empirical industrial organization - lancaster.ac.uk notes 2013 u of p-1.pdf · io‐de silva...

99
IODe Silva Empirical Industrial Organization Dakshina G. De Silva Short course Université Paris – 1, PanthéonSorbonne March – April, 2013 These notes are mainly based on Timothy Dunne’s IO – II lecture notes collected when I was a grad student and various other sources. All errors are mine.

Upload: hathu

Post on 31-Jan-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Empirical Industrial Organization 

 Dakshina G. De Silva 

Short course 

Université Paris – 1, Panthéon‐Sorbonne 

March – April, 2013 

 

These notes are mainly based on Timothy Dunne’s IO – II lecture notes collected when I was a grad student and various other sources.  All errors are mine.   

Page 2: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Introduction A definition of industrial organization:  

“Industrial organization is concerned with the workings of markets and industries, in particular the way firms compete with each other."  Two branches of I.O. a) Theories of Markets and Market Structure.  This branch treats the firm as a black box and focuses on how firms compete with each other.  

b) Theories of the Firm.  This branch investigates why some transactions are conducted through markets while others are conducted within a firm. Attempts to look inside the black box and explain things like firm size, the boundaries of the firm, and incentive schemes within the firm.  

Page 3: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Main question:  Are the real world markets perfectly competitive? (Most of the time this is the null.)    Theory          –  Cournot behavior 

Markov perfect games  

  Econometric tools    –  Simultaneous equations                 Panel data techniques                 Discrete choice models                  Nonparametric methods   Institutional framework    Data          –  Often disappointing   

Page 4: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

A Brief History of Industrial Organization  1. Harvard Tradition (1940 – 1960: Joe Bain) 

  Structure‐Conduct‐Performance  

Structure (i.e., how sellers interact with each other, buyers, and potential entrants) is a function of number of firms, technology, existing constraints, products... CR4, HHI  

Conduct (i.e., how firms behave in a given market structure) includes price setting, competition, advertising...  

Performance (i.e., technological efficiency, social efficiency, dynamic efficiency) includes consumer surplus, optimal variety, profits, social welfare...  

 

 

Page 5: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

2. Chicago School (1960 – 1980: Robert Bork … “The Antitrust Paradox")  

Firms become big for particular reasons  

Emphasis on price theory (markets work)  

More careful application of econometric techniques  

Use different market structures to understand different industry settings or markets  

Monopoly is much more often alleged than confirmed; entry (or just the threat of entry) is important  

When monopoly does exist, it is often transitory 

 

 

Page 6: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

3. Game Theory (1980 ‐ 1990)  

Emphasis on strategic decision making  

Modeled mathematically using Nash equilibrium concept  

Produces a huge proliferation of models which are often very intuitive theoretically  

However, it is difficult to know which model is the right one for a real world industry 

 

 

 

 

 

Page 7: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

4. New Empirical I.O. (1990 ‐ )  

Combines theory and econometrics in a serious way  

Sophisticated, computationally intense, complex empirical models  

Not all I.O. economists think this way or use the same methods  

This view of the world is constantly evolving 

   

Page 8: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Contemporary Issues in I.O.  

Entry and exit  

Merger analysis  

Product choice: characteristics and location  

Retail Markets  

Price discrimination  

The role of information and monitoring technologies  

Advertising  

Learning‐by‐doing  

Technological innovation  

R & D spillovers    

Page 9: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Evolution of Empirical IO I’m sure you have seen these before.  So this is going to be quick. 

Bain (1951) 

,   If we run   and look at  ⁄ 0 (This is the Lerner index.)  Main weakness:  Market structure is exogenous  

Profits and CR (+)  

  They do not depend on LI  

The (+) relationship may be due to one firms efficiency.  Therefore, 1 does not show the correct market power. 

Page 10: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Comanor and Wilson (1967) 

, , , … , )  

This is at industry level  Growth was measured in change in revenue  If p increases revenues will increase.  Then ‘Growthi’ and  are correlated 

  

 

   

Page 11: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 Orr (1974) 

 Examine net entry (Canadian manufacturing industries)  Measured net entry as an index and lagged variables.   But lagged shock is not a current year shock  Dropped variables due to collinearity — leads to specification error. 

