Here is a paper by IIMA working group that argues for automatic detection of market manipulation near expiry.

The essence of the paper is that, SEBI has to detect fraud and punish the manipulators, rather than putting measures to prevent fraud( which has proven inadequate in general). This paper is inspired by an episode in the Indian market where a group of entities resorted to manipulative trading, who were later barred by SEBI from trading in the capital markets.

In the Indian scenario, where the futures are cash settled, an arbitrageur tends to be price insensitive in closing out the position at the expiry. Let’s say an arb guy shorts futures and goes long in the cash market at time t1. On the day of expiry in the last half hour, he merely has to sell stock in the cash market at a price that is equal to the average settlement price in the futures market. He uses slicing algo or a participation based algo that is dependent on the volume and not price. He doesn’t care about the price as he will have locked in the profit by selling at VWAP in the last 1/2 hour, regardless of the price movement. This situation can be exploited by market manipulators. What does  a manipulator do? He tries to inflate the price in the last 1/2 hour by making a large set of entities allegedly trade with him(circular trading). These different entities that are all somehow related to the manipulator place buy and sell orders to artificially inflate the price so the settlement price goes up. At the same time, the market manipulator in the cash market does not hold any large inventory at the end of expiry. What are the other preconditions for this kind of manipulation to happen ? Trades in the last half hour should be high frequency trades so that genuine bears are out of the market activity.

How does one automatically detect the market manipulation ? One needs to look at abnormal trading volume during specific minutes near the close ? The model employed is a standard regression model that takes in to account the intra-day seasonality ( flexible fourier form), more specifically the last 1/2 hour seasonality

  • Dependent variable is the actual trading volume of a stock at every minute in the last half hour

  • Independent variables are decrease in open interest in the current month, increase in open interest in the next month, roll over, cash market

The way to classify unusual trades is whenever the residual from the regression is more than 3 sigma from the mean of the residual. This basically means identifying outliers in the simple regression model. The authors of the paper use close to 600,000 observations collated from 136 expiry days( last 11 years data) to build a model that can serve as a detection mechanism for unusual trades. Besides using the model, there are several supplemental measures employed by the authors to create indicators for market manipulation. In their training sample, they have found some instances where the model and the indicators point towards market manipulation. These are the cases where nothing has been done. May be they were false positives but the fact that nothing was done to even check whether they were false positive or not, is a case of concern.

The takeaway of the paper is that the regulator needs to employ an econometric model for automatically detecting market manipulation. Also there might be a need to reconsider VWAP as the sole determining factor of the cash settlement price for the futures contract, as VWAP seems to have been manipulated in all the cases reported in the paper.