In a post someone just asked about the assumptions that must be satisfied in the classical linear regression model so that the parameter estimates are unbiased/consistent/efficient/etc. My question here is similar: what assumptions for the market for some asset must be satisfied so that there is no arbitrage? What are the implications of violations of these different assumptions? Is there a standard set of assumptions or do different authors use slightly different formulations? (I know in the classical linear regression model all authors seem to work with slightly different sets of assumptions).Here's the fun part: can you relate this to "Stamp Arbitrage"? That is, do you think there's really no arbitrage in the market for stamps? If not, what assumptions do you think are violated?Hopefully a relation to the stamp market will help a number of people (including me ) who are wrestling with the intuition behind the assumptions need for no arbitrage.

Hey,I've just worked this out, but I think it makes sense...For arbitrage to exist one must be able to buy in one market and sell in another for a higher price risklessly. That is a fair enough assumption on its own. I dont think arbitrage has the same set of technical assumptions as econometric techniques such as OLS and Gauss Markov.So lets see what assumptions we need for the stamp arbitrage example to work Firstly it did not say that we knew when stamp prices would go up, so when we bought our stamps we were exposed to the risk of no price increase for a number of months possibly.Assume we are in the real world and we knew the price was going to go up; we buy the forever stamps and the price for stamps go up 5% the day after. For this to be arbitrage our costs must be less than 5%. The main costs I can think of are1) Our time. Standing outside a post office or selling on ebay takes alot of time2) Transaction costs, we may need to buy alot of stamps and this will cost money in storage and logistics and depending on where we sell them we may incur rents etc (ebay)3) Opportunity costs, we give up the ability to use our money elsewhere, eg savings account at 2% (If your lucky)So weighing up the costs of performing this transaction, this cannot be seen as an arbitrage because you are not guaranteed a riskless profit due to the inherent costs involved which people wisely see will outweigh the benefits involved.However if we assume no transaction costs and we have perfect information of when the prices will increase arbitrage can be assumed to exist, as long as the riskless rate of savings are not greater than 5% This is not the technical answer I think you were looking for but it gives a good intuition of how I see arbitrage.There is a nice technical proof of arbitrage existence in yield curve shifts when we assume parrell shifts in the yield curve using duration and covexity which might be useful if you want a technical outlook (Martellini et all,2008, Fixed Income Securities)Paul