Porter&Zona_Ohio School Milk Markets

OHIO SCHOOL MILK MARKETS: AN ANALYSIS OF BIDDING

Porter, R. H. and Zona, J. D. (1997)

Dr. Hwa Ryung Lee

Industrial Organization and Competition Policy

Ivan Jivkov, Stefanie Marty, Matthias Hafner, Daria Zürcher-Bevza

University of Zurich

HS2012

Contents

1. Introduction

2. The Market Characteristics

3. Bid Rigging Mechanism

4. The Competitive Market Model

5. The Behavior of the Defendant Firms

6. Estimating Damages to Schools

7. Conclusion

8. Discussion

2

• Collusion is a pervasive phenomenon in school milk auctions – Guilty pleas in at least a dozen states – 90 people sent to jail – Fines in excess of $90 000 000

• Porter & Zona examine school milk procurement process, bidding data, and supply in Ohio during 1980’s, and compare bidding behavior of defendants with a control group

• The bidding behavior of the accused firms is consistent with collusion

• Average effect of collusion: 6.5% on market prices

3

Introduction

The Market: The Nature of Competitive Interaction between Suppliers

• Each of over 600 school districts in Ohio solicits bids on annual supply contracts of milk

• Firms that elect to bid submit a list of prices

• The lowest bid wins

• Coolers, straws or napkins might be included in the contracts; escalator clauses may be provided

• Potential suppliers: processors and distributors

• Entry barrier: the costs of a processing plant

4

The Nature of Possible Collusion: Market Characteristics Facilitating Collusion

5

Market Characteristics Their Role in Facilitating Collusion

Demand is inelastic Constant amount of milk purchased even at elevated prices

Predictability of demand Credibility of punishment threats

Competition only on price (homogeneous good)

Ease of the coordination of cartel behavior

High transportation costs Barrier for distant competitors

Public announcement of bids and the identity of bidders

Sure detection of deviation

Small and stable set of firms submitting bids

Higher likeliness of an agreement on joint behavior

The Nature of Possible Collusion: Market Characteristics Facilitating Collusion

6

Market Characteristics Their Role in Facilitating Collusion

Easy definition of markets according to school district boundaries

Allocations by assignment of territories

Availability of competitors’ price list Communication of intentions

Mutual custom, membership in dairy trade associations

Direct communication, pretext for meetings

Assignment of districts to levels in price decks

Ease of complementary bidding

Disorganized letting of contracts Adjustment of cartel behavior, immediate retaliation

Multimarket contact

Competition is not fierce, collusive schemes are more feasible

The Nature of Possible Collusion: Bid Rigging Mechanism

• How could conspirators limit competition?

1. Refrain from bidding against each other

2. Submit bids at inflated levels to create the appearance of competition (Phantom bidding)

• Coordination of bid rigging behavior:

1. Scheme of respecting incumbencies

2. Assignment of geographic territories to individual firms

7

The Competitive Market Model

• Control Group vs. Defendants

– Bid submission decision

– Level of bid given submission

• Approach:

1. Create a control group

2. Build reduced form model for bid submission decision and level of bid for the control group

3. Compare control group with defendants

8

Control Group Bid Submission Decision

Factors affecting the bid submission decision: Probit-Model

• Findings:

- Processor (+) (≠Distributor)

- Direction (+)

- Distance (-)

9

Control Group Bid Level Contingent on Submission

• Trade-off of higher profit vs. lower probability of winning

• Bidding behavior:

1. Variables that contribute to cost e.g. Distance, Processor/Distributor, # of deliveries

2. Variables that reflect the competitive characteristics of the market e.g. Closest/second closest firm

10

Factors affecting the bid level: OLS-Model

Findings:

• Distance (+)

• Processor (-)

• Firms closest to school district have competitive advantage

Control Group Bid Level Contingent on Submission

11

• Bidding by non-defendants is consistent with competitive bidding under standard models of spatial competition.

Comparing to Control Group

Approach:

– Bid submission decision (test 1)

– Level of bid given submission (test 2)

Test 1: difference in bid submission decision (likelihood):

For each defendant:

1. (model 1) H0: Slope coefficients are the same for control group and defendant firm (intercepts may differ)

2. (model 2) HA: Separate slope coefficients for the defendant

3. Likelihood ratio test statistic (= which model is more probable to fit the data better)

12

Comparing to Control Group

Test 2: Differences in bid level:

LaGrange Multiplier test of equal slope coefficients in the bid level:

H0: The slope coefficients for control group and defendants are zero

For both tests H0 is rejected:

behavior by SW Ohio defendants differs from the control group!

