The story goes: an economist was walking on the road with a friend, who said:
"Look, there's a 100 dollar bill lying on the ground, Professor"
"It can't be. If it was, someone would have taken it already", the economist replied.
While that sounds absurd, there still needs to be someone picking up those bills. In market parlance, the act of buying something and selling it almost immediately at a higher price is arbitrage — what might seem to be "risk-free" money. Let's just look at the field, from an Indian market context.
The BSE and NSE are the two most popular exchanges in India, and many stocks trade on both of them . If there is a price difference in a stock's quoted prices in each, you can buy on one exchange and sell on the other. You should theoretically not lose any money regardless of which way the market goes. As more people do this, the prices on the two exchanges should converge.
That's technically true, but there are many reasons why there is a gap in the prices, and no one is picking up the free money. Transaction costs can be too high — apart from brokerage, you have clearing costs, exchange fees, stamp duty and securities transaction tax — sometimes high enough to have a healthy gap. And then you have execution risk — what if you get one trade but not the other because prices moved?
In India, you must square off an inter-exchange arb trade within the day for structural reasons. With a T+2 rolling settlement system in India, you need to deliver stock that you sell on the next day of the trade, but you only receive what you buy two days after — so it's perilous to hold this trade through the end of the day. (If you reverse the trade within the day, it doesn't need to be settled) But then, an intra-day "square-off" means that prices must converge within the day — and if that doesn't happen, you might have to square off at a loss at 3:25 p.m.
Buying a future is like buying a stock, except that it is settled at a later date. The futures market therefore quotes separately from the cash market — and there might be a 1-2% difference between the two. On the expiry date of the future, the future is "settled" at the closing price in the cash market — which means the prices in the two markets converge. Additionally, you can sell a future without owning any stock; the arbitrage, then, is to buy stock and sell a future at the same time, pocketing the difference. On the expiry date, everything should converge and you'll make your profit.
The risks? Apart from transaction costs and execution risk, you have another structural issue — the "price of convergence". The closing price of a stock isn't the last traded price of the day — which can be manipulated — but a weighted average of all the trades in the last half hour. You need to sell your stock at what you think will be the closing price on expiry day. But the last half hour can see wild swings — wild enough to make your returns zero, or even negative.
There's also the cost of capital. At the very least, you want the 8% return the banks offer you at virtually no risk. That's 0.67% per month, below which you actually lose money — any arbitrage not yielding that much is just not worth it.
Statistical Arbitrage (Stat-Arb)
The above forms of arbitrage come with a reasonable probability of convergence — that is, what you buy and what you sell will meet each other in price some day. A different way is to bet that two prices will converge, because they "tend to do so". Looking at past data, you might analyze two stocks in the same sector, like Infosys and TCS, and determine that they tend to move in tandem — if one stock goes in a direction faster, the other eventually catches up. The trade then is to identify "divergence" from an equilibrium point and taking a trade by shorting one stock (using futures) and going long the other, betting that prices will again move back to the equilibrium. But a pair trade can go horrendously wrong, as investors who were short Infosys and long Satyam discovered in December 2008, when Satyam fell 50% in a single day on discovery of a fraud.
There is a way to simulate a position using other instruments. For instance, you could buy a stock by selling a put option and buying a call option at the same strike price. Match that against buying the stock or future outright, and you might have an arbitrage.
Some Indian stocks are listed in the US as ADRs, and the ADR price, after conversion, can be arbitraged with the Indian price with a hedge on the USD-INR exchange rate. But some US ADRs quote at a significant premium to India; the reason? In most cases, you can buy ADRs in the US and convert them to shares for the Indian exchanges, you're not allowed to go the other way around, so what seems like an opportunity, isn't.
Two stocks that decide to merge might quote independently until the merger goes through, and you could capitalize on the difference between the market price ratio and the merger ratio. There is the downside of a breakdown in the merger, which happened when Bear Stearns was initially offered to be bought by JP Morgan for $2 a share, but they revised it to $10 later.
So, are there more risks?
Apart from the risks above, you have the risk of liquidity. Typically, arbitraged positions are considered lower risk due to their nature of being market-neutral. So much that even regulators put in lower "margin" requirements during stable periods; consider for instance that to buy a nifty contract worth around 300,000 you need to put in a margin of only 10% - or about 30,000 — despite the Nifty having moved 10% in a single day four times in the last three years. To make 1% returns on Rs. 30,000 means only needing Rs. 300 in a month. But should the market get really volatile, NSE will increase the margin requirement — to what might end up being Rs. 60,000. Suddenly, you end up with a good trade getting you less than you would have made in a safer bank deposit. Worse, if you only had Rs. 30,000 to begin with, where would you get the additional money?
This problem killed a large hedge fund — LTCM — in the late 90s. The principals of the fund said they would be a giant vacuum cleaner picking up nickels in arbitrage from all over the world; others have likened what they were doing to picking pennies in front of a steamroller. You might get rich, but you will eventually die. The arbitrageurs' answer is, "Oh, but we'll all die. At least I'll die rich". The battle rages on.
Most human arbitrage is being replaced with computers and technology. If it's free money, it is best done fast, and you can't beat technology with a human hand going "F6-Buy 100-Enter-Enter". The result is that computers end up trading with each other, which is how the foreign exchange markets work. That's one reason I wouldn't advise many of you to investigate the arbitrage route; at the retail level, our transaction costs are too high as well. But it's useful to understand why there seems to be free money to be made in the markets; sometimes it's an opportunity, and at other times it's not really free.