Considering Multiple Vintages

Joseph Breeden

In the previous chapter, we explored intuitively the dynamics affecting vintages. The next step is to develop a modelling framework that can reveal those dynamics, which we explore in this chapter. If we move down to the account level, we will, of course, reveal more structures explaining consumer behaviour, but we will not get any closer to quantifying the lifecycle and environmental effects merely by studying individuals. Also, as we saw at the beginning of the previous chapter, receding to the portfolio level only obscures the lifecycle and environmental dynamics.

So, we need to understand the drivers of the vintage, but going either up or down in aggregation provides no assistance. The answer lies in modelling multiple vintages simultaneously and making some assumptions. Figure 3.1 is a plot comparing multiple vintages. Each vintage is one “ribbon”. Since a new vintage begins each month, a triangular data set is created. Visually this suggests immediate analogies to spatio-temporal data sets and our solution will be similar.

Figure 3.1 A two-dimensional perspective on vintage performance showing how vintages can be compared across age and date to gain understanding of the