Skip to content

Checking a Book's Beta

A signal made beta-neutral with regression_neut uses estimated betas; the book is exposed to realized ones, after its caps and scaling. This checks how much market exposure is really left. Why it can be large, and how to fix it, is in Market Beta.

from qex import qex, forward_returns, BookSettings, build_book, hold_untradable

signal = qex("""
    let b = ts_beta_shrunk(returns(1), returns(1, 'market'), 252)
    in regression_neut(rank(-returns(5)), b)
""")

settings = BookSettings(delay=1, universe='member', truncation=0.05)
w = hold_untradable(build_book(signal, data, settings), data['close'])   # the book actually held
fwd = forward_returns(data['close'])
pnl = (w * fwd.reindex_like(w).fillna(0)).sum(axis=1)
mkt = forward_returns(data['market']).reindex(w.index)

pnl.corr(mkt)                                  # realized correlation with the market
pnl.cov(mkt) / mkt.var()                       # realized beta of the book

b = qex("ts_beta_shrunk(returns(1), returns(1, 'market'), 252)").evaluate(data)
(w * b.shift(1).reindex_like(w)).sum(axis=1)   # ex-ante beta, day by day

The data needs a market series (Data Format) and a member universe field. To evaluate the signal on its own with the same universe the book uses:

from qex import cross_section_universe

with cross_section_universe(data['member'].notna()):
    z = signal.evaluate(data)          # regression_neut fits over the members only