Something old, something new, something borrowed, something blue

Sorry, I couldn't resist that title.

The oldest and newest "vintages" of data for the series "Household Debt Service Payments as a Percent of Disposable Personal Income" at ALFRED:

Graph #1

What I expected? Maybe the data would go back some years before 1980 and I could use it to extend the current series back to an older start date.

What I got? Everything starts in 1980, so it doesn't do me any good. Plus, the two datasets don't match up at all.

Besides that, the difference varies:

Graph #2

The old series runs from more than 122.5% of the new series, down to about 111%, then quickly up, and down.

I wonder what accounts for the difference.

In How much income is used for debt payments? A new database for debt service ratios by Mathias Drehmann, Anamaria Illes, Mikael Juselius and Marjorie Santos (13 September 2015), the authors discuss the methodology of developing DSR data series:

... amortisation data are generally not available and hence present the main difficulty in deriving aggregate DSRs.

... it is possible to construct meaningful aggregate DSRs with a relatively sparse set of aggregate data ...

... potential mistakes mainly lead to a shift in the level of estimated DSRs, but will not affect their dynamics over time.

Potential mistakes mainly lead to a shift in the level of estimated DSRs, but do not affect their dynamics over time.

To me this means the numbers might all be high or might all be low, but the patterns of the high data and the low data should be quite similar. I think we see this in the graph. And yet, as the second graph shows, the patterns are not as similar as one might hope.