Adjusted PIN and the Duarte-Young Model

The classic PIN (probability of informed trading) model explains a stock's daily buy and sell order counts using two possible states: an "information event" day where informed traders push order flow one-sided, or a normal day where only uninformed traders trade and buys roughly balance sells. Duarte and Young noticed a problem: some stocks have chronically imbalanced order flow for reasons that have nothing to do with private information — thin markets, inventory-driven market maker behavior, or clientele effects that simply produce more buys than sells on average. The original PIN model has no way to represent that, so it mislabels this ordinary illiquidity as informed trading, inflating PIN estimates for illiquid but not necessarily "toxic" stocks.

Their Adjusted PIN model (AdjPIN) adds a second, symmetric shock parameter to the underlying process: alongside the informed-trading imbalance, it lets uninformed order flow itself have episodes of elevated (but non-directional, or differently-directional) intensity, so a spike in buy-sell imbalance driven by ordinary liquidity effects no longer has to be absorbed by the informed-trading parameter to fit the data.

Empirically, Duarte and Young found that once this adjustment is made, the "informed trading" component of standard PIN loses most of its power to predict returns, while their added illiquidity-shock component picks up that predictive relationship instead — evidence that some of what PIN calls informed trading was actually priced illiquidity risk.