bayesline.api.equity.BrinsonColumnSettings#
- model BrinsonColumnSettings#
A Brinson allocation, selection or interaction effect of the active book.
Each day, each group
iof the queried assets has start-of-day weightsw_P = N_P(i) / N_P(R)andw_B = N_B(i) / N_P(R)and returnsr_P = PnL_P(i) / N_P(i)andr_B = PnL_B(i) / N_B(i), whereNis the start-of-day NAV of a leg’s holdings,Rthe reference book, and the benchmark is carried onto the portfolio’s NAV; the reference book’s own returnsr_P(R)andr_B(R)are its two PnLs overN_P(R), so NAV the benchmark does not cover is cash. Under Brinson-Fachler, allocation is(w_P - w_B) (r_B - r_B(R)); under Brinson-Hood-Beebower it is(w_P - w_B) r_B. Selection isw_B (r_P - r_B)and interaction(w_P - w_B) (r_P - r_B)under both. Over groups that partition the reference book the three effects add to its active returnr_P(R) - r_B(R), under Brinson-Fachler where the benchmark covers the whole NAV, so an effect is not additive over a partial set of groups. Where the benchmark holds nothing in a group, its return there is taken asr_B(R), and where the portfolio holds nothing, asr_P(R); a group neither leg holds has every effect zero. A book with no benchmark has every effect NaN, and a fully benchmarked book that is its own reference, taken as one group, has zero allocation and interaction and a selection equal to its active return.denominator="net"gives the effects as returns on the reference’s NAV,nonein currency, each timesN_P(R), andgrossas currency over the reference’s gross.dates="sum"adds the daily effects;dates="compound"links them with the namedlinkagainst the reference’s two daily returns, so it needsdenominator="net"and one ofcarino,mencheroorfrongellowherever both are set. An effect is not summed over its window.variantnames the decomposition.Fields# Field
Type
Required
Default
Literal[‘sum’, ‘compound’, ‘sum_over_window’] |DefaultNo
NoneLiteral[‘none’, ‘net’, ‘gross’] |DefaultNo
NoneLiteral[‘brinson’]No
'brinson'Literal[‘book_difference’, ‘carino’, ‘menchero’, ‘frongello’] |DefaultNo
NoneMetricTYes
Literal[‘own’, ‘root’] |DefaultNo
NoneLiteral[‘Brinson-Fachler’, ‘Brinson-Hood-Beebower’] |DefaultNo
None- field metric: MetricT [Required]#