    

Page 12: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Domowitz, Hubbard, & Petersen (1987)         Looked at the CR, and profit relationship.        A panel of 284 industries – first to use a panel   

                industry effect  

Used fixed effects.  Therefore, looking at as CR what will happen to  (time variation.) 

 Measure  from price cost margin. 

 (Rev – VC)/Rev        left with FC, Capital ratio, and .  This is a weakness. 

 Nice paper to read:  Caves, Richard E. (2007) “In praise of the Old I.O.” International Journal of Industrial Organization. 

Page 13: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Generic models of Oligopoly – Bresnahan (1987)      Homogeneous product industry    Perfect competition:  p = MC    Monopoly:      MR = MC    Oligopoly framework: 

  q12 firms  q2

   mc1(q1) = mc2(q2)       Q = q1 + q2      Market demand function: p(Q)   

  

Page 14: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Suppose a Cournot model  

F.O.C 

                  MR    MC  

 

 

2 2  

 12  

          1/

121 

Page 15: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

For n firms  

1 1 

If you chose Bertrand   

,

if12 if

0if

 

 NE     pi = pj = c  , therefore LI = 0     

Page 16: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Data  Only have p and Q  Demand function:    Q = D(p,y)  MC function:      c(Q,w)  

Competitive supply 

Monopoly supply

, ,  

Quasi supply curve 

,  

 If = 0 then market is perfectly competitive (= 1 then monopoly)  We like to estimate  but we have unobservable    . 

Page 17: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Example – not identified   1) Market demand:                  (1) 

 2) Marginal cost:                   (2) 

 

3) Quasi supply:                 

   MC 

Note that   ⁄   

             (3) 

    

Page 18: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Two equation model (solve using eq. 1 & 3)   

 

Q

MCm ( , 1

MCc ( , 0

MR

D

Q1

p1 

Page 19: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

  

 

Take Eq 1 and put a rotational variable (z) 

                   (4)  Supply relationship with    

                    (5) 

 This is identified since   and  comes in different forms.         

Page 20: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

      

 

 

 

 

 

 

 

Q

MCm ( , 1

MCc ( , 0

MR

MR’

D

D’

Q1

p1 

Page 21: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

 

 

Bresnahan (1987) 

If you only have data on market price and quantity (no cost data) and endogenous demand and supply shifters, you can identify the degree of market power from market demand and supply relationship if the demand curve structure (in the real world) is rich enough to allow for the rotation of the demand curve. 

 

   

Page 22: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Appelbaum – JE (1982) 

Industries (US): Rubber, Textile, Electrical machinery, & Tobacco 

Industry demand:    ,  

Quasi‐supply:       1  

Industry cost:       ,  

Must model , , & MC 

 

FOC for firms: 

1 ,  

 

 

Page 23: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Assumed MC constant and equal for all firms. 

, ∗  

1 ∗  

 

This means    is equal for all firms – homogeneous cost function. 

Equilibrium output level at oligopoly 

1 ∗  

Industry demand curve: 

ln ln ln  

 

Page 24: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Cost function: 

c /  

(Leontief cost function of the Gorman polar form) 

Quasi supply curve: 

1 /  

                          MC 

He estimates: 

/ 1  

 

Page 25: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Then h is approximated linearly as  

 

p & Q endogenous—simultaneous equation method.  Industry  ∗ ⁄Rubber  .019  .056 Textile  .370  .670 E. machinery  .200  .196 Tobacco  .402  .651  Contribution:  Econometrics and Data Issues:    Same demand function for all industries       Industry classification—sawing machines and power generators    Also check: Schroeter, John R. (1988) Estimating the Degree of Market Power in the Beef Packing Industry, REStat 

Page 26: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Detecting collusion 

Porter (1983) –Bell 

Porter (1983) analyzes the pricing behaviour of the US railroad cartel  (The Joint Executive Committee –1880‐1886.)   Cartel controlled freight shipment from Chicago to the Atlantic seaboard.  Collusion was legal (pre Sherman act) and the workings of the cartel are very well documented. 

