SYSTEMS
Future markets analysis
By Barbara Jewett, NCSA -- Participating in the futures market can be a risky business. But a liquidity cost estimator developed by a pair of University of Illinois agricultural economists can help manage the risk. 
What do you do if you have a problem to solve, but it will take you almost a month to get the results on your desktop computer? Would you simply give up and move on to a new problem? Or would you have enough faith in what you were doing to slog slowly along, getting a fraction of the data each day? Slogging slowly along was the path Julieta Frank was on until she discovered NCSA. Frank, a doctoral student in agricultural economics at the University of Illinois, is researching a method to measure liquidity costs in agricultural futures markets. She set her program to run knowing that each day's worth of data she had would take a day to run on the desktop computer in her office. And she had a month's worth of data for two commodities. Then a fellow student told her about NCSA. "This has been so helpful to me," said Frank. "It still takes me a day to run my program, but at the end of the day I have the calculations for a commodity all of the days of the month, not just one." The problem
For her doctoral dissertation, under the direction of professor Philip Garcia, Frank investigates the cost of liquidating a position in the futures market. Price movement is a risk to be managed, she says. An important factor in trading contracts is the cost of not having enough buyers or sellers, which could cause undesired price movements. "If you go to the market and you want to sell or buy futures contracts but you don't have a buyer or seller, then you have to modify your price. But that modification is hard to estimate, it is something you have to extract from the price of the actual transaction. That price modification then becomes a cost of establishing or liquidating your position," Frank explains. "That is the whole challenge of this research, how to estimate that cost." While there has been some study of liquidity costs in financial markets, very little has been done in the agriculture area; that work is older and doesn't reflect the current state of the commodities markets. The challenge in her work, says Frank, comes from the lack of data needed to perform the estimation. "In agricultural commodities pit trading is still common," she says. "Traders yell out the bids and offers and those are not recorded. You only have the final transaction price to infer the bids and the offers and the costs." The process
To overcome this problem, Frank and Garcia built on a Bayesian Markov Chain Monte Carlo (MCMC) algorithm, the Gibbs sampler, implemented by Joel Hasbrouck of New York University in 2004. After comparing their liquidity cost estimator with others that had been used in the past and finding that theirs was more accurate, they proceeded to the next step. The team gathered microstructure data at the Chicago Mercantile Exchange for each day of trading during a month for two commodities, lean hogs and live cattle. The microstructure data basically includes all transaction prices observed during a day, second by second. Then they fed the data into the estimator. Estimating liquidity costs using their algorithm and data takes a considerable amount of time, and even longer when data for many days are used, explains Frank. "I was using the computer in my office. I would start the program and it would run the problem, but it would take all day. And that was to process one day's worth of data for one commodity. If I needed to make a change, I had to start the waiting process all over. Then I learned about NCSA. I called and the people were so helpful. The staff modified our Matlab program so it would run on the NCSA machines and assisted me in getting the information transferred from my office onto the machines." By using the computing and data analysis resources of NCSA, Frank says she was able to include both simulated and actual market data for lean hogs and live cattle in her analysis. She found a consistent pattern of behavior between the real prices and prices generated. She was also able to describe how liquidity costs change over time during the life of the futures contracts. The significance
Estimates of liquidity costs are of interest to both exchanges and market participants. For instance, the team's findings might help an exchange develop strategies for developing new products as well as regulating existing products. For market participants, estimates of liquidity costs in different markets and exchanges are useful in making operational decisions, such as, when to buy or sell. Frank designed the estimator to be used by the various futures market participants. "The results of the research could be used to assist in developing marketing strategies for producers and other market participants," she says. She presented her research at the 2007 NCCC-134 conference and has submitted a paper for publication. She is ramping up for the final phase of her project, which introduces volume into the variables and then attempts to measure how prices change when a trader purchases or sells a large quantity of futures contracts. And she will once again turn to NCSA for assistance in managing her large data sets. This work is partially funded by the Office for Futures and Options Research in the Department of Agricultural & Consumer Economics at the University of Illinois at Urbana-Champaign. Team members:
Julieta Frank
