The probability of informed trading (PIN) variable is simply a proxy for the proportion of trades executed by informed investors. The authors define informed investors as investors trading with superior knowledge of the probability distribution of share prices, through either access to private information or skillful processing of public information. Uninformed investors do not have the same information but are aware of this disadvantage and attempt to profit by following trades of informed investors.

The authors show that when analysts in the Spanish equity market issue a change of recommendation, informed trading increases but uninformed trading experiences a larger increase, causing PIN to decrease. This decrease in PIN happens whether the analyst issues an upgrade or a downgrade. The arrival of uninformed buyers is not significantly different from the arrival of uninformed sellers, suggesting herding behavior. When there is no change, or there is a reiteration, of analyst recommendation, informed trading decreases slightly, but uninformed trading experiences a larger decrease, causing PIN to increase. This effect is most significant for small-market-cap stocks, which are determined to have a higher initial PIN. Large-market-cap companies and stocks with high share turnover ae determined to have a lower PIN and less variation in PIN across all situations.

After a change of analyst recommendation, the presence of informed investors fails to improve efficiency because there is no significant effect on the price discovery process. This finding is evidenced by a delay measure of how long it takes an individual stock to incorporate marketwide information and may be attributable to the analyst recommendation’s attracting uninformed investors who are naive about what the equilibrium price should be.

Knowledge of factors affecting the PIN variable is valuable for market makers and market participants who manage the risk that the counterparty is trading with superior information. Market regulators may also be interested in knowing when trading with information asymmetry is taking place.

Sociedad de Bolsas S.A. provides data on the Spanish equity market for 1997–2010, including transaction date, time, broker codes, price, and trading volume. The authors obtain analyst recommendation data from FactSet. They choose the Spanish market because of prior research in which an environment of lower-quality information and significant herding behavior in this market was identified. Stock information—including market capitalization, average trading volume, and book value—is obtained from the Datastream database by Thomson Financial. These data produce a sample containing 49 average observations for each of the 161 months.

The PIN variable estimates the arrival of informed investors as a percentage of the arrival of all investors, whereby the probability of informed trading is modeled as a function of excessive buying or selling order flow attributable to information events. Informed investors place trades motivated by speculation—buying on good news and selling on bad news—and these order flows are assumed to follow a Poisson process. Uninformed investors are essentially liquidity traders, buying and selling for reasons unrelated to the model according to a Poisson-independent distribution. Whether a transaction is initiated by a buyer or a seller is unavailable in the dataset and is estimated using a tick test, whereby an uptick indicates a buyer and a downtick indicates a seller.

To address heteroskedasticity, the authors use pooled regressions under a variance–covariance matrix over the entire firm-month sample. These regressions include the logarithm of variables—including market cap, share turnover, annualized standard deviation of daily yields, and book-to-market ratio—previously shown to have a significant relationship with PIN. Regressing daily returns on contemporaneous and lagged market returns, the authors create a delay measure similar to an F-test, representing individual stock return variation explained by lagged market returns.

Classifying market participants as informed or uninformed on the basis of a presumption about their knowledge of the probability distribution of share prices presents a challenge when market participants have various time horizons. Automated high-frequency-trading systems and day traders may have sophisticated knowledge about the probability distribution of share prices in the short term and may adjust their holding periods accordingly to profit from very short-term price fluctuations. Although these market participants may not clearly fit the authors’ informed or uninformed classification, they may account for significant order flow—particularly during the later years of the study period, starting around 2005.

The authors mention that company news announcements or other information disclosures that occurred during the study period may have influenced PIN measures, although they focus specifically on analyst recommendations. The PIN variable is susceptible to the estimation bias of nonobservable parameters, the misclassification of trades as buyer initiated or seller initiated, and problems with applying the measure to very high-volume stocks and high-frequency trading. Despite its limitations, PIN remains a useful analytical tool and an area for continued research.