Faculté des sciences de l’agricultureet de l’alimentationUniversité Laval

Keep it Down: An Experimental Test of the Truncated Uniform Price Auction

2009 NAREA Workshop on Experimental Methods in Environmental, Natural Resource, and Agricultural

Economics

Maurice Doyon

Daniel Rondeau

Richard Mbala

2

Outline

• Context

• Research objectives

• The truncated uniform price auction

• Experimental design

• Results

• Closing remarks

3

Context

• Since 1972, Canadian egg producers are required by law to hold a production permit (quota) for the laying hens they own.

• Supply control generates oligopoly rents and makes individual transferable quota highly valuable.

• In the egg industry (as in other agricultural quota markets such as milk and chicken), prevailing conditions tend to provide sellers with a distinct advantage and tend to yield high quota prices.

4

Context• High quota prices are a concern for Quebec’s egg producers

for two central reasons. • Large sums of money are capitalized into this non-productive

asset (Boots et al 1997, Alvarez et al 2006). • Detrimental to a renewal of the industry, making it difficult for

young individuals to become producers.

• No official market exist for «laying hens» and a few feed suppliers have become de facto quota brokers

• Feed suppliers have an informational advantage and purchase much of the available quota and tie its resale to exclusive long terms input contracts

5

Context

• It is within this context that the Quebec Federation of Egg Producers decided to create a centralized market institution for quota trading.

• The objective of the Federation is to implement a market mechanism that fosters efficient trading of a relatively low volume of units and provides the strongest possible protection against aggressive oligopoly pricing emerging from a small number of sellers when demand substantially outstrips supply.

6

Objectives

• Doyon and Rondeau (2006) explored various market designs that could potentially meet the prerogatives of Quebec egg producers.

• They concluded that a truncated Uniform Price Auction (UPA) has interesting characteristics that should be tested.

• Therefore, the objective of this study is to test the effect of the truncation rule on market price, buyers and sellers bids, the number of unit traded and economic efficiency

7

Truncation rule• Uniform price auctions are efficient since they ensure that buyers

with the highest subjective valuations for goods obtain those goods and, in exchange, sellers receive returns that approximate the totality of this value (Ausubel and Cramton 2002, Khrisna 2002).

• Any intervention aimed at modifying the outcome of a uniform price auction is likely to reduce the efficiency of that auction

• Doyon et al. (2008) have explored modifying a UPA with a tax on units offered but not sold and excluding highest bidders and offers from the bid and asks schedules. Experimental results indicate that they achieved little price reduction at the expense of important loss of efficiency

8

Truncation rule

• Under the T-UPA, both buyers and sellers are invited to submit bids and asks consisting of the number of units they wish to buy or sell, and the price at which they are prepared to buy or sell respectively.

• While multiple units are involved, a restriction is imposed that participants can only submit a single bid or ask

• The intersection of the bid and ask curves determines the quantity that will be supplied to the market. In standard UPA, the units to be sold are those starting with the lowest asking price on the supply curve, moving sequentially up to the intersection of the bid and ask functions. This rule is not changed in presence of the truncation rule.

9

Truncation rule

• Whereas a standard UPA simply allocates the units in decreasing order of bid price, the T-UPA favors bidders on extra-marginal units.

• Bidders are declared eligible to purchase some of the traded units as long as their bid price is equal or greater than the price asked by the seller of the last unit (the unit at the intersection of the bid and ask curves).

• In a reversal of normal rules, the units are sold in priority to bidders who submitted the lowest price bids rather than the highest

10

Truncation rule

G

11

Experimental design

• Two sets of experiments were conducted to compare the impact of the T-UPA against the UPA. In all cases, the underlying design and parameters were identical across all experiments.

• In all sessions, the market was made up of seven buyers and five sellers who kept the same role for a total of thirty periods of market interaction with each other.

• Subject could visually determine that there were twelve participants in the market, but were not told how many were buyers and sellers

• All information on quantities and reserve prices (induced values) was private and confidential

12

Experimental design• The subject’s task in each round was to submit a bid or ask

containing two piece of information: 1) the number of units offered (seller) or sought (buyer); and 2) the price they were prepared to transact at.

• At the end of each period, each individual privately received their personal results in the form of the number of units he/she transacted, the market clearing price and their personal profits.

• The first six periods of this experiment were unpaid practice rounds. Each of these periods implemented different vectors of market parameters (individual quantities and reserve prices), allowing participants to familiarize themselves with their task, the operations of the computer interface and the impact of their decisions

13

Experimental design

• The experiment then proceeded with three sets of eight periods each, generating real earnings. During each set, the same overall vector of market parameters were kept, but rotated across participants with each new period.

14

Experimental design

15

Experimental design• The experiments were implemented using z-tree and conducted at

the CIRANO laboratory for experimental economics in Montreal, Canada.

• Subjects were university students from all disciplines recruited via email invitations using the laboratory’s standing list of student volunteers.

• Individuals were randomly assigned to their role of buyer or seller. They received oral explanations and instructions.

