Profit maximization, industry structure & competition

Steve Keen (UWS)

Russell Standish (UNSW)
 From finance to economics?
• To date, econophysics has been the physics of finance
– Attractions:
• Fantastic data sets
• Market behavior clearly contrary to neoclassical beliefs
• Relatively easy application of existing toolkit
• A plea: please consider a physics of economics too!
– Attractions:
• Lousy data sets…
• Market behavior clearly contrary to neoclassical beliefs
• You have the toolkits that economists lack…
 “In the beginning was Marshall…”
• Example: “theory of the firm”
• “Baby” version of theory of firm “Marshallian”
– Firms are profit-maximizers
– Diminishing marginal productivity—rising marginal cost
– Maximize profits by equating “marginal revenue” (MR)
to “marginal cost” (MC)
– If industry “competitive”, constant MR=Market Price
• “Supply=Demand” optimal output level
– Marginal Benefit to society (Price=MB) equals MC
– If monopoly, face falling marginal revenue
• Sub-optimal output level
• MB exceeds MC
• Higher output would be socially beneficial…
 “And Marshall said, ‘let there be competition’”
• Monopoly vs perfect competition distinction “atomism”
– Individual firms so small that
• don’t react to what others do
• are “price-takers”:
– can produce any amount without affecting price
• So… q j  0
qi
dP  Q 
– Therefore… while for the market 0
dQ

dP  Q 
– for the individual firm 0
dqi

• Model forms base instruction of all economists

– Believed true by 90% of more of them…
 “A reading from the bible according to Gregory”

• This & next 4 slides from popular introductory economics

textbook, Mankiw, Principles of Economics…
– …
• Comparing Monopoly and Competition
– For a competitive firm, price equals marginal cost.
P = MR = MC
– For a monopoly firm, price exceeds marginal cost.
P > MR = MC
 Demand Curves for Competitive and Monopoly Firms...

(a) A Competitive Firm’s (b) A Monopolist’s

Demand Curve Demand Curve
Price Price

Demand

Demand

0 Quantity of 0 Quantity of
Output Output
The Efficient Level of Output...

Price
Marginal cost

Value Cost to
to monopolist
