In futures there exist exchange traded calendar spread contracts, which trade as a single unit (think May/June Crude Oil). The bid ask spread for the spread contracts is the same as that of the outrights, which typically makes trading the spread contract (if you would like to trade a spread) cheaper than trading the outright contracts individually.
Is there a way to take this into account in a portfolio optimization problem. For example, perhaps without transaction costs my mean variance optimization says to buy 55 June (M) crude oil contracts and Sell 45 July (N) crude oil contracts. This amounts to buying 55 spread (M/N) contracts and buying 10 July (N) contracts. However, because trading the M/N spread contracts are 50% cheaper than trading the outright contracts, the optimal solution knowing the transaction cost difference may be just to buy 50 M/N spread contracts.
Is there a way to formalize this? I know the case in which you have individual assets people typically use the 3/2, volume based transaction model by Chriss. But in cases in which these spread contracts exist is there a formal way of modelling the transaction costs?