Transcript Market Architecture - Market Design Inc.

Market Architecture
Robert Wilson
Stanford University
New Markets in Basic Industries
• Privatization / “Liberalization” (worldwide)
Deregulation / “Restructuring” (in U.S.)
– Communications, Energy, Transport, Water
• Motives:
 Efficiency &  Investment
Arguments:  Scale   Contestability
Markets “get prices (incentives) right”
Light-handed regulation suffices
• Implementation requires Market Design
2
Elements of Market Design
• Inputs: Scarce resources. Outputs: Products, services.
Instruments: Tradable rights and contracts.
Markets: Physical forward and spot. Financial hedges.
Prices: Bids. Allocation rules. Settlement rules.
• Efficiency: Gains in short-run. Investments long-run.
• Incentives:
– Procedures invulnerable to gaming
– Rules + Incentives  Efficiency (2nd Best)
– Mitigate or control market power
3
Wholesale Electricity Markets
• Energy: Long-forward. Day-ahead, Day-of. Real-time.
– Markets: Bilateral. Exchange. Spot
[ = Real-time]
• Transmission: Day-ahead. Real-time.
– Managed by SO
[ = System Operator]
– Markets: Congestion pricing, or purchase counterflow
• Reserves: Capacity = Day-head. Energy = Real-time.
– Markets: Auctions by SO. Real-time control by SO.
• Hedges: Energy (Futures, CFDs).
Transmission.
– Markets: Auctions by SO. Secondary markets.
4
Simplified Electricity Market
Transmission
Capacity BA
Congested Price in A
P(A)
Supply Incs
in A || in B
Uncongested
Energy Price
Congested Price in B
P(B)
Supply Decs
in B || in A
Transmission: Uses A-Incs & B-Decs
[ Transmission Charge = P(A)  P(B) ]
Spot Market: Uses all Incs & Decs
Supply into A net of
Demand in B
Demand into A net of
Supply in A
Imports into Region A
Exports from Region B
Note: actual market has many
zones or nodes
5
Design Issue #1
Centralized v Decentralized
(Complete v Incentives)
• Centralized = consolidated markets
SO optimizes everything: energy, trans., reserves
Prices = shadow prices on constraints
– Prices are right if model good & data accurate
• Decentralized = separated markets
PX clears markets, SO conducts auctions
Prices = clearing prices
– Prices are right if markets are complete &
perfectly competitive
6
Institutional Perspectives
• Centralized designs are based on
– Relational contracting. SO = Traders’ agent.
Organizational structure inherited from utilities.
– Objectives = reliability, coordination, pricing.
– Incentives = sanctions. Abuses are penalized.
• Decentralized designs are based on
– Voluntary participation, bidding, availability.
Few markets based on simple market clearing.
– Objectives = efficiency, competition, min-SO.
– Incentives = market prices for deviations.
7
Theoretical Perspectives
• This debate is like 1930s Lange-Lerner.
Basic fact is primal-dual equivalence of
quantity and price mediated mechanisms.
• Can incomplete and imperfect markets
match optimization?
– Requires good price discovery & repeat mkts.
• Is decentralization necessary to promote
competition and strengthen incentives?
– Centralized designs cannot provide sufficient
incentives if pricing is constrained (Vickrey).
8
Examples of Incompleteness
• Simple contracts with retail customers
– So demand highly stochastic (and no storage!)
• Energy: Forward markets clear each hour
independently. No contingent contracts.
• Transmission: Scarce resources are
not priced when large zones are used.
• Reserves: Capacity is priced imperfectly on
relevant quality dimension -- response time.
These are a few among many.
9
Example of Gaming
• Centralized: set price at each node.
– Nodal Price = Energy Price + Injection Charge
from [Demand = Supply] & [Trans.  Capacity]
• Decentralized: 1st clear DA energy market,
2nd adjust energy in zones, 3rd use incs/decs.
– Gaming: bidder sells huge quantity day-ahead
that the SO is forced to sell back at low-priced
dec in the spot market. Enter at congested node!
• This is inevitable consequence of unpriced
scarce resources (i.e., incomplete market).
10
Other Effects of Incompleteness
• Impaired Efficiency
– Intertemporal effects are ignored.
Startup costs, ramping constraints,…
– Flexible resources & reserves are under-priced.
Due to lack of contingent planning or contracts.
– Entrants attracted to wrong locations.
Due to unpriced transmission constraints.
– Sequence of markets depends on rational
expectations. Anticipate transmission charge.
• Impaired Competition
– Each incompleteness invites gaming.
11
Remedies for Incompleteness
• Sequence of markets
– Repeated trading of simple contracts
(long-forward, day-ahead, hour-ahead, spot)
approximates complete market.
Transmission is similar if pricing is locational at nodes.
• Better price discovery
– Iterative auction approximates Walrasian model
but requires good activity rules.
Example: “use it or lose it” options.
• Complex market-clearing procedures
– Requires consolidated markets run by SO ?
12
Summary
• Can centralization work well?
[complete]
– Strengthen competition, or use incentivecompatible settlement rules (e.g., Vickrey).
• Can decentralization work well? [incentives]
– It works - systems are operating successfully.
Most are seen as promoting competition.
– But to price efficiently and prevent gaming
requires effectively complete markets.
– Can this be done without consolidated market?
Working hypothesis is ultimately yes - after
many improvements to control gaming.
13
Design Issues #2, #3
(Addressed in text of paper)
• Design details of each market
– Energy: bilateral contracting, or an exchange.
– Transmission:
• Implementing nodal pricing in decentralized market
• Markets for hedges, firm transmission rights
– Reserves: design of two-dimensional auctions
(available capacity, contingent energy supply).
• Mitigation of market power
– Divestiture. Contract cover. Supply auctions.
14
Comparisons with Other Industries
• Transport industries with similar issues but
different solutions:
– Gas pipelines
– Rail networks
– Telecommunication networks
• To what extent are solutions derived from:
– Technical differences?
• Point-to-point transport
– Ownership and control?
• Each network privately owned and managed, no SO!
15