Competition for Order Flow

In the competition between Coke and Pepsi, imagine a day on which some technical snag causes Coke to be unavailable all over the country. On that day, sales of Pepsi would roughly double, and the top management of Pepsi would celebrate.

In the competition between NSE and BSE, on the first day that a technical snag caused NSE to be unavailable all over the country, volumes at BSE dropped by 40%.

The surprising outcome of this `natural experiment' has served as a powerful reminder of the non-obvious complexities of the securities industry, of the ways in which this competition differs in a basic way from the `normal' competition that we are all used to in product markets.

In order to fit the exchange into familiar concepts, let us fix notions of firm, product and price in the securities market. The exchange is a firm. The product that an exchange sells is the ability to do transactions, and the price is the transactions costs imposed upon the user (i.e. buyer of liquidity services). This is not immediately obvious - transactions costs are a silent bleed that all users suffer. Speculators are used to forecasting prices and taking positions: whether they are right or wrong, transactions costs have to be paid. The term `market liquidity' is often used loosely; economists define a liquid market as one with low transactions costs.

It turns out to be very useful to decompose transactions costs into two parts, which may be termed systemic cost and impact cost.

Systemic costs are comprised of a host of problems, like inaccessibility, brokerage fees, bad paper, counterparty risk, gala, etc. For the major part of India's history, basic administrative failures have plagued markets and generated large systemic costs. Systemic costs have been drastically reduced since 1994 using new ideas as VSAT technology, free entry into brokerage, the depository, novation by the clearing corporation, screen based trading, dissemination of quotes over the Internet in realtime, etc.

If we narrowly focus upon order placement itself, then impact cost dominates. The buy and sell prices on a market, (called bid and offer prices), are never equal: buyers on a market are always willing to pay less than what the sellers on the market demand. For example, a share might trade at 9.5/10.5, which means that purchasing 100 shares costs Rs.10.5 but sellers get only Rs.9.5. The "midway quote", Rs. 10 in this case, is often used as a benchmark price. In this example, the buyer faces a price which is Rs.0.5 worse than the benchmark, i.e. an impact cost of 5%. The moment someone buys 100 shares at Rs.10.5, he has immediately made a loss of Rs.1, because the liquidation value of the shares is only Rs.9.5.

Impact cost worsens dramatically for larger transactions. While the impact cost on Infosys on NSE is typically 0.5% at 100 shares, it degrades to 1.5% at 1000 shares. Speculative transactions suffer two impact costs: one on buy, and one on closeout. A speculator who takes a position of 1000 shares of Infosys always loses 3% (plus systemic cost), whether or not his forecast pans out correctly.

The behaviour of market liquidity, and the competition between exchanges, hinges on five key facts.

  1. For squaring-off speculators, impact cost is the dominant cost. Systemic costs are dramatically lowered for squaring-off transactions (which helps explain the appeal of this mode of functioning), so the major cost faced by squaring-off speculators is the impact cost at buy and impact cost at sell.
  2. Impact cost is determined by the order flow. A market has low impact cost if numerous buyers and sellers are standing at the screen, ready to buy or sell. This leads to an extremely paradoxical fact about the securities industry: impact cost is determined purely by the flow of orders that come to an exchange.
    This is in sharp contrast with product markets. When one person purchases a car, the price that he pays depends weakly upon the number of people who are buying the same car. In the securities industry, the impact cost that one person pays when using an exchange depends entirely upon the mass of users that the exchange is able to attract.
  3. Impact cost has a natural monopoly character. Hence the securities industry has a natural monopoly character, in the sense that once users are going to one exchange, it has low impact cost, which gives incentives to other users to come to this exchange. When a second exchange tries to attract users, it faces the vicious cycle of low usage leading to high impact cost leading to low usage.
    NSE faced this when NSE first got started. NSE faces this today, when users in India continually trade for physical delivery even though trading for settlement through the depository is now available. The BSE faces this problem: NSE has the liquidity, and it is very hard to take liquidity away from an existing leader.
  4. Lowering systemic cost helps improve the impact cost. For a user of the market, the total cost of transacting is the sum of systemic cost and impact cost. Hence every reduction in systemic cost makes transacting cheaper, so people transact more, which helps reduce impact cost.
    NSE is one of the only examples in the world of a second exchange successfully taking liquidity away from the leader. The essence of understanding NSE's success is the administrative improvements, as compared with the BSE, which led to a collapse of systemic cost. This left users with lowered costs when they first contemplated using NSE, even though impact cost at NSE was high. This led to a first trickle of users coming to NSE, which improved impact cost on NSE, which further brought users, etc. This sort of dynamics explains how NSE suddenly moved from the second slot to the first.
  5. Arbitrageurs support weak exchanges. The picture painted so far is fairly gloomy as far as competition goes: this is a picture where orders rapidly converge on the exchange with lowest impact cost, thus starving other exchanges of revenue, and eventually leading to their closure.
    The survival of exchanges other than NSE is partly based on "product differentiation", such as weaker margin requirements and implementation, differences in trading cycles and settlement practices, etc. In addition, a major factor which protects the survival of other exchanges is arbitrage. Arbitrage equalises prices between exchanges, and serves to transmit liquidity from one exchange to another. The natural monopoly character of liquidity pulls all orders to NSE, but arbitrageurs ferry the prices and liquidity from NSE to the satellite exchanges, and help them be a viable alternative, atleast for small orders.

Using these ideas, we can interpret the experience of 6 October. BSE's volume on Monday was augmented by some of NSE's intrinsic order flow which moved to the BSE, and hurt by the absence of BSE/NSE arbitrage. The ratio of NSE vs. BSE in trading volume is normally around 1800 to 1000, but BSE's trading volume dropped to Rs.600 crore on Monday. This suggests that BSE's intrinsic order flow is less than Rs.600 crore, and NSE's intrinsic order flow is more than Rs.1400 crore.

Back up to Ajay Shah's media page