Onions and Oil: Flawed Analysis

May 20th, 2011 | Categories: Uncategorized | Tags:

I came across a blog post (link, and also a column: link) that talks about the effects of futures market on the oil and onion market. Many think that futures market is responsible for the oil price fluctuation (link). The author notes that despite the absence of futures market in the onion markets (due to Onion Future Act (link)), the price fluctuations are more severe than that is seen in oil trading. The authors uses this example to argue some academics’ belief that futures markets control extreme price swings. I think the analysis is poor for the reasons I describe below. I am not an economist; but, being trained as an engineer, I can say that their analysis is flawed.

First, the statistical analysis of price fluctuations in the posts is poor. The author of the blog post compares the monthly changes in the price of onion to that of oil. We have seen that oil prices have a two- or three-year cycle. Obviously, the monthly changes in oil prices is not as big as the changes in onion prices. A better statistical analysis is needed. Also, volatility of onion prices before and after the Onion Futures Act was passed is open to debate (link).

Second,  it is not a controlled statistical sampling of commodity value because there are a number of factors that affect each the price of commodity. The oil market has a seasonal supply-and-demand cycle, whereas the onion market experiences abrupt supply issues due to inclimate weather. There are plenty of other factors, including the price of oil, that determines the price of onions. Oil prices are globally uniform while onion price vary geographically. Future trading is not the only factor that differs between the two markets.

Third, if the cycle is two to three years long, ten-year data of oil prices fluctuation is not enough to statistically conclude anything about futures trading. We either need a different form of analysis for the data from the limited time period or need data from a larger time period. But it is hard to control all the factors when the time period is large. Thus, statistical analysis would likely be meaningless.

The authors may be right about their conclusion, but they have not provided sound evidence to back it up.

No comments yet.