experimental testbedding market mechanisms for environmental … · 2017. 9. 20. · field lab...
TRANSCRIPT
Experimental Testbedding Market Mechanisms for
Environmental Policy DesignTim Cason
Professor of Economics and Director, Vernon Smith Experimental
Economics LaboratoryPurdue University
An Absurd Headline
• “First Test Flight of Air Force Transport Plane Crashes with 43 Passengers on Board– Physicists Puzzled: ‘It should have flown’– Aeronautical Engineers Wonder Why they Were Not Consulted
– Air Force Hails Test as a Stunning Success”
• What did they forget to do?
Central Tenet for Testbedding: Markets are influenced by the same forces (although perhaps
in different ways) in the lab and field• Experiments provide insight into whether predictions developed through theoretical reasoning can be applied to more complex field conditions
• Laboratory markets are populated by profit‐motivated human agents, just as markets are in the field
• Wind tunnel testbedding– Imagine testing a new wing design on an airplane without first assessing its actual aerodynamic properties in controlled wind tunnel testing
– Note that such testing focuses on specific components of a new aircraft, not the entire system
Field TheoryLab Experiment
Central Tenet for Testbedding: Markets are influenced by the same forces (although perhaps
in different ways) in the lab and field• Experiments provide insight into whether predictions developed through theoretical reasoning can be applied to more complex field conditions
• Laboratory markets are populated by profit‐motivated human agents, just as markets are in the field
• Wind tunnel testbedding– Imagine testing a new wing design on an airplane without first assessing its actual aerodynamic properties in controlled wind tunnel testing
– Note that such testing focuses on specific components of a new aircraft, not the entire system
Field TheoryLab Experiment
The Market Design Problem and the Laboratory
• In some cases trading institutions may already exist, and have plausible alternatives, which can be modeled theoretically with some simplifications
• Complementing this, experimental designs can mimic the institution rules that are in effect in the field market
• I.e., “experiments are used as models of the real‐world system about which knowledge is sought” (Bardsley et al., 2009).
Overall Research Objective
• Make a correct (or at least a better‐informed) guess about what will happen if a particular policy is adopted
• (Testbedding experiments can also be used for market design and market engineering—such as for creating entirely new and complex combinatorial auctions with package bidding—but my examples will not include those)
Example #1: Auction Rules can Influence Allocative Efficiency and Price Discovery
• Earliest government emission permit auctions (U.S. SO2 permits, starting in 1993) resulted in biased price signals for this emerging market– Cason (1995); Cason & Plott (1996)
• Riskiness of the EPA auction likely was one factor contributing to its abandonment (by traders) in favor of a negotiated price market with higher transaction costs (Joskow et al., 1998)
• Uniform price auctions (such as an ascending price clock auction) reduce incentives to strategically manipulate bids to influence price
Unique Rules for the EPA Annual SO2Permit Auction
• Auction ~2.8% of permits each March (0 reserve price)• Private holders could offer permits as well
– Buyers pay bid price (discriminative, “pay as bid”) and lowest offer prices matched with highest bids
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14
Pric
e
Quantity
Offers to SellBids to Buy
Uniform Price
EPA Auction
EPA reserve(offered at 0)
Multiple equilibria exist for a variety of underlying environments
Do Human Subjects Recognize the Auction Incentives? What is the Performance Impact?
Within minutes, subjects recognize the incentives to offer low prices to obtain the higher bids
(Bid schedule begins to flatten)
Combination of downward bias in bids and offers results in a price bias for the EPA rules.
Source: Cason & Plott (1996)
(Other, less stable environments led to similar conclusions.)
So What Ever Happened to the EPA Auction?
• EPA considered changing to uniform price rules, but by the time the proposed change appeared in the Federal Register the non‐auction market had matured– Virtually no private allowances were offered 2003‐2009
• New regulations after 2010 made SO2 market obsolete
$0.00
$50.00
$100.00
$150.00
$200.00
$250.00
‐ 50,000 100,000 150,000 200,000 250,000 300,000 350,000 400,000 450,000 500,000
Bids
Offers
2001 Spot Auction Results
(Enron)
(Lawsocieties)
(AEP)
(EPA)
(Actual pricesof private offersare unknown)
97.8% of sold permits were offered at 0 by EPA, and 2.2% were privately offered.
