To answer your post title, I would define decline is defined as - a change in market conditions over time in which an unchanged property would most probably re-sell for less money than it previously sold for, or; given two properties of equal quality, the one that sells later would most probably sell for less.
I agree that one thing won't work on every type of data.
1. A mean or median trace would usually be applied to samples that are larger than the immediate neighborhood. It can be effective with homgenous samples. Variability creeps in as homogeneity decreases. The weakness is applying a broad "rule of thumb" to specific properties can miss things (ie, the older properties are losing value, but the newer ones are not).
4. Price-indexing can be effective. Simple price indexing would be to use certain known properties or to create hypothetical composite comps and appraise them periodically, like once per quarter or once per month. A more complex approach is to quality-point ranking as an appraisal method, then you'd always be show the price change of a similar quality level over time. The weakness here is that this can involve a lot of work.
3. Trend analysis would be graphing (or regresing) sales in the immediate area, comp candidtes. It is similar to mean trace, but usually refers to more the immediate neighbood. This would tend to be more accurate than mean trace and less work than price indexing.
