Price discovery in the Indian castorseed market

Transcription

1 Indira Gandhi Institute for Development Research 1/13 Price discovery in the Indian castorseed market Susan Thomas and Kiran Karande susant@igidr.ac.in

2 Questions we are addressing How are prices on the futures and spot markets related? How do the Ahmedabad and Bombay markets affect each other? 2/13

3 Methodology for thinking about spot/futures 3/13 prices We watch fluctuations of daily spot and futures prices. Basis = F t S t. We watch for changes in the basis. Question: Does a jump in the basis today forecast an increase in the futures price tomorrow? Question: Does a jump in the basis today forecast an increase in the spot price tomorrow? We do empirical analysis of this basic idea using two methods: (a) the Garbade-Silber model and (b) Granger causality.

4 Methodology for thinking about Bombay/Ahmedabad prices Spread = F b F a. We watch for changes in the spread. Question: Does a jump in the spread today forecast an increase in the Bombay price tomorrow? Question: Does a jump in the spread today forecast an increase in the Ahmedabad price tomorrow? We do empirical analysis of this basic idea using two methods: (a) the Garbade-Silber model and (b) Granger causality. 4/13

5 Data Daily data only. Data from 1985 to Bombay: BOOE prices. All data from FMC. 5/13

6 Our findings 1. We find that futures prices lead spot prices for three of the four contracts. In the case of the March contract, there is no clear flow from futures to spot. In the Bombay markets, futures consistently react before spot price. In contrast, Ahmedabad spot prices seems to react as rapidly as futures prices; most of the information flow seems bidirectional. 6/13

7 2. Geographical location of markets appear to matter in the price discovery process. Bombay futures prices lead the Ahmedabad futures prices for three out of four futures contracts. Ahmedabad futures prices lead Bombay futures prices in March. 7/13

8 Results for Ahmedabad and Bombay futures prices 8/13 Coefficient March June September December Pooled β fa (-0.54) (2.39) (1.97) (2.70) (8.78) R D-W β fm (2.82) (-0.63) (0.42) (0.68) (3.97) R D-W ρ s,f (t statistics in parenthesis)

9 A graph of the market and the contracts 9/13 C4: Dec Sep C1: Mar Dec C2: Jun Mar C3: Sep Jun C4: Dec Sep C1: Mar Dec Harvest Harvest

10 A graph of the results A: S <=> F A: S <=> F B: S > F B: F > S A > B A: S > F B: F > S B > A A: F > S B: F > S A = Ahmedabad B = Bombay F = Futures S = Spot 10/13 B > A B > A C4: Dec Sep C1: Mar Dec C2: Jun Mar C3: Sep Jun C4: Dec Sep C1: Mar Dec Harvest Harvest

11 Questions that arise from these findings 1. Do Bombay futures prices lead Ahmedabad futures prices because news and information about the demand for castor is revealed in Bombay first? 2. Do Bombay futures prices lead Ahmedabad futures prices because capital is easier to source in Bombay? 3. Why do the prices in Bombay lead the prices in Ahmedabad when the volumes in Ahmedabad are so much larger? Has this always been the case, or is it a recent phenomenon? 11/13

12 4. For the March contract, does the Ahmedabad futures prices lead Bombay prices because production of castor is in Ahmedabad, and information about the harvest hits the Ahmedabad markets first? 5. Is this also the reason that the spot market prices lead futures markets prices in both Ahmedabad and Bombay for the March contract? 6. If information is being revealed in one market before the other, why is no-one arbitraging between the two markets? Either Ahmedabad and Bombay, or futures and spot? 7. Are high costs of transactions an issue in doing arbitrage between futures and spot markets? Where are the costs higher? 8. Have these costs been increasing or decreasing over time? 9. Have easier information access reduced these apparent arbitrage opportunities over the last decade? 10. Is counterparty risk an issue in doing arbitrage trades? 12/13

13 What should we do next? These models can be used for computing initial margin on the futures exchange, using Value at Risk (VaR). 13/13