What does it mean for a risk model to treat systematic risk as idiosyncratic?

Risk models split each holding's movement into a part explained by shared factors and a part unique to that holding. If the factors chosen do not capture a source of common movement, that shared component gets misfiled as unique. The portfolio then looks more diversified than it is, because risk the model recorded as independent is in fact moving together.

Is correlation between two strategies the same as correlation between their results?

Not necessarily. A paper in this week's roundup separates the two: correlation measured across holdings on a given day and correlation measured across time between the amounts two approaches earn are computed over different index sets. The first is often used as a stand-in for the second, and the paper sets out exactly what that substitution captures and what it leaves out.

How can a company be exposed to a currency it never borrows in?

Through its suppliers. Research summarised this week describes large firms borrowing cheaply in a foreign currency and then extending credit to domestic customers through accounts receivable. The customer never touches the foreign currency directly, yet its financing conditions move with it, because the terms it receives depend on a supplier that does.

This week in research: the dependence you did not measure

Diversification is a claim about dependence — a statement that the things you hold do not all move together. The research crossing the wire this week is largely about how that claim goes wrong, and it goes wrong in the same way each time: dependence exists in a place the measurement was not looking. Nothing here is a market call. It is a set of reminders that a portfolio is only as diversified as the model used to check it.

A risk model can quietly reclassify shared risk as unique risk. Conditional risk models estimate each holding’s risk relative to a chosen set of information — a factor set, a set of state variables. A revised paper, Screening-Off Information and Conditional Risk in Portfolio Choice, points out that these models rarely test whether the chosen information actually removes common dependence across assets. When it does not, systematic risk gets recorded as idiosyncratic, which distorts both the resulting portfolio and the efficient frontier the investor believes is available. The authors formalise the condition — screening-off — and separate the causal, distributional and second-moment versions of it. The methodological point is unusually clean: a diversification measurement is a conditional statement, and it is only as good as the information it is conditioned on.

Correlation between inputs is not correlation between outcomes. Signal Correlation, IC, and PnL Dependence, posted this week, examines a substitution practitioners make routinely. The correlation between two signals is measured across holdings at a point in time; the correlation between what those two approaches earn is measured across time. They are correlations over different index sets, and the first is habitually treated as a proxy for the second. The paper gives an exact decomposition showing what the proxy captures and what it discards. For a long-term investor the analogue is familiar: two funds whose stated methods look different are not thereby independent in what they deliver, and the only way to know is to measure the thing you actually care about.

Exposure travels through channels a holdings list does not show. A research summary on firm-to-firm financial linkages and dollar risk transmission describes a mechanism that bypasses the usual categories. Large firms borrow cheaply in dollars and then finance their domestic customers through trade credit — accounts receivable. The customer neither borrows in dollars nor exports, yet its financing conditions move with dollar funding markets anyway, because its supplier’s do. The exposure is real and it is invisible to any classification based on what the company itself owes. This blog has made a related argument about instruments whose labels overstate their independence, in when a hedge isn’t a hedge.

Judgement depends on the comparison you happen to be holding. Comparisons, a working paper by Thomas Graeber and Benjamin Enke, addresses a long-standing puzzle in how people decide: sometimes a reference point pulls a decision toward it, sometimes it pushes the decision away. The authors propose a taxonomy that predicts which happens, resting on how difficult the absolute evaluation is. When translating an input into a judgement is hard, people read the comparison point as information about how to navigate the tradeoff. Investing supplies these comparison points constantly — a benchmark, a neighbour’s result, last year’s number — and this work is a reminder that they are not neutral background. They enter the assessment.

Designing a long-horizon system to absorb shocks rather than relitigate them. For readers closer to drawing an income, Peter Diamond’s Pensions and Turbulence examines automatic and semi-automatic adjustment mechanisms in long-term pension design — rules that respond to financial shocks, longevity change and shifting labour markets without requiring a fresh political decision each time. The chapter weighs those mechanisms against adequacy, equity and legitimacy. Read as method rather than policy, it describes something a household plan also needs: a decision made once, in advance, about how the plan reacts when conditions move, so that the reaction is not improvised under pressure.

The throughline. Risk can only be budgeted where it has been correctly attributed. Every item this week is a case of attribution going astray — a common factor misfiled as idiosyncratic, an input correlation standing in for an outcome correlation, a currency exposure arriving through a supplier, a comparison point entering a judgement unannounced. None of this argues for a different portfolio. It argues for a sharper question about the one you have: not “how many things do I hold”, but “what would have to be true for these to be as independent as I am assuming”. That question is the whole of risk-budgeting, which is the subject of what a bank treasury knows about risk.

Curious where your portfolio’s risk structure stands? The free ETF Portfolio IQ Score is one way to see.

Related