Ulen (1978) said there are several instances where they thought cheating occurred. 

Cartels can increase profits by restricting output from competitive levels.  JEC agreement involved allocation of market shares. 

However, members face an incentive to cheat because price is above marginal  cost. 

JEC collected and disseminated info to its members on total quantity and average price. 

Page 27: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

In practice, it may be difficult to monitor competitor’s output.  Firms’ market share depended not only its own price, but also prices of other firms and demand conditions. 

 Model Assumptions (Green and Porter—1984)  Firms set their own production level  Firms do not know the quantity produced by any other firms ‐ they only observe the market price 

Firms’ output is homogenous (they face a common market price)  Pattern of observed prices consistent with firms following trigger strategies: periods of high prices followed by periods of low prices  “Cheaters” are punished by an industry‐wide switch to noncollusive (e.g. Cournot) quantities for a fixed period of time, resulting in lower revenues for all firms.  

Since firms do not observe one another’s output, this switch occurs once the market price falls below a previously decided “trigger price”.  

Page 28: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Collusion supported by an appropriately chosen “punishment pair” (trigger price, punishment period length).  To be effective, the “punishment pair” must make the cost of cheating be at least as large as the benefits on expectation.  Costs go up by longer punishment period or lower trigger price.  Porter wants to establish that  

(i) price wars occurred,  (ii) price wars were triggered by demand shocks 

 Econometric model is design to test whether significant switches in supplier behaviour occurred and to identify the periods in which it took place.     

Page 29: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

More on JEC  Porter argues homogenous good. 

  Grain was 73% of dead tonnage. 

  Even though endpoints of rails differed, overseas shipping rates adjusted. 

  Attention to grain without loss of generality. 

  Entry occurred multiple times in this sample. 

  New entrants were accepted into the cartel and allocated market shares. 

  JEC office took weekly accounts so that the shipments could be monitored. 

  Demand was quite variable and hard to predict. 

  Lake ships were primary competition. 

 

Page 30: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Few notation change from Appelbuam* 

∗ & ∗   Market demand: 

ln ln  Cost: 

  

1  

 

1 1  

                  MC  

Where  ∑ /   

Page 31: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Take logs Quasi Supply:  

ln ln 1 ln ln 1  

 Demand curve: 

ln ln   Reversionary period    

ln ln 1 ln ln 1  

 

ln ln ln 1 1 ln  

 ln β β ln  

Page 32: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Collusive period   

ln ln 1 ln ln 1  

 

ln ln ln 1 β ln ln 1 ln 1  

 

ln ln β ln ln  

  (Supply shifter is generated by regime shift)  

      

Page 33: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

   

SC

SR

D

Q

 

PC

PR

QRQC

Page 34: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

R:    ln β β ln   C:    ln ln β ln β       0  if reversionary period    Pr = 1 ‐  

    1  if collusive period      Pr =  

 Switching model:  

ln   

| 1 1 | 0                   SupplyC         SupplyR  

 

 

Page 35: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

   

Page 36: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Results  

All signs are as expected.  

Demand slopes down, supply up.   Lakes shifts (residual) demand of cartel down. 

  Note that R2 on demand is 0.31 

  Hard to predict. 

   

Page 37: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Results (cont...)  

Entry drives prices down in supply.  

The estimate of 3 is roughly consistent with Cournot behavior when collusion is taking place.  

The breakdown in collusion leads to significantly lower prices.  

Setting all variables equal to their sample mean (using the PN estimate), we get the numbers in Table 4. 

   

Page 38: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Table 4 Variable Lakes  0 1Price      PN  0 .1673 .1612  1 .2780 .2679  +66% Q      PN  0 38680 25904  1 25775 17261  ‐33%TR      PN  0 129423 83514  1 143309 92484  +11%PN: Estimated cheating dummy variable   

 

Page 39: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

PO often reflects a price war before PN, but they normally switch back to unity together.   

This is consistent with Green and Porter not picking up secret price cuts, so there is a lag in the PN estimate. 

  On average, non‐cooperative periods lasted about 10 weeks. 

