudinal  

10: structure function $F_L$ and the cross sections $\sigma_W, \sigma_Z$  

11: for $W$ and $Z$  

12: hadroproduction as examples.   

13: For $F_L$ we find relatively poor convergence, with  

14: increasing order, at small $x$; whereas $\sigma_{W,Z}$  

15: are much more reliably predicted. 

16: \end{abstract}