Note that alternative, combinatorial designs (e.g., Porter et al. 2009) could have improved value of an EPA-sponsored auction
(Bids indicatingrevealed permitdemand)
Experimental Models to Investigate Implications of Alternative Design Details
• Experimental models are informative for institutional design in part because purely theoretical modeling is constrained by the theorist’s imagination (Smith, 2008)– An experiment leaves room for other, unimagined factors to influence outcomes
• Experiments can serve as test beds to try out new rules and institutions, even those guided only by the designer’s intuitions– A very low‐cost way to identify weaknesses and strengths of alternatives
Example #2: Banking and Price Volatility
0
20
40
60
80
100
120
140
160
0 20 40 60 80 100 120 140
Pric
e
Transaction Number
Figure 2: Transaction Prices for Session BUN3, with Banking and Uncorrelated Emissions Shocks
Prices (Pre-Shock)Prices (Post-Shock)Mean PricePeriod Break
Prices are relatively stable over time when traders can use banking to store unusedemissions permits or carry forward permits as a buffer against emissions shocks
Especially if emissions shocks are correlated (e.g., due to weather), price spikes and crashes occur if banking
is prohibited.
0
20
40
60
80
100
120
140
160
0 20 40 60 80 100
Pric
e
Transaction Number
Figure 3: Transaction Prices for Session NBCO3, with No Banking and Correlated Emissions Shocks
Prices (Pre-Shock)Prices (Post-Shock)Mean PricePeriod Break
Examples of price crashes following negative emissions shocks
Examples of price spikes following positive emissions shocks
Source: Cason & Gangadharan (2006)Other designs, such as overlapping permit validity periods, can also reduce price volatility (Carlson et al., 1993)
Similar crash seen at end of EU ETS Phase I (2007) due to overallo-cation of non-bankable permits
Example #3Price Controls in Permit Markets
• Price ceilings can help firms avoid exorbitant costs associated with price spikes due to volatility or overly‐aggressive abatement targets
• Price floors stimulate investment in emissions abatement technologies
• Various proposals for different price controls in GHG permit trading legislation in U.S.
• Price controls can be implemented in various ways, such as hard (Cantwell‐Collins) versus soft (Waxman‐Markey) price ceilings
• Hard ceiling is an absolute price control– No emissions limit—effectively
converts to an emissions tax
• Soft ceiling: A reserve auction with a minimum reserve price
PI
QI Quantity
Price Supply
Demand
PM=PSC
QCQM
InitialAuction
ReserveAuction
PI
QI Quantity
PriceSupply
DemandPSC
QM = QC
PM
InitialAuction
ReserveAuction
PI
QI Quantity
Price
Supply
Demand
PT = PC
QM
Summary of Late‐Period Prices
Hard ceiling (not surprisingly) controls prices effectively.Soft ceiling does not control prices as well, since traders do not fully
anticipate the availability of lower-priced reserve permits.Reserve auction, without a minimum price, actually results in lower
prices than the soft ceiling.
Source: Perkis, Cason & Tyner (2013)
Example #4: Compliance and Enforcement
• Compliance responds to enforcement efforts, and it interacts with other permit design features– E.g., noncompliancemay increase with more flexible banking and trading rules (Cason & Gangadharan, 2006; Murphy, Spraggon & Stranlund, 2009)
– Permit market equalizes compliance costs across firms, but experiments reveal that net permit buyers tend to have lower average compliance (Murphy & Stranlund, 2006)
– Increased enforcement has both a direct positive effect on compliance incentives, but also an indirect negative effect by forcing permit prices higher (M&S)
Permit Transaction Price Differences for Low and High Monitoring Intensities
100
120
140
160
180
200
220
0 1 2 3 4 5 6 7 8 9 10 11 12
Pric
e
Period
Average Median Permit Transaction Prices for Low Monitoring and High Monitoring Treatments
Low MonitoringHigh MonitoringFull Compliance Equilibrium Range
LowMonitoring
HighMonitoring
Source: Raymond & Cason (2011)
Example #5:Across‐Market Trading of Permits
• Direct firm‐to‐firm across‐market trading may have low transactions costs, and helps maintain thickmarkets, but raises (cross‐jurisdictional) enforcement concerns (Stavins, 2007; Metcalf, 2009)
• Restricting trading to occur through intermediaries may raise transactions costs, and could also harm price discovery by creating very thinmarkets
• Inter‐market trading allows prices to reflect overall (across market) marginal abatement costs, increasing efficiency
Example Autarky Session(4 separate markets)
160
180
200
220
240
260
0 50 100 150 200 250 300 350 400
Price
Transaction Order
Market 1Transactions
Market 2Transactions
Market 3Transactions
Market 4Transactions
Market 1 Median
Market 2 Median
Market 3 Median
Market 4 Median
Market 1 & 3Equil.
