Shane Mage wrote: > What meaning can "the true probability distribution" have here? > Probability distribution (objective) can apply to the outcome of a > series of "random" events like throws of dice or spins of a roulette > wheel, or to (subjective) *an* estimate of the "likelihood" of > various (mutually incompatible) future possibilities.
The way I look at it, ultimately, all financial assets are contingent claims on physical productive assets: means of production (MP) and labor power (LP). This is a plain accounting fact. The price of MP and LP (in social settings where they're commodities) have objective centers of gravity: "values." That's my view at least. Some respectable PEN-L members believe that valuing MP and LP is inherently impossible or self-contradictory. Not me. I've looked into the argument a few times and remain unconvinced. I'm with Marx on this. Clearly, these values are the social aggregate (average) of private expectations about the amounts of social labor required to reproduce the MP and LP in "normal" conditions -- expectations conditional on existing information available to whoever is doing the valuation. Ultimately, as Marx also believed, the valuation of MP and LP is social, that is, the computation or aggregation of expectations is conducted by trial and error in/through market exchange. Those values are what I call the "fundamentals." They're social objects. They're objective: independent of the subjectivity of individuals. That said, we need to keep in mind that the objectivity of values is social. When people do things (e.g. produce in given social settings), their actions crystallize into new or transformed social objects (not necessarily tangible, but material or physical in the most general sense of these terms): products, social structures and their emerging properties, institutions, broader social conditions, culture, history. The degree of social objectivity of a given social object depends on its particular nature. Ultimately, these social objects are all historically contingent. Some are contingent upon a change in people's minds or habits or customs. Others are contingent on a change in laws or political conditions. Others are contingent on changes in more hardened economic structures. Etc. Keeping track of this, probability theory and statistical inference can be used productively. > But (first > case) "variables" in the social sciences are essentially unique > (100% probable), like the present value of an asset as determined > ex post. I really don't understand this. What is the PV of an asset "as determined ex post"? Are you talking about an actual asset price? Are you talking about the results of backtesting some asset pricing model? In those cases, indeed, you can think of those prices (actual or predicted) as unique realizations of the true random price under some probability distribution. But neither of these is a PV. PV is your best estimate of the price of the asset as of now. The one that dictates your next move (or lack thereof) regarding the asset. And that's an expectation, conditional on the information available to you (e.g. history of prices of the asset, etc.). > And (second case) nobody ever even tries to estimate a > true probability distribution comprising all possible values of > such a "variable." Well, it depends on your practical needs and resources. I don't think the "all possible values" is a big deal. There's nothing that stops you from assuming that the price of an asset can vary along the entire set of real numbers. I don't see why that'd be more costly than restricting the variation to a narrower range. The costly thing is the estimation of the parameters of interest of your distribution. There are trade-offs involved here, but simply put, how good is your data? The better your data, the more costly. Which parameters do you need to estimate? The higher the moments (center, dispersion, skewness, kurtosis, etc.), the more costly. Which assumptions about probabilistic behavior are you willing to make? The weaker your assumptions, the more costly. Etc. It boils down to a cost-benefit calculation. My impression is that, usually, most people in finance play the game at a level that only requires estimating the first 2-4 moments (center, dispersion, skewness, kurtosis). And, given their goals, they are willing to make strong assumptions. Still, that's a lot of compressed information about the true (unknown) behavior of the rv. More than most people use in their practice. > So how can the people who make up the > "markets" get it all wrong? The reason is that "all" is very long term, and > their only interest is very short term because that is where the loot is. > And in *every* short term they (collectively) make out like the > bandits they are. The huge costs are borne by other people, including > classsical shareholders, not only workers and their pension funds. Again, I don't think the issue is "all." The rest, I think, is well said. For some cases at least.
