# Kyle’s lambda

Kyle’s lambda is a measure of price impact per unit of signed trading volume, estimated from how much prices move in response to net buying or selling.

## What it means

Kyle’s lambda describes how strongly a stock’s price responds to net buying or selling over a specified interval. In the linear price-impact framework associated with Kyle (1985), a larger lambda means a given signed trading quantity is associated with a larger price change. It is one way to describe market liquidity: a market with greater price impact is less able to absorb that trading flow without moving the price.

The number has no universal unit. Its units depend on whether the calculation uses shares, US$ of trading value, or another measure of order flow—and whether the price change is measured in US$, percentage points, or basis points. A lambda estimated using one convention cannot be compared directly with one using another.

## How it is calculated

A common empirical estimate uses a regression across equal-length time intervals:

`ΔP_t = α + λQ_t + ε_t`

- `ΔP_t` is the stock’s price change during interval `t`, in US$ per share in this version.
- `α` is the fitted intercept, in US$ per share.
- `λ` is Kyle’s lambda: the fitted price change per net share traded, with units of US$ per share of price change per net share traded.
- `Q_t` is signed net trading volume, in shares: buyer-initiated volume minus seller-initiated volume. Positive values indicate net buying; negative values indicate net selling.
- `ε_t` is the part of the interval’s price change not explained by this linear relationship, in US$ per share.

With an intercept, the fitted slope is `λ = Cov(Q_t, ΔP_t) / Var(Q_t)`, provided signed volume varies across intervals. Estimation requires observations from multiple intervals and a method for classifying trades as buyer- or seller-initiated. Different classification methods, price measures, interval lengths, and market conditions can produce different estimates. An observed association does not prove that all the price movement was caused by the trades.

## Worked example

Consider a **hypothetical** stock with three equal-length intervals. Use net shares traded and the change in a reference price, both measured over each interval:

| Interval | Net signed volume, `Q_t` | Price change, `ΔP_t` |
| --- | ---: | ---: |
| 1 | −10,000 shares | −US$0.10 per share |
| 2 | 0 shares | US$0.00 per share |
| 3 | +10,000 shares | +US$0.10 per share |

The mean of each column is zero, so the fitted intercept is zero. A 10,000-share difference in signed volume corresponds to a US$0.10-per-share difference in price change. The fitted lambda is `US$0.10 / 10,000 shares = US$0.00001 per net share traded, per share of price change`.

Under this simplified linear relationship, net buying of 5,000 shares corresponds to a fitted price change of `5,000 × US$0.00001 = US$0.05 per share`. This is a model illustration, not an execution quote: an actual order could face spread costs, changing depth, other traders’ activity, and nonlinear impact.

## Why it matters before you trade

Kyle’s lambda connects order size to potential price movement. For the same signed volume, a higher estimated lambda implies greater fitted impact. That makes it relevant when checking whether a proposed quantity is large relative to available liquidity and when comparing possible execution conditions.

It does not, by itself, give total transaction cost. The bid-ask spread, commissions, timing, and differences between expected and actual fills may also matter. Historical lambda can change, especially when trading activity or market conditions change.

## How PreTrAIde uses it

Kyle (1985) is among the market-microstructure references behind PreTrAIde’s estimates. The product facts do **not** say that PreTrAIde calculates or displays Kyle’s lambda. Instead, its pre-trade analysis estimates bid-ask spread, participation in daily volume, and market impact using an Almgren-Chriss-style model, alongside other basket and stock-level measures. It returns decision-support verdicts and reasoning before users place orders with their own broker.

## Common mistakes

- **Treating lambda as a spread:** Lambda relates price change to signed trading flow; the bid-ask spread measures the difference between quoted buying and selling prices.
- **Ignoring units:** A per-share-volume estimate is not interchangeable with a per-US$-volume estimate.
- **Treating the fitted slope as a guaranteed cost:** It is a historical, model-dependent relationship, not a fill price or a prediction for a specific order.
- **Comparing unlike estimates:** Different intervals, trade-signing methods, and price definitions can change the result.

---
Source: https://pretraide.com/glossary/kyle-lambda