RISK MODEL
Enter your holdings and this free factor risk model splits their risk across 13 style factors, the industries they sit in, and what is left over for each company on its own. It runs in your browser, so your positions never leave it.
THE TOOL
One line per position: a ticker and a weight. Percentages, decimals and dollar amounts all work, and the tool prints which reading it used, which names it could not cover, and the week the model was built from. Holdings from a Portfolio123 strategy work the same way as any other list of US-listed stocks.
Research and education only.
NEWSLETTER
New research as it is published. Sign-up and delivery are handled by Substack.
Loads the Substack sign-up form here. Substack then sets its own cookies, listed in our cookie policy.
Factor models, risk frameworks and research tools for your own universe, built in a quant consulting engagement with an independent practice.
WHAT THE MODEL MEASURES
Every company in the universe carries a position on each of these, measured against the average name that week. An exposure of zero is average; the further from zero, the more unusual the tilt. A company with no usable data for a factor sits at zero.
How cheap a company is against its book value, free cash flow, sales and earnings, each measured against similar companies and read together. Free cash flow and sales do not count for financial companies. A positive exposure means the holdings are cheaper on those measures than their peers.
The past year of price return, skipping the most recent month. A positive exposure means the holdings have been climbing for longer than the average name.
Profitability, read from gross profit, cash flow and earnings against assets or equity, gross margin and low accruals. Financial companies are read on return on equity and return on assets only. A positive exposure means more profitable businesses than the average name.
How much profitability has improved over three years: the change in gross profit, earnings and cash flow per share against assets or equity, and in gross profit against sales. Financial companies are read on the change in return on assets and return on equity only. A positive exposure means profitability has improved more than for the average name.
How steady a company's return on equity has been over the last twelve quarters. A positive exposure means steadier profitability than the average name.
The changes analysts have made to their earnings estimates over about the past six months, each measured against the share price. A positive exposure means estimates that have been rising rather than falling.
How much a company moves when the market moves. This is the market sensitivity of the model, and a positive exposure means calmer names than the average.
Shares sold short against shares outstanding. A positive exposure means the holdings are more lightly shorted than the average name.
Cash returned to shareholders, through dividends and net buybacks of stock. A positive exposure means more of it than the average name.
How much of a company's market value changes hands on an average day, measured against what its short interest already explains.
Liquidity is orthogonalized against Short Interest every week, so this exposure is not how much a stock trades but how much it trades beyond what its short interest implies.
The past month of dollar volume against the ten months before it. A positive exposure means trading in these names has picked up rather than faded.
Market capitalization on a logarithmic scale, standardized across the universe. A positive exposure means larger companies than the average name.
How much a company has moved on its own account over the past year, once its industry and the other exposures are accounted for. This is a company's own specific risk, not its market sensitivity, which is Low Beta.
HOW IT IS BUILT
What the model is fitted on, what comes out of it each week, and where its limits are.
The model is a weekly cross-sectional factor model of about 3,500 US stocks including micro caps. Once a week, every company's return over that week is explained by the industry group it belongs to and by where it sits on the 13 style factors above. Whatever the explanation misses is that company's own return, and the size of it is estimated separately for every name. The weekly regression is fitted on the 3,000 largest of these companies. Every company the model covers, micro caps included, still carries its exposures and a forecast of its own risk, and counts in the three comparisons.
Two things come out of that. The first is a covariance matrix of the factors, estimated over the whole archive with one decay for the correlations and a faster one for the volatilities, so the model changes its mind about the level of risk sooner than about the shape of it. The second is a forecast of each company's own risk for the coming week, built from its history of surprises and from what the company looks like today.
Your holdings are matched to that cross-section by ticker, with the usual class-share spellings resolved to one name. Anything inside the universe is covered; anything outside it, funds and instruments that are not US-listed stocks included, is reported as uncovered rather than estimated.
The model estimates how far a portfolio can move, never which way. Two portfolios with the same risk figure can end a year a long way apart.
Exposures are read from the most recent weekly cross-section. A portfolio that has changed since then is described as it stood on that date, not as it stands now.
How stocks move together is estimated from the weekly archive, which starts in 2006. In a sharp crisis they move together more than an estimate built on calmer years expects.
Names outside about 3,500 US stocks including micro caps are listed as uncovered. Coverage is printed on every run, with a warning when much of the book falls outside the model.
Walk through the numbers in a free 45-minute introductory consultation.
Request an introductory consultationOr write to: c.morales@quantsolvings.com