  In this sample, price wars (using either PO or PN) were not preceded by adverse demand shocks. Normally incidents began after entry of another firm, though they were not immediate (average 40 week lag time). 

  This is consistent with theory, as the increase in number of participating firms leads to increased enforcement problems for the cartel. 

  Reversions also became more frequent as the number of firms increased. 

  Uses likelihood ratio tests to determine whether structural change has occurred in the industry, or if changes in price can be attributed to outside demand shocks.  

Page 40: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Tests the null hypothesis that the coefficient on It is equal to zero (no regime change). Uses a  distribution with 1 degree of freedom.  

Test‐statistic = 554.1 ‐ the null is overwhelmingly rejected!   

Main conclusion: Price and quantity changes cannot be attributed solely to exogenous changes in demand and structural conditions.   

JEC – led to the Sherman act    

Page 41: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Ellison (1994)—Rand   This is a refinement for Porter.  Focuses on empirical implication of trigger price strategies.  

Expanding output of i‐th firm.  For punishment j firms reduce prices.  If we cannot observe price change then we should see an expansion of the output. 

 This means 

High market share for one firm  Low market share for other firms  Low market share for one firm (other are cheating) 

 Two predictions 

Price wars should occur  Price wars should occur following demand signals used as strategies 

Page 42: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Model  Demand 

ln ln   

  

1AR 1  Supply 

ln ln β ln β β   New Eq (logit):  Probability of moving from one state to the next state.  

Pr 1| 1  

 

Page 43: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Wt     – constant     –  : allows for state dependence     –  : the trigger variables   Demand shocks:   and      unanticipated component increases price wars decrease.   Conclusion:  Green and Porter:  Price wars should occur 

This is true  But price is more alike a monopoly  Regime transition probabilities—provide evidence for the existence of the trigger strategies that are central to the theory. 

 Serially correlated demand shocks.  Therefore, such trigger effects can be distinguished from predictions of the RS model. 

Page 44: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

probability of price war should not be constant but depend on whether the previous period was cooperative.  Finds market power = 0.848, much closer to collusion   Probability of collusion this week is 0.975 if behavior was cooperative last week, but probability of collusion is 0.067 is there was a price war last week  Quite interestingly, concludes that actual cheating was rare, hence more evidence that Trigger strategies were followed as a way to sustain collusion     

Page 45: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Graddy  (1995) – Rand  Imperfect competition         Price discrimination  Old Fulton Fish Market:  1822‐2005 (now a restaurant)  

 

Page 46: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

   

   

Page 47: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

New Fulton Fish Market ($85 million facility) in the Bronx  

 

Page 48: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

   

Page 49: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Graddy collects data from a dealer  Fish: the common name whiting    

There will be no difference in the cost      Sales to consumers within a day  Large variations in the price across days    

$  .33 / lb May 1st $1.75 /lb May 8th 

 Systematic days of the week effects  Substantial variations in prices across customers in the same day      Are these price differences systematically related to customer characteristics?      

Page 50: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 Estimates  

, , , , . ,   i = transaction, t = day, and   = day dummy  Finds that prices charged for Asians are about 6‐7% lower!  Empirical model: Asian 

 

 Average price on the day at time t  White 

 

   are unobservable but know   for one firm.  Assume MC cost for all firms 

Page 51: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 MCiAt = MCAt  &  MCiWt = MCWt    For i      

 

ln1

ln ln  

 

ln ln1

ln  

 Same for    

ln ln1

⋯ 

 

ln ln1

⋯ 

  

Page 52: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 Quasi supply relationship 

1   

1   Note that  and   are elasticities.  Set    

11  

 Assume     

11  

 .598 .213  .753 .361  

Page 53: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

.388 .640  11

. 708

. 768 .922  

 This shows that    is about 8.5% above    

≅ 1.07 

 Sub into (A)  

11.07

1 .751 .6  

  .35 calculated & .388 estimated  

 IO data does not come easy!  Sometimes you may need to get your hands dirty!    Issues: ... 