• Sessions lasted approximately 90 minutes and participants received between $18 and $50, (including a $5 show up fee).

16

Results

Number of Unique Participants and Number of Individual Observations (decisions) by Model Economy Model A Model B Model C Total Unique Buyers/Sellers UPA 70/50 21/15 21/15 182/130 Unique Buyers/Sellers T-UPA 77/55 21/15 21/15 119/85 Total Unique Buyers/Sellers 147/105 42/30 42/30 301/215 Number of Bids/Asks UPA 560/400 168/120 168/120 1456/1040 Number of Bids/Asks T-UPA 616/440 168/120 168/120 952/680 Total Bids/Asks 1176/840 336/240 336/240 2408/1720

17

Results-Transaction Prices

• Result 1: The uniform transaction price under the T-UPA allocation rule is significantly lower than under the UPA rule.

18

Results-Number of Units Traded

• Result 2: The T-UPA results in a smaller number of units traded than the UPA allocation rule

19

Results-Market Efficiency

• Result 3: For the experimental conditions implemented in this study, the T-UPA captures 15 points of percentage less of the potential gains from trade than the UPA

UPA

T-UPA

Theoretical prediction for perfect competition

P-value

Periods 1 à 8 (Model A) 88,8% 76,1%

100%

< 0 Periods 9 à 16 (Model B) 97,1% 77,9%

100%

< 0 Periods 17 à 24 (Model C) 93,3% 74,3%

100%

< 0 Total 91,3% 76,1%

20

Results-Market Efficiency

• Two factors contribute to the loss of efficiency. • First, is the loss associated with a sub-optimal number of units

traded already noted• Second, the “misallocation” of units to extra-marginal buyers

and sellers.

• Result 4: Units defined as extra-marginal based on the underlying distributions of reserve prices are more likely to be traded in the T-UPA auction and contribute significantly to lowering the efficiency of the market rule.

21

Results-Market Efficiency

•For the regular UPA, and across all markets in the dataset, only 1.05% of the 5,507 units exchanged in our experiments were extra-marginal on the demand side.

•In contrast, this proportion climbs to 10.3% of the 5,542 units traded in the T-UPA.

•This confirms that the T-UPA rule provides significant opportunities for less efficient buyers to purchase units

22

Results-Buyer Behavior

• Result 5: Buyers submit significantly lower bids in the T-UPA than in the UPA

23

Results-Seller Behavior

• Mixed results

24

Results-Seller Behavior

• the T-UPA produced much less volatility in Model B and substantially reduced average sellers bids to reach 174%, a value in line with the other Models.

Average Bids Submitted by Sellers as a Percentage of their Induced, Aggregated by Models

UPA

T-UPA

Theoretical prediction for perfect competition

P-value

Periods 1 à 8 (Model A) 152% 159%

100%

0,61

Periods 9 à 16 (Model B) 267% 174%

100%

< 0

Periods 17 à 24 (Model C) 184% 164%

100%

0,25

25

Closing remarks• The truncated rule significantly reduces market clearing price. It

does so by significantly reducing buyers’ bids and by reducing the asking price of sellers in extreme market conditions, when sellers have shown an ability to exercise market power in a standard uniform price auction.

• In situation of large excess demand and wide difference between the reserve prices of the marginal buyer and seller, the truncated rule was able to dramatically reduce the ability of sellers to push the equilibrium price up.

• This is an important finding since this type of conditions underlying thin markets such as quota market in agricultural sectors or in other sector could benefit from this type of market design.

26

Closing remarks

• However, reduction in market clearing price and the ability to mitigate market power of sellers in certain market conditions come to a cost.

• This cost is a drop in the number of units transacted and an undesirable allocation of some units to buyers who value them less than some excluded participants. These combined effects result in a loss of efficiency in the order of 15% in average in our experiment.

27

Closing remarks

• Nevertheless, these results compare advantageously with the modifications to the uniform price auction market tested by Doyon et al (2008).

• To reach a mere 5% reduction in market clearing price, 50% of the unit traded and 52 point percentage in efficiency were sacrificed.

• In Model B, a 51% reduction over the non competitive market price is obtained in exchange of a loss 4% in the number of unit traded and an efficiency loss of 19 points percentage.

28

Closing remarks

• These differences could be explained by the flexibility of the truncated rule, as opposed to an exclusion rule as tested by Doyon et al (2008).

• While the exclusion would be fixed and would always be blindly applied, the truncation rule does not apply in the absence of excess demand and its effect is modulated with the level of excess demand.

29

Closing remarks• Nevertheless, efficiency losses are significant.

• Impact on long term competitivity of the Quebec egg production sector?

• However, it represents a sizeable improvement over current practices, where a large portion of the gains from trade are likely captured by feed suppliers

• Efficiency loss should be seriously considered and weighted against the perceived benefits of lower market prices before implementing a truncated uniform price auction in the field.

• First real auction, based on our design was realized June 1st

Questions ?

Maurice DoyonProfessorDepartment of Agricultural Economics and Consumer ScienceLaval University

Email: maurice.doyon@eac.ulaval.ca