I investigate whether, as is commonly believed -- and if so how -- firm disclosure of so-called "toxic" releases, required since 1987 by the federal "Toxics Release Inventory ("TRI"), has brought about the reductions in toxic releases that have occurred since that time. Existing literature, consisting principally of event studies of stock market returns, suggest that dirty firms experience abnormal negative returns. Using a micro-level data set that links TRI releases to plant level Census data for petroleum refineries, I study plant-level behavior, exploiting state variation in toxics regulations, and exploring the relationship between TRI releases and concomitant regulation of non-toxic pollutants. I find that, although TRI induced public disclosure may have contributed to the decline in reported toxic releases, that alone has not been the cause of those reductions: the evidence is strong that changes in toxic emission intensity are a byproduct of more traditional command and control regulation of emissions of non-toxic pollutants. I find that (1) since 1987, refineries have become substantially cleaner in terms of over-all toxic releases; (2) the clean-up has not occurred through substitution away from TRI listed substances as inputs or alteration in the mix of outputs; and (3) refineries in states with more stringent supplemental regulation of toxics (e.g. with specific state-wide goals for toxic reductions) have significantly lower toxic emission intensity levels than refineries in other states. I find also that (4) TRI air releases are highly correlated with levels of criteria air pollution; (5) both toxic pollution levels and intensity fall with increases in pollution abatement (operating and maintenance) expenditures for non-toxic air pollution; and (6) TRI air releases are affected by being in more stringent regulatory regions for the criteria air pollutants. Finally, I link my data-set with CRSP data to re-evaluate the effect of TRI reporting on company stock market valuation, correcting for a methodological shortcoming (stemming from the fact that all reporting firms face a common event window) of prior event studies of the impact of the TRI. Correcting for that shortcoming, I find that (7) the evidence of negative abnormal returns around TRI reporting dates for petroleum companies is not significant. My findings suggest that the most probable mechanism through which TRI reporting may induce firms to clean up is local and state governmental use of TRI disclosures. They suggest also not only that the perceived effectiveness of TRI regulation has been overstated, but perhaps more importantly that the benefits of command and control regulation of non-toxic pollutants have been underestimated.