Philip Garcia More information:
http://www.farmdoc.uiuc.edu/nccc134/ http://www.ace.uiuc.edu/ofor/ http://www.ace.uiuc.edu/
What do you do if you have a problem to solve, but it will take you almost a month to get the results on your desktop computer? Would you simply give up and move on to a new problem? Or would you have enough faith in what you were doing to slog slowly along, getting a fraction of the data each day? Slogging slowly along was the path Julieta Frank was on until she discovered NCSA. Frank, a doctoral student in agricultural economics at the University of Illinois, is researching a method to measure liquidity costs in agricultural futures markets. She set her program to run knowing that each day's worth of data she had would take a day to run on the desktop computer in her office. And she had a month's worth of data for two commodities. Then a fellow student told her about NCSA. "This has been so helpful to me," said Frank. "It still takes me a day to run my program, but at the end of the day I have the calculations for a commodity all of the days of the month, not just one." The problem
For her doctoral dissertation, under the direction of professor Philip Garcia, Frank investigates the cost of liquidating a position in the futures market. Price movement is a risk to be managed, she says. An important factor in trading contracts is the cost of not having enough buyers or sellers, which could cause undesired price movements. "If you go to the market and you want to sell or buy futures contracts but you don't have a buyer or seller, then you have to modify your price. But that modification is hard to estimate, it is something you have to extract from the price of the actual transaction. That price modification then becomes a cost of establishing or liquidating your position," Frank explains. "That is the whole challenge of this research, how to estimate that cost." While there has been some study of liquidity costs in financial markets, very little has been done in the agriculture area; that work is older and doesn't reflect the current state of the commodities markets. The challenge in her work, says Frank, comes from the lack of data needed to perform the estimation. "In agricultural commodities pit trading is still common," she says. "Traders yell out the bids and offers and those are not recorded. You only have the final transaction price to infer the bids and the offers and the costs." The process
To overcome this problem, Frank and Garcia built on a Bayesian Markov Chain Monte Carlo (MCMC) algorithm, the Gibbs sampler, implemented by Joel Hasbrouck of New York University in 2004. After comparing their liquidity cost estimator with others that had been used in the past and finding that theirs was more accurate, they proceeded to the next step. The team gathered microstructure data at the Chicago Mercantile Exchange for each day of trading during a month for two commodities, lean hogs and live cattle. The microstructure data basically includes all transaction prices observed during a day, second by second. Then they fed the data into the estimator. Estimating liquidity costs using their algorithm and data takes a considerable amount of time, and even longer when data for many days are used, explains Frank. "I was using the computer in my office. I would start the program and it would run the problem, but it would take all day. And that was to process one day's worth of data for one commodity. If I needed to make a change, I had to start the waiting process all over. Then I learned about NCSA. I called and the people were so helpful. The staff modified our Matlab program so it would run on the NCSA machines and assisted me in getting the information transferred from my office onto the machines." By using the computing and data analysis resources of NCSA, Frank says she was able to include both simulated and actual market data for lean hogs and live cattle in her analysis. She found a consistent pattern of behavior between the real prices and prices generated. She was also able to describe how liquidity costs change over time during the life of the futures contracts. The significance
Estimates of liquidity costs are of interest to both exchanges and market participants. For instance, the team's findings might help an exchange develop strategies for developing new products as well as regulating existing products. For market participants, estimates of liquidity costs in different markets and exchanges are useful in making operational decisions, such as, when to buy or sell. Frank designed the estimator to be used by the various futures market participants. "The results of the research could be used to assist in developing marketing strategies for producers and other market participants," she says. She presented her research at the 2007 NCCC-134 conference and has submitted a paper for publication. She is ramping up for the final phase of her project, which introduces volume into the variables and then attempts to measure how prices change when a trader purchases or sells a large quantity of futures contracts. And she will once again turn to NCSA for assistance in managing her large data sets. This work is partially funded by the Office for Futures and Options Research in the Department of Agricultural & Consumer Economics at the University of Illinois at Urbana-Champaign. Team members:
Julieta Frank
Philip Garcia More information:
http://www.farmdoc.uiuc.edu/nccc134/ http://www.ace.uiuc.edu/ofor/ http://www.ace.uiuc.edu/