I plotted all my data from the last 10 years and did a curve fit. The km rate showed a 1 second a year slow down for that period. So average training pace is 10 seconds per km slower than 2010. Not sure about the specific decline in tempo pace as I used the average of all data logged.
There are two confounding factors for me. One is the basic age related decline. The other is the impact increased injury prevalence. The latter would probably account for a greater proportion of the variance in the full data set. Might have a go at the analysis. I joined AP in 2006 and I have ten years of data before that somewhere in an Access file.