Market 2 & 4Equil.
Period Break
Source: Cason & Gangadharan (2011)
Example Session with Intermediaries Required for Inter‐Market Trades
160
180
200
220
240
260
0 100 200 300 400 500 600 700
Price
Transaction Order
Market 1TransactionsMarket 2TransactionsMarket 3TransactionsMarket 4TransactionsInt'l MarketTransactionsMarket 1 Median
Market 2 Median
Market 3 Median
Market 4 Median
InternationalMarket MedianMarket 1 & 3 Equil.
Market 2 & 4 Equil.
Int'l Market Equil.
Period Break
Source: Cason & Gangadharan (2011)
The costs of requiring intermediation for across‐market trading
• The treatment that requires that trades occur through intermediaries can be compared to one with direct firm‐to‐firm trading
• This reveals that requiring intermediation– does not substantially increase (eventual) price deviations from the competitive equilibrium, but it does
– (1) slow convergence to equilibrium; – (2) lowers efficiency; – (3) increases volatility; and – (4) raises transaction costs
Example #6: Using the Wind Tunnel to Refine Conservation Auction Details
• State of Victoria, Australia, Department of Natural Resources & Environment sought to deliver funding for conservation works and protect biodiversity: BushTenderTM auctions– Landholders bid for payments to carry out conservation activities
What Information to Provide Bidders before the Auction?
• Field Officers assess significance and quality of native vegetation, collect data, and discuss management options with private landholders
• Should they provide information on their project’s value to the landholders?
Testbed this Choice in Laboratory
• In some sessions bidders received info about the environmental “benefits” of alternative projects they could offer
• In an alternative treatment, they learned only their projects’ cost
• Government wants them to bid near their cost, so they don’t “overpay” for improvements in environmental quality
Results Indicated that Bidders Exploit Information on Quality to Obtain Higher Payments
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
All Qualities 25 to 40 40 to 80 80 to 120 120 to 160 160 to 200 200 to 240 240 to 280 Over 280
Off
er/C
ost R
atio
Quality Range
Median Offer/Cost Ratio by Information Treatment and Quality Range: All Final Round Offers for Periods 1-5
Benefits Revealed
Benefits Unknown (All Sessions)
This quality range represents 19% of the projects, but 63% of the SOLD projects.
Costs are directlyobserved in labtestbedding
Source: Cason, Gangadharan and Duke (2003)
Results Indicated that Bidders Exploit Information on Quality to Obtain Higher Payments
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
All Qualities 25 to 40 40 to 80 80 to 120 120 to 160 160 to 200 200 to 240 240 to 280 Over 280
Off
er/C
ost R
atio
Quality Range
Median Offer/Cost Ratio by Information Treatment and Quality Range: All Final Round Offers for Periods 1-5
Benefits Revealed
Benefits Unknown (All Sessions)
This quality range represents 19% of the projects, but 63% of the SOLD projects.
Costs are directlyobserved in labtestbedding
Source: Cason, Gangadharan and Duke (2003)
Implementation in the field:
Hides information from
bidders about the benefits
provided by their projects.
What about “External Validity”?
• External validity is not a binary condition, either on or off, but rather it should be considered as having various degrees and dimensions
• If both lab and field behavior are governed by the same principles (as is usually assumed in economics), it makes little sense to distinguish “external” and “internal.” All the behavior is “internal.”
When should we be more confident about external validity?
• An external validity concern can just as easily (and in most cases more justifiably) be raised for any theoretical model of an economic process
• Experimental models can help inform theoretical models, which helps us understand potential outcomes in the field– External validity to the field may not even be a main goal
• We can be more confident about external validity for types of conclusions where mechanisms and/or logic for the underlying process is likely to have a fieldparallel (e.g., price spikes following correlated excess emissions)
Summary: So do we learn anything from these experiments?
• These examples illustrate that experiments can help researchers and regulators understand the implications of many market design details, which can influence the success or failure of market regulations– evaluate specific trading institutions (e.g., auction formats)– how permit banking affects price volatility in the presence of correlated emissions shocks
– Performance of alternative price control mechanisms– how enforcement affects compliance incentives– implications of across‐market restrictions on permit trading– what information should be revealed in conservation auctions
Conclusion: Do we also need field experiments and naturally‐occurring data to get the right answers?
• Of course we do!• Parallelism with the field could be enhanced by
• including non‐student subjects • (but outcomes are often not sensitive to the subject pool)
• incorporating realistic institutional details, such as the political processes leading to market rules, and;
• actual environmental consequences from abatement, compliance and conservation decisions