Expenditures for environmental protection in the U.S. are estimated to exceed $150 billion annually or about 2% of GDP. This estimate, based on largely self-reported information, is often cited as an assessment of the burden of current regulatory efforts and a standard against which the associated benefits are measured. Little is known, however, about how well reported expenditures relate to true costs. The potential for both incidental savings and uncounted burdens means that actual costs could be either higher or lower than reported expenditures. A significant literature supports the notion that increases in reported environmental expenditures probably understate actual economic costs. Estimates of the true cost of a dollar increase in reported environmental spending range from $1.50 to $12. This paper explores the relationship between reported expenditures and economic cost in the manufacturing sector in the context of a large plant-level data set at the four-digit SIC level. We use a cost function modeling approach which treats both environmental and non-environmental production activities as distinct, unrelated cost minimization problems for each plant. We then explore the possibility that these activities are, in fact, related by including reported regulatory expenditures in the cost function for non-environmental output. Under the null hypothesis that reported regulatory expenditures accurately measure the cost of regulation, the coefficient on this term should be zero. In ten of eleven industries studied, including all of the heavily regulated industries, this null hypothesis is accepted using our preferred fixed-effects model. Our best estimate, based on an expenditure weighted average of the four most heavily regulated industries, indicates that an incremental dollar of reported environmental expenditure reduces non-environmental production costs by eighteen cents with a standard error of forty-two cents. This is equivalent to saying that total costs rise by eighty-two cents for every dollar increase in reported environmental expenditures. Using an alternative pooled model we find uniformly higher estimates. Although consistent with previous results, we believe these higher estimates are biased by omitted variables characterizing differences among plants. Summarizing, our results enable us to reject claims that environmental spending imposes large hidden costs on manufacturing plants. In fact, our best estimate indicates a modest though statistically insignificant overstatement of regulatory costs.