David-Evans

David Evans

 

[Image of a split circle appears with photos in each half of the circle flashing through of various CSIRO activities and the circle then morphs into the CSIRO logo]

 

[Image changes to show a new slide showing David Evans inset talking in the top right and text appears: Designing carbon markets fit for the Australian land sector, Andrew Reeson, Todd Sanderson and David Evans]

 

David Evans: Thanks. Hi everyone. I’m David Evans and I work in the Decision Sciences Arm of Data61 and I’m just going to talk about three pieces of research that Andy Reeson, Todd Sanderson and I have done on designing carbon markets for the Australian land sector.

 

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: The Australian carbon market, Facilitates generation and exchange of carbon credits, 1 carbon credit = 1 tonne of CO2 equivalent emissions stored or avoided, Supply side, Government issues carbon credits to landholders for emissions reduction activities, Demand side, Australian government buys carbon credits through Emissions Reduction Fund auctions, Businesses by carbon credits via over-the-counter markets]

 

So I’ll just start with a brief overview of the Australian carbon market. So this market facilitates the generation and exchange of carbon credits, where each carbon credit represents one tonne of CO₂ equivalent emissions stored or avoided. On the supply side of the market, the government issues carbon credits to land holders for undertaking emissions-reduction activities. These include things like reforestation to sequester carbon. And on the demand side of the market the federal government is the major buyer of carbon credits and makes these purchases through its Emissions Reduction Fund auctions to meet its environmental targets. Some businesses also buy carbon credits via an over-the-counter market to offset their emissions and this seems to have been increasing a bit, recently.

 

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: Design of procurement auctions, Auctions to pay landholders to change their land use for environmental benefit, e.g. ERF auctions, Rationale is that competitive bidding, Decreases the price the government pays, Reveals bidders’ true opportunity costs, We explore the performance of repeated auctions under different designs, Discriminatory price, Uniform (second) price]

 

So just on to our first piece of research, this sort of considers how to design efficient procurement auctions for ecosystem services. So governments frequently use these auctions to pay landholders to change their land use to deliver some type of environmental benefit. These auctions involve landholders just submitting bids where they offer to undertake projects at certain prices and the government just selecting the best bids. The most notable recent example of this is the Federal Government’s Emissions Reduction Fund auctions, through which the government has paid landholders a bit over $2 billion to undertake projects that have delivered about 200 million tonnes of CO₂ equivalent emissions reductions.

 

Our study explored the performance of different auction designs for repeated or multiple round procurement auctions. These auctions are often used by governments so the repeated auctions I mean. They have tended to have been under studied with most of the research focussing on single-round auctions. So in our study, we compared the two main formats used in procurement auctions. The first format is the widely used discriminatory price auction where each successful bidder is paid its bid price. The problem with this auction design is that it gives bidders an incentive to add mark-ups to their bids to extract greater profits. And the second format we looked at is the uniform, second price auction, where each successful bidder is paid the lowest price of the unsuccessful bids. So the benefit of this is that it removes the incentive, bidder’s incentive, to add mark-ups to their bids but the problem is that it may require the government to, it requires the government to pay all bidders the potentially high uniform price.

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: Design of procurement auctions, Agent-based model of the auction system, Agents adapt in the Discriminatory format, Agents bid truthfully in the uniform format, Compared formats on two measures, Cost-effectiveness, Efficiency]

 

So we developed an agent-based model to simulate a repeated auction under both of these formats and we used the results from auction experiments to inform the agent’s behaviour in each format. So in the discriminatory auction we had bidders incrementally adjust their prices in the direction of greater profits over time as observed in experiments, and in the uniform auction we just let the bidders bid their true opportunity costs in each round which is consistent with what’s been observed in experiments. We’ve only compared the formats on two measures. One was cost-effectiveness or just the dollars spent per unit purchased and the other was efficiency or the opportunity cost per unit purchased.

 

[Image changes to show a new slide showing two graphs showing the Discriminatory and Uniform formats measured in price paid per unit and opportunity cost per unit and David can be seen inset talking in the top right and text appears on the slide: Design of procurement auctions, Discriminatory format more cost-effective in early rounds, then less cost-effective due to bidders’ learning, Discriminatory format reasonably efficient despite agents’ bid inflation]

 

So the next few plots just show the results of our simulations across 20 auction rounds. The left panel shows the average price paid per unit and the right panel shows the opportunity cost per unit. In both cases, the lower the better, and since agents’ behaviour is partly random, in our model we’ve used repeated simulations for the comparison. So, our first experiment , or sorry our first result shown in the left panel here, is that the discriminatory format in red is more cost-effective in the early auction rounds than the uniform format but quickly deteriorates as bidders learn to adjust their prices to extract greater profits over time. And our second result shown in the right panel is that the discriminatory auction maintains reasonably good efficiency despite agents inflating their bids above their true costs but with the auction still being able to select most of the low cost bids.

 

[Image changes to show a new slide showing two new graphs and David can be seen inset talking in the top right and text appears on the slide: Design of procurement auctions, Greater competition widens the performance gap between the two]

 

So, our third result here is that increasing the level of competition by adding more bidders to the auction widens the performance gap between the two formats on both measures. So it tips things further in favour of the uniform auction.