Page 54: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Entry, Growth, and Exit  A good starting point is:  Dunne, Timothy, Roberts, Mark and Samuelson, Larry (1989) "The Growth and Failure of U.S. Manufacturing Plants," Quarterly Journal of Economics, November, 671‐698.   Bresnahan‐Reiss (1991) – JPE   They look at the increase in the number of firms in a market as market size increases.  The pattern of entry should tell us about how mark‐ups decrease as market size increases.  They  look at entry patterns across geographically differentiated markets—these are small markets all over US  (towns which are located far away from other towns in the US.) 

Page 55: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Industries:  

Dentists  Tires  Car Dealers  Plumbers 

  What can these entry data tell us about mark‐ups and competition?  Demand: 

, ∙   

,  = representative consumer &    = size of the market (number of cunsumers).  Z = demographic characteristics & P is price.  

Page 56: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Note: Consumers are separable—this demand function is good for a private goods market.  Firm cost:  Fixed cost:  F  (this is a sunk cost)  

,   

,   = firm output 

  Single firm case  

, , ∙   

percustomervariation 

Page 57: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

0 ⇒  

Entry if  0 

 As F increase or v decrease you need a larger market to sustain firms.   is the entry threshold for monopolist. 

 Duopoly case  Question is if you have two firms what level of s is need to support two firms.  

  

, , ∙     Price in duopoly but not 

specific about the nature of competition 

Difference in production costs for firms—think of big box stores! 

Implicitly assume that the 2 firms split the market —necessary assumption but ... 

Difference in fixed costs  

Page 58: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

   If  0 

2  

If  0 

2  

 is the duopoly entry threshold.      

Page 59: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

       

1

2

Page 60: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Few notes:    and   differ due to  

  MC may differ (b2 > 0)  Depend on price in monopoly vs. price in duopoly. 

 If no cost differences    B2 = b2 = 0  Perfect cartel case    

P2 = P1    

v2 = v1  

 

Page 61: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

*  Per firm level demand supports 2 firms or firms are identical.  Then get the 2nd firm when the market doubles.       

Page 62: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 *    B2 = b2 = 0 & perfect competition.  Slope gets flatter as P = MC                       

Page 63: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

*    B2 = b2 = 0 & Cournot competition.               

 

1

2

Page 64: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

B & R:  

∙ 1 

 If  0  then this ratio is about completion and/or MC change as N increase.  Per customer  fall as N increase and if no cost differences then it reflects a competitive effect.  More generally:  

∙  

    

Observable Unobservable

Page 65: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Empirical work:    202 geographic markets   Ordered probit  

The profits are the latent variable  

, , , , ,    

 

  

 

Page 66: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Y  =  Population variable (market size)  

Z, W  =  Market demand and cost shifters (County level variables)     WL  = fixed cost shifters (farm land prices)  Note that   can be estimated since this is interacted with pop variables.  This breaks the orthogonal condition.  This allow for separability.   

∑∑  

 

Page 67: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

      

Page 68: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

 

Page 69: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

These differences are due to entry cost, mark‐ups, and cost differences.  Key interpretation:     After the 1st and 2nd entry the mark‐up does not differ.     Few issues to think about:  

When looking at service industry like doctors it is hard to identify when they add a new doctor to staff 

     Market definition: people drive to find a doctor from one city to the other.        

Page 70: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Mazzeo (2002)—Rand  Use ‘Motel industry’ to extend B&R 1991 paper.  Extend the model by allowing firms to chose which type of firms they enters as.  Firms can differentiate between each other by choosing to enter in different areas of the product space (for example one might build a French restaurant if another decides to build an Italian restaurant.)  For this example Mazzeo chose high (h) and low (l) quality (i) Motels.    Market is defined as near highway exits.    