But behavioral differences are not necessarily caused by anticompetitive motives !!!

13

* 1. Compute the predicted bid for the firm based on control group bid estimates; 2. Construct a residual (difference of calculated and real bid), for each district and year (if possible); 3. Regress residuals on the independent variables of the control group equation; 4. F-Test: Are all slope coefficients zero?

Comparing to Collusive Strategy

Approach:

• Examine deviations (predicted - actual) of defendants bidding behavior from control group

14

Defendants submitted more bids if the plant was less than 30 miles away from the school district.

Comparing to Collusive Strategy

Bidding in parallel fashion between defendants

• Model is based on public information and on the specifications of the bid level equations

Test:

• H0: Knowledge of one firms bid should not help predict what another firm will bid

• HA: Knowing that one cartel member bids higher than the predicted level helps predict that another will bid above that level (complementary bidding)

15

Comparing to Collusive Strategy

16

• Sample sizes largest for Meyer, Trauth, and Coors

• Result supports testimony by representatives of Meyers and Coors.

Outcome • Consistency of

complementary bidding scheme, i.e. if one bids high, the others bid high as well.

Estimating Damages to Schools

• The authors are interested in the collusion effect on prices charged to schools.

• Estimate the percent mark-up in price due to collusion in various auctions

17

Conclusion The Main Findings

• A number of the characteristics of the auction mechanism facilitate collusion

• Systematic deviations between predicted and realized bids have been observed; the pattern of deviations is consistent with collusion

• The winning bids of conspirators are higher than competitive bids which causes consumers (taxpayers who subsidize students’ meals) to overpay for milk by 6.5%.

18

Conclusion Ohio School Milk Procurement vs. New York State Highway Paving

19

Auctions New York State Highway Construction (Porter and Zona, 1993)

Ohio School Milk Procurement (Porter and Zona, 1997)

Similarity Phony higher bids by ring members unrelated to cost measures

Difference No access to conract-specific information Analysis of bid level is not possible

Information about school districts and individual contract terms are available Focus on the decisions whether to submit a bid AND on the level of submitted bids

Critique Descriptive character of the results. A public adoption of the test procedure would incentive cartels to

tailor their phantom bids to disguise their collusive behavior.

Discussion

Henriques and Baquet, “Evidence Mounts of Rigged Bidding in Milk Industry” (1993):

• “Some of the sharpest questions about corporate tolerance of bid-rigging focused on the fact that many companies were paying their executives' legal expenses in the investigations. "When you have a problem that creeps across state borders like this, you either have upper-management involvement - or you have an upper-management problem," said Mary Sue Terry, the former Virginia Attorney General (…). Indeed, that attitude may have been supported through official inaction.”

• → How could corporate tolerance of bid-rigging be dealt with?

• “The wave of antitrust scrutiny of the milk industry began in Florida in the mid-1980's, when (authorities) began to apply computerized bid-analysis techniques, initially developed to analyze highway construction bids, to examine milk bidding. (…) E. Linwood Tipton, president of the Milk Industry Foundation, said he believed the industry's local problems had been exaggerated by the "spotlight" of the Federal prosecutions and state civil suits. "If you put the same scrutiny on any other industry, you'd find a significant number of cases there, too," Mr. Tipton said. "When there are enormous investigatory powers directed at you, it looks like a more abnormal situation than it probably is.”

• → Why is collusion pervasive in school milk auctions? How could collusion-facilitating market characteristics be improved?

20

Literature

• Diana Henriques and Dean Baquet, “Evidence Mounts of Rigged Bidding in Milk Industry”, New York Times, May 23, 1993, www.nytimes.com/1993/05/23/us/evidence-mounts-of-rigged-bidding-in-milk-industry.html, retrieved on 11.11.2012.

• Robert H. Porter and J. Douglas Zona, “Detection of Bid-Rigging in Procurement Auctions,” Journal of Political Economy, Vol. 101, June 1993, 518-538.

• Robert H. Porter and J. Douglas Zona, “Ohio School Milk Markets: An Analysis of Bidding,” National Bureau of Economic Research, Working Paper 6037 , 1997, 30-33.

• Robert H. Porter and J. Douglas Zona, “Ohio School Milk Markets: An Analysis of Bidding,” The RAND Journal of Economics, Vol. 30, No. 2, 1999, 263-288.

• Michael D. Whinston, Lectures on Antitrust Economics, The MIT Press, Cambridge, Massachusetts, 2006.

21