 

[Image changes to show a new slide showing two new line graphs plotting the Risk-neutral, Risk-averse, Risk-seeking, and Uniform lines and David can be seen inset talking in the top right and text appears on the slide: Design of procurement auctions, Relative performance is highly sensitive to collective risk appetites]

 

The final result is that the relative performance of the format is highly sensitive to small changes in bidders’ collective risk appetites so in particular if agents have mild collective risk-aversity the discriminatory format which is shown here in green achieves similar cost-effectiveness and efficiency to the uniform format which is shown in purple. So these results show that the widely-used discriminatory format can perform quite well relative to the uniform format but only within certain conditions. In particular, having a small number of auction rounds limits bidders’ ability to, sort of, learn and makes the discriminatory format relatively cost-effective and also making bidders uncertain about the timing and budget of each auction round may make them risk-averse and improve the discriminatory format’s performance.

 

[Image changes to show a new slide showing an Australian map and David can be seen inset talking in the top right and text appears on the slide: Temporal crediting for carbon sequestration, Biosequestration can make large contribution to Australia’s GHG mitigation, Forestry could sequester tens of millions of tonnes of carbon per year, Potential remains untapped, Limited uptake of reforestation and carbon sequestration projects]

 

OK, so just now moving on to a second study that we’ve done which proposes a new approach to incentivising carbon sequestration through temporal crediting. So biosequestration can make a significant contribution to Australia’s greenhouse gas mitigation efforts. Forestry in particular has the potential to sequester tens of millions of tonnes of carbon per year over several decades and this figure on the right gives an indication of the amount of land that could be used for carbon forestry and the associated sequestration rates. But the problem so far is that most of this potential remains untapped with very limited uptake of reforestation projects in the Government’s Emissions Reduction Fund.

 

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: Temporal crediting for carbon sequestration, Limited uptake due to strong ‘permanence’ requirements, Achieving permanence required landholders to forego the (valuable) option to change land management practices, Temporal crediting system, Recognise and reward the value of temporarily sequestered carbon, Allow landholders to maintain management flexibility, reducing the costs of participation]

 

So we argue that this limited uptake may be due to the Government’s strong permanence requirements which specify that sequestration projects must maintain stored carbon for a period of 25 or 100 years and this means that there is zero-value placed on carbon stored for shorter periods, even though every year the carbon’s not in the atmosphere has environmental value or may have environmental value.  So these strong permanence requirements greatly increase the opportunity costs to land holders undertaking carbon sequestration projects and this is because achieving permanence requires landholders to forego the option to change their land management practices such as switching to different crops in response to market conditions and studies have shown that landholders place substantial value on this sort of management flexibility.

 

So, to increase the uptake of sequestration projects we’ve proposed a temporal crediting system that involves renting carbon sequestration from landholders while it is ongoing. This basically allows landholders to retain the option to switch management practices and it would probably increase the uptake of the method and deliver some environmental benefits.

 

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: Recognising multiple carbon values, Current carbon market does not recognise multiple carbon values, Different carbon credits may have different values due to, Co-benefits, Risks, Challenge is to have mechanism that recognises multiple carbon values without fragmenting the market, We ran a market experiment to test different mechanisms]

 

So, just now moving on to our final study, which is on a market mechanism that can recognise multiple values for carbon. So in Australia, a carbon credit just represents one tonne of carbon emissions avoided or stored. So a limitation of the current market is that it doesn’t really recognise that different carbon credits can have different values. These different values may arise in a couple of ways. One is due to co-benefits such as where a carbon credit generated via a tree planting project also generates bio-diversity benefits adding to its value and another way a different value can arise is through risks associated with certain projects such as the risk of fire reducing the value of credits generated via carbon forestry projects.

 

So, the challenge here is to design a mechanism that could recognise the multiple values of carbon without fragmenting the market into lots of small markets for each type of carbon credit and so we ran an experiment that tested different mechanisms for doing this.

 

[Image changes to show a new slide and David can be seen inset talking in the top right and text appears on the slide: Recognising multiple carbon values, Experimental set up, Three sellers, each offering a type of carbon, Six buyers, each with own preferences for each type of carbon and demand level, Mechanisms tested, Mixed market – single price of carbon, Parallel auction – separate market for each carbon type, Parallel conditional auction – separate auction for each carbon type, but buyers can use AND/OR functions, Results, Parallel conditional auction increased volume of carbon purchased and increased total market surplus, Also recognised the different values of carbon, reflecting buyers’ preferences]

 

So, in the experiment we had three sellers each offering a different type of carbon and there were six buyers each with their own budget and preferences for each type of carbon and the experiment ran for 12 rounds and we tested three mechanisms. The first was just a mixed market which represents the current market where all types of carbon are mixed together and there’s just a single price for one type of carbon credit. The second mechanism was a parallel auction where there’s a separate market for each type of carbon credit allowing for different prices and each buyer can just submit bids for any type of carbon or a combination of types and the third mechanism was a parallel conditional auction where a buyer could submit bids for any type of carbon but with and/or functions.

 

So an example is, a buyer could submit a bid for x units of type A and y units of type B or z units of type C. An optimisation algorithm then selects the, selected the buyer/seller matches that maximise the total market surplus. We found that relative to the other formats, the parallel conditional auction increased the volume of carbon credits purchased and increased the total market surplus and it was also able to recognise the different values of carbon reflecting buyers’ preferences as did the parallel auction to a lesser extent. So it seems to be a promising mechanism for being able to recognise different, different carbon values while maintaining high volume of trade in the market. And that’s me, finished.

 

[Image changes to show a white screen and the CSIRO logo and text appears: CSIRO, Australia’s National Science Agency]