Hi, I'm a programmer who writes a lot of financial model code (a financial programmer?) for my company's quants. I understand the vocabulary, and I can handle the math (BS in Electrical Engineering . Recently I have become interested in modeling the difference between puts and calls in commodity markets. From searching this site I read that commodity markets tend to be demand driven, and there is a skew between puts and calls. I noticed this before signing up here, in most commodity markets the vol for itm puts is higher than the call at that strike and vice versa. Also there is the effect of the price floor for otm options - they can't settle below 1 tick, which results in higher than expected implied vols. Plotting the difference between the p/c vol's shows a clearly defined (and almost always the same) odd-ordered graph (like x^3, x^5, or whatever). Has anyone created a usable model for this behavior? In my mind the difference should be an exponential function, something proportional to exp(abs(K-U)) or maybe exp(K/U) for the right hand side and exp(U/K) for the left. I've tried several regressions but their results are unsatisfactory. The best I've come up with is B * sqrt(T) * exp(K/U) or U/K for the opposite side. B is some number that seems to be constant per commodity. Any success with this, or other ideas for me to try?

QuoteOriginally posted by: IgnignotAlso there is the effect of the price floor for otm options - they can't settle below 1 tick, which results in higher than expected implied vols. The bid/ask spread will always give you an implied vol spread. How you pick a single iv from within that spread is essentially arbitrary in the absence of other information that might help. When you are so far otm that the bid price is zero, then the iv is anywhere between zero and the ask iv.Be skeptical of any notion of a single market iv (e.g. the bid/ask mean) that ignores this spread. Edit:A similar situation applies deep itm where bid prices may even be less than intrinsic value. So again use zero for the minimum iv.In any tail situation like this, applying some sort of smoothness condition as you go from strike to strike may give the extra info you need.

I am currently producing smooth implied volatilities with a number of functions. The tail sigma differences between put and call are easy to predict, because deep itm options tend to have 0 extrinsic value, while deep otm options tend to have 1 tick extrinsic value. However there is a transition area between the center of the vol smiles where there is little or no difference between the two and the tails. In this transition there is a clear movement away from p/c parity while still in a region where the discrete price movements should not be an important factor. I'm familiar with the effects of converting between a continuous (where the price should be) and a digital (price can only be on ticks) system, and this divergence from parity doesn't appear to be caused by rounding. Basically my goal here is to produce some numbers to answer the question "if the market were trading contracts at these strikes, what would they say the price or vol was?" So producing prices in between ticks really isn't a solution. When there are other contracts to build a model with (say I need a certain put's iv while I already have 10 put iv's) then I already have tools to produce something. But in some markets only one type of contract will be popular, like in Chicago Wheat May 05. Until today there were no puts traded. It is hard to build a model for the puts without any put data, so instead I want to take the call data (which is very deep) and put it through some sort of put/call parity model. I have a general sense of what the data should look like, but I don't have a deeper understanding of why and I believe that is necessary for me to produce a good model here.

QuoteOriginally posted by: Ignignot When there are other contracts to build a model with (say I need a certain put's iv while I already have 10 put iv's) then I already have tools to produce something. But in some markets only one type of contract will be popular, like in Chicago Wheat May 05. Until today there were no puts traded. When you say "popular" what do you mean? How many contracts per day? What might a typical ATM bid/ask spread be?Matthew

By popular I mean that the call contracts are relatively popular in comparison to the put contracts. The put contracts currently have 1 open interest in a single contract, whereas the call contracts have around 1000 open interest spread across 20 contracts. Again this example is Chicago wheat in May 05, but it happens in other months and commodities as well. On average, the volume is very light for calls (about 10 contracts a day) and it will remain that way until Dec gets closer to expiry. The puts have no volume.

QuoteOriginally posted by: IgnignotBy popular I mean that the call contracts are relatively popular in comparison to the put contracts. The put contracts currently have 1 open interest in a single contract, whereas the call contracts have around 1000 open interest spread across 20 contracts. Again this example is Chicago wheat in May 05, but it happens in other months and commodities as well. On average, the volume is very light for calls (about 10 contracts a day) and it will remain that way until Dec gets closer to expiry. The puts have no volume.When we get closer to expiraton, what might the daily volume be? How many market makers are there? This seems to be a pretty thin market.Matthew

QuoteWhen we get closer to expiraton, what might the daily volume be? How many market makers are there? This seems to be a pretty thin market.It is a pretty thin market, that's why I'm having trouble fitting the puts to a model, because until recently there was absolutely no put data. In 2 months I'd expect the volume to go to about 400 contracts a day, and 2 months after that I'd expect it to be around 1000 contracts a day.

QuoteOriginally posted by: IgnignotQuoteWhen we get closer to expiraton, what might the daily volume be? How many market makers are there? This seems to be a pretty thin market.It is a pretty thin market, that's why I'm having trouble fitting the puts to a model, because until recently there was absolutely no put data. In 2 months I'd expect the volume to go to about 400 contracts a day, and 2 months after that I'd expect it to be around 1000 contracts a day.I understood you to say previously that current call volume is 10 contracts per day and that current open interest is 1000 contracts. And you are projecting volume in 2 months to be 1000 contracts per day?Matthew

Oh my God! You're interested in May Wheat! Focus on the March and July instead. The market will trade out of the March and "skip over" the May and build up volume & open interest in the July. If you're really interested in grain options, focus on Corn and Soybeans instead. Don't waste your time with illiquid contracts.

QuoteOriginally posted by: nazzdackOh my God! You're interested in May Wheat! Focus on the March and July instead. The market will trade out of the March and "skip over" the May and build up volume & open interest in the July. If you're really interested in grain options, focus on Corn and Soybeans instead. Don't waste your time with illiquid contracts.As I said further down, I'm mainly a programmer, not a trader. I don't get to decide which markets to be interested in - I just want to be able to handle any of them. It is not uncommon for my company to be the first to really trade some markets. I wasn't sure if May was one of the heavily traded contracts, so my guess of 1000 contracts a day is obviously high. But in 4 months I'd expect at least 100 a day. It should still be traded more than serial months. The important thing here is that this month has a unique property - that there is a good amount of data on call options, but almost none on the put options. Since I want to be able to guess what a put would cost, I want to use the call data to help flesh out a put model. Right now I use a stupid hack of a model - basically p/c parity plus some general deviation based on time to expiry and the distance between the strike and atm. I got it through some trial and error with a linear regresssion, but I think that with a better understanding of why there is a systematic difference between puts and call volatilities I can come up with a useful model. In other words, right now I have a little bit of the what and I want some of the why so I can figure out the rest.

Sorry for the very late reply.Put call parity applies at forward.Excluding the vol bid/offer spread, some differences in implied vol will be noticeable because of the market makers on the floor "manipulating" prices for transformations (a call vs a put of same strike should equal intrinsic but these guys are making markets on these...), but this is a reamain from the old CBOT past when you HAD to physically exercise your ITM options with your clearer at expiry, thus creating a need for strike management. The spd is not justified anymore.If you dont have the put price, just use calls, it works.

put-call parity should hold for options on futures as there is no problem shorting futures as there might be for equities. are you using quoted markets or are you using settlements? are you using the corresponding future (e.g., May wheat for may options)? i figure you have gone through this, but i am just trying to find out where the problem might be.

another thing to check if you are seeing consistent call-put iv differential for the same strike -- interest rates. if that is not your problem, let me know