When a contribution table misleads a careful reader

Contribution is the friendlier cousin of full attribution. It asks: given the weights you actually held, which groups added or subtracted pounds in the period. It does not ask whether those weights were ‘right’ versus a model portfolio. That second question needs beginning weights, ending weights, and a benchmark that can be sliced the same way. Many charity files cannot support it.

The first way a table lies is coarse grouping. ‘Equities’ as a single row in a portfolio that holds a concentrated UK smaller-companies fund next to a global tracker will show a number that belongs to neither. We would rather print six rows the board already uses in its policy than three elegant ones.

The second way is mid-period cash. A large redemption sitting in sterling for three weeks will drag ‘cash’ and will also change the weights of everything else. If we do not show the cash row, the equity row looks guilty. If we do show it, someone will ask why cash ‘underperformed’. Both reactions are human; the caption has to do the work.

The third way is interaction. When a sleeve is both overweight and the worst performer, a Brinson-style split can be written. When we lack the benchmark constituents, we refuse the split. A made-up allocation effect is worse than no allocation effect.

We also watch for look-through. A fund-of-funds line labelled ‘absolute return’ may contain listed equity. Contribution at the line level will not match a trustee’s mental picture of the sleeve. Where we have look-through from the manager, we print a second table and we say which table the minute should cite.

The sentence we try to leave the board with is modest: these pounds moved in these groups, under these weights, with these holes in the file. Anything grander belongs in a strategy paper, and that is not the pack we are hired to write.