 

 

Page 71: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Questions:  

He wants to understand how firms decide the “location” of the products that they produce. This is driven in large part by the incentive to differentiate ‘my’ product from those of ‘my rivals.’  Allow for different types of producers to compete with each other: not just a homogenous good.  Profit in reduce form: 

, , ,  

 

    Demand characteristics that affect profits  

Competitor effects that influence profits  

Page 72: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Equilibrium concept:  Follow Sutton (1998) – Consider firms in each market to be playing a generic two‐stage game.   In the first stage (the "investment" stage), firms decide whether to enter and choose to offer either low‐quality or high‐quality services.  In the second stage (the "competition" stage) ensues in which payoffs are determined.   Assume that there are infinite number of potential identical entrants in each market (that is, for a given market structure, the same product‐type choice gives the same profits for every firm.)  In other words they get the same shocks (this means you don’t have to specify the number of potential entrants in a market).  Thus the equilibrium conditions in this market are:  Firms that are in the market make positive profits: 

Page 73: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

, , 0  

, , 0  If an additional firm entered, it would make negative profits:  

, 1, 0  

1, , 0  

To close the model, he needs to impose and additional assumption, that the effect of competition is higher for firms of your own type that the other type:   

, 1, , 1,   

1, , 1, ,  

Page 74: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Estimation:   Equilibrium is unique and therefore, the sum of the probabilities for all market configurations always equals one.   Maximum likelihood selects the profit function parameters that maximize the probability of the observed market configurations across the dataset.  The model is estimated using simulated maximum likelihood.   Industry and data set:  492 oligopoly markets  Motels began to prosper during the first half of the 20th century. 

Page 75: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 System of Interstate and Defense Highways, a 42,500‐mile network of freeways (1956‐present)  In the early years, most motel properties were independent‐often a single family designed, built, managed, and operated the motel.  Last several decades, more motel owners have affiliated themselves with regional and national franchises and chains.  Consumers were attracted to chain‐affiliated motels, known to have a consistent and predictable level of quality.  Generally motels provide the same basic services but differ in the level of service quality they have chosen to supply.  

Page 76: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

American Automobile Association (AAA) has established ratings systems to provide consumers with accurate information about the quality of motel services.  This is what Mazzeo use for quality index.  The nature of demand for highway motel services complicates the selection and definition of markets to analyze empirically.  Highway motels serve both visitors of residents and businesses in the town nearby each exit, as well as long‐distance travellers.  Observe geographically isolated clusters of motels along most interstates—this limits the extent to which motels at one exit compete with motels at other exits.  Also collect data only from small, rural exits.     

Page 77: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Page 78: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Page 79: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

   

Page 80: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva  

Page 81: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Page 82: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Page 83: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Results:  Operating a high‐quality motel is on average more profitable than operating a low‐quality motel ( CH = 2.5252 versus CL = 1.6254)  With no competitors, the payoffs to operating a low‐quality motel are on average higher (L = 1.6254+ ( ‐2.303) * (.2711) = 1.001) than those to operating a high‐quality motel (H = 2.5252 + ( ‐2.303) * (0.6768) = .9668)  Issues:  An issue with the Mazzeo (2002) model is that it is difficult to estimate the model when there are multiple “types” of firms, since the number of inequalities which need to be satisfied rises exponentially.     

Page 84: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Seim (2006)—Rand  To overcome the issue discussed in Mazzeo’s paper Seim introduces ‘private information’ into the model.  This means firms have information about the payoffs of entering a market that other firms do not see.   This is a very common strategy in empirical I.O—especially in auctions!  But ...  Private information shocks will lead to “mistakes” in the sense that two firms enter when only one firm would make positive profits. This could also happen in the perfect information case if firms are playing a mixed strategy equilibria.     

Page 85: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Questions:   How do firms choose their geographic locations: trade‐off between density of demand (lots of consumers) versus competition.  Choice of different type of modems technologies  The model  Simultaneous Moves.  Asymmetric Information.  Firm f’s payoff in location l in market m:  

Π Γ , n ε     

Market shock  Private shock 

Page 86: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Few more assumptions 

Assumption 2:   Γ , n ∑  

 

Assumption 3:   if ,  

Where  and  denotes cutoffs for the distance band.  Assumption 3 is okay—US Census tract level data. 

 

So the payoff function is  

Π  

Where   is the impact of competitors in distance band b.  

 

Page 87: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Each firm decides to locate in a location (l) that maximizes it’s payoff given its own type.  So expected profits in location l is: 

 

Π  

 Π  

Assume that the private information shocks   are distributed as i.i.d. logit draws.  

This results in a multinomial logit probabilities for firms’ beliefs conditional on entry.  

 

  

Page 88: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

So the entry probability with private information shocks distributed as i.i.d. logit draws is:  

Prexp ∑ exp Π , , ,

1 exp ∑ exp Π , , , 

 

 

Data: 

Video rental industry  Homogenous good.  Relatively inexpensive $2‐$4.  Store heterogeneity – inventory, rental period, drop‐off convenience, and mainly spatial location.   Average consumer travels about 3.2 miles to rent a movie! 

Page 89: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Medium size cities – 40 to 150K pop.  Include cities where the largest city outside of the market within 10 miles has a pop of less than 10K and the pop of the largest city within 20 miles does not exceed 25K people.   

Page 90: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

Page 91: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

  

 

Page 92: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

Figure 2 – Sample market – Greater Falls, MT

 

Page 93: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 

 

 

 

Page 94: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

  

Page 95: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

  

   

Page 96: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 Readings:  Bresnahan, T. F. and P. C. Reiss (1991). “Entry and Competition in Concentrated Markets.” Journal of Political Economy, 99(5): 33.  Dunne, Timothy, Roberts, Mark and Samuelson, Larry (1989) “The Growth and Failure of U.S. Manufacturing Plants,” Quarterly Journal of Economics, November, 671‐698.  Mazzeo, M. J. (2002). “Product choice and oligopoly market structure.” Rand Journal of Economics, 33(2): 221‐242.  Seim, K. (2006). “An Empirical Model of Firm Entry with Endogenous Product‐Type Choices.” RAND Journal of Economics, 37(3): 619‐640.    

Page 97: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

 Auctions:  Milgrom, Paul R.; Weber, Robert J., (1982) “A Theory of Auctions and Competitive Bidding” Econometrica, 50(5): 1089‐112 

De Silva, Dakshina, G., Timothy Dunne, Anuruddha Kankanamge, and Georgia Kosmopoulou. “The Impact of Public Information on Bidding in Highway Procurement Auctions” European Economic Review, 2008 (http://www.lancs.ac.uk/staff/desilvad/EER%20‐%202008%202065.pdf) 

 De Silva, Dakshina, G., Georgia Kosmopoulou and Carlos Lamarche “The Effect of Information on the Bidding and Survival of Entrants in Procurement Auctions,” Journal of Public Economics, 2009, 93(1‐2): 56‐72.  (http://www.lancs.ac.uk/staff/desilvad/JPubE%202008.pdf)  

Page 98: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

De Silva, Dakshina, G., Timothy Dunne, Georgia Kosmopoulou, and Carlos Lamarche “Disadvantaged Business Enterprise Goals in Government Procurement Contracting: An Analysis of Bidding Behavior and Costs”, forthcoming in International Journal of Industrial Organization. (http://www.lancs.ac.uk/staff/desilvad/IJIO%20‐%202012_DBE.pdf)  Guerre, E., Perrigne, I., Vuong, Q., 2000. “Optimal nonparametric estimation offirst‐price auctions.” Econometrica 68 (3), 525–574.  Haile, P.A., Hong, H., Shum, M., 2006. “Nonparametric Tests for Common Values in First Price Sealed‐bid Auctions,” Working Paper  R & D and Spillovers:  Abramovsky, L., Harrison, R., and Simpson, H. (2007). "University Research and the Location of Business R & D", The Economic Journal, 117: C114.C141.  

Page 99: Empirical Industrial Organization - lancaster.ac.uk notes 2013 U of P-1.pdf · IO‐De Silva Introduction A definition of industrial organization: “Industrial organization is concerned

IO‐De Silva 

De Silva and McComb. (2012) “Research Universities and Regional High‐Tech Firm Start‐up and Exit,” (with Robert McComb.)  Economic Inquiry, 50(1): 112–130.  George Deltas, Dakshina G. De Silva, and Robert McComb. “Agglomeration Spillovers and Industry Dynamics: Firm Entry, Growth, and Exit in the Software Publishing Industry,” (I’ll give a copy)