Almost all economists are in favor of taxing carbon. In survey after survey, they support the idea.

It's easy to see why. The theoretical argument is convincing. Negative externalities -- social costs created, but not considered, in private transactions -- ought to be "internalized" through taxes on the activities in question. Any other outcome is inefficient, meaning that you could make at least one person better off without making anybody (including the polluters) worse off.

There's just one problem. Nobody knows how big carbon's negative externality -- if any -- really is. Richard Tol, an economist at the University of Sussex and an expert on carbon-tax research, has compiled 588 estimates of this number from 75 different studies. In an e-mail he tells me that the interquartile range (which excludes the most extreme estimates) is approximately -$25 to $325 per metric ton of carbon (a negative number means a positive externality).

This huge variation reflects (a) disagreement about the proper social discount rate and (b) what economists call Knightian uncertainty. On the first, the discount rate expresses future costs and benefits in terms of present value. A lower discount rate raises the present value of any future externality: If carbon is damaging, a lower discount rate implies a higher carbon tax. The second, Knightian uncertainty, refers to risks that can't be measured. (The concept comes from Frank Knight, an economist of the Chicago school.) Earth is a highly complex system, and climate models are necessarily poor approximations -- so poor that the degree of uncertainty can't be reliably estimated.

Both create a big problem for the carbon tax. Set it too high, and the economy is harmed. Set it too low, and the economy is harmed.

Nicolaus Tideman and Florenz Plassmann, economics professors at Virginia Tech and SUNY Binghamton respectively, have an ingenious solution. In a paper published in 2010, they say polluters should be made to buy a special 30-year, zero-interest bond from the government for every unit of pollution they emit. The government would set the principal at a reasonable upper-limit estimate of the per-unit cost of pollution. The bond's redemption value, though, would be set in the future. Bondholders would receive what is left of the principal after subtracting the actual cost of the pollution as determined by an independent agency at the bond's maturity.

Investors could trade the securities in a secondary market, creating a prediction market for the cost of emissions. Investors who think the bond market is overestimating future costs will buy the bonds. Those who think the market is underestimating the costs will sell them.

"If you're going to deal efficiently with pollution, it's appropriate to put an incentive on people to economize," Tideman said in an interview. "But you've got to get the price right. Experience tells us that the best way to get information about the future is a futures market. Our proposal adapts a futures market to address climate change."

Tideman's and Plassmann's insight is that policymakers don't know the actual future costs of pollution and don't need to guess. Their market would determine the actual liability of most polluters, except for those who chose to hold the bonds to maturity. This solution uses the ability of prediction markets to synthesize a price from disorganized, disconnected bits of knowledge.

It has four other advantages as well. First, the prices create useful forecast data for environmental policymakers. Second, bondholders remain liable even for surprisingly high costs -- so the costs of a worst-case scenario are privatized, not in effect socialized as with a carbon tax or cap-and-trade program. Third, a bond market accommodates different preferences for risk, allowing those who like it to acquire it from those who don't. Fourth, polluters will make their production decisions based on expectations of actual social cost, rather on what the government thinks today.

The trouble with the carbon tax, for all its strengths, is that we don't know enough about the climate's future to set it correctly. The advantage of Tideman's and Plassmann's proposal is that it addresses that problem directly.

(Evan Soltas is a contributor to the Ticker. Follow him on Twitter.)