# Idiosyncratic risk

Idiosyncratic risk is the risk of a stock moving for company-specific reasons rather than because of broad market movements.

## What it means

A stock can move because the whole market moves or because something particular to its company changes. The second source is idiosyncratic risk. Examples include an unexpected earnings result, a product recall or a change in management. Broad changes in interest rates or market sentiment are generally sources of systematic risk instead.

Company-specific risk can be reduced by holding stocks whose company-specific outcomes do not move together. It cannot necessarily be eliminated: a portfolio concentrated in one stock remains exposed to that stock, and stocks within the same sector may share risks that a simple market model does not capture.

## How it is calculated

One common estimate uses a regression of a stock’s returns on a broad market index. The variation left in the regression residuals is the estimated idiosyncratic component:

`r_stock,t = alpha + beta × r_market,t + epsilon_t; idiosyncratic variance = Var(epsilon_t)`

- `r_stock,t` is the stock’s return during period `t`, usually expressed as a decimal.
- `r_market,t` is the index’s return over the same period and in the same currency.
- `alpha` is the fitted intercept, in return units per period.
- `beta` measures the stock’s sensitivity to index returns; it has no units.
- `epsilon_t` is the residual: the stock’s actual return minus the return fitted by the model.
- `Var(epsilon_t)` is the variance of those residuals, in squared return units. Its square root, `sqrt(Var(epsilon_t))`, is idiosyncratic volatility, in return units per period.

Use a consistent return frequency and historical window. The result depends on the chosen index and model: a residual can include industry-wide movements, measurement noise or effects the model missed. This is an estimate, not a permanent characteristic of the stock.

## Worked example

Suppose, hypothetically, a regression on daily returns produces an intercept of 0% per day and a beta of 1.2. On a day when the index returns 1%, the model’s fitted stock return is `0% + 1.2 × 1% = 1.2%`. If the stock actually returns 3.2%, that day’s residual is `3.2% − 1.2% = 2%`.

One residual does not measure risk. Suppose the residuals across a longer, hypothetical set of daily observations have a variance of `0.0004` when returns are expressed as decimals. Daily idiosyncratic volatility is `sqrt(0.0004) = 0.02`, or 2% per day. This describes historical residual variation, not the stock’s likely return tomorrow or its total daily volatility.

## Why it matters before you trade

A basket’s number of stocks alone does not show how much company-specific exposure it carries. A large weight in one stock can make one company’s news important to the whole basket. Several stocks in the same industry can also respond to the same event, even if their ticker symbols differ.

A pre-trade check can examine position weights, concentration and sector exposure alongside volatility. Those measures answer related but different questions: concentration shows where exposure sits; an idiosyncratic-risk estimate attempts to separate stock-specific return variation from market-linked variation. Neither measure says whether a trade will succeed.

## How PreTrAIde uses it

PreTrAIde’s published product facts do not say it calculates regression residuals or reports a dedicated idiosyncratic-risk estimate. Its Pre-trade analysis instead provides basket and stock-level risk assessments and 0–100 volatility scores. Its quantitative layer also calculates daily volatility, a 1-day 95% value at risk, portfolio breakdowns by sector and other categories, and a Herfindahl-Hirschman concentration index. These can help put company-specific exposure in context, but they should not be read as a direct measurement of idiosyncratic variance. Its GO, NO-GO and MODIFY verdicts are decision support, not recommendations.

## Common mistakes

- **Equating it with total volatility.** Total stock volatility includes both market-linked and residual variation.
- **Assuming diversification removes every risk.** Shared sector exposures and broad market movements can remain.
- **Treating a single news event as a risk estimate.** An event may explain one move; risk estimation requires a defined model and a series of observations.
- **Treating a residual as purely company-specific.** An incomplete market model may leave broader influences in the residual.

---
Source: https://pretraide.com/glossary/idiosyncratic-risk