When environmental damages from emissions are spatially nonuniform, pe
rmit trading has been modeled most often as a ''pollution offset progr
am'' in which emission permits are traded between agents, subject to c
onstraints on ambient air quality. To date the institution envisioned
to implement such a program involves trading on a bilateral and sequen
tial basis. However, simulation studies indicate that the sequence of
trades may alter the outcome and undermine the cost savings from a pol
lution offset program. This paper identifies a design for the trading
institution that tends to overcome this phenomenon and improve the eff
iciency of equilibria obtained in a simulation model. We model a bilat
eral trading process for the reduction of sulfur dioxide emissions wit
h a stochastic description of the sequence of trades within groups of
nations in Europe. When trading takes place between disaggregated, sty
listic representations of economic enterprises, rather than between na
tional governments, a significantly greater portion of potential savin
gs is achieved. In fact, under most sets of assumptions, approximate f
irst order stochastic dominance is achieved wherein the more decentral
ized the trading agents, the greater the expected savings from a tradi
ng program.