In a general sense this analysis has been concerned with the extent of a market and the effect of limiting the extent on the prices of assets in that market. One example of the type of limited market extent with which we have been concerned is provided by nonmarketable assets and another is provided by market segmentation.
Unambiguous statements of the effect of nonmarketable assets and market segmentation on the level of prices require that and be oppossite in sign (unless one or both are zero). Only two cases have been identified where unambiguous statements can be made. The first, the case of constant absolute risk aversion, implies there is no effect of nonmarketable assets or market segmentation on the level of asset prices. The second case, constant relative risk aversion, where the coefficient of relative risk aversion is equal to or less than one, implies that prices are lower in the presence of these imperfections.
Arrow  argues that the coefficient of relative risk aversion must “hover around 1.” Thus, if constant relative risk aversion is a reasonable approximation to reality we should accept the implication of the latter case. Certainly, if a choice had to be made, the latter case would be the more palatable of the two. That is, constant relative risk aversion does imply decreasing absolute risk aversion, which appears more acceptable than the hypothesis of constant absolute risk aversion.
The constant relative risk aversion case has implications for such things as the organization and operation of markets and corporate merger decisions. For example, higher margin requirements that inhibit diversification would be expected to lower asset values. Also, as a matter of corporate policy, it would appear that, ceteris paribus, mergers that increase the extent of a market would be preferable to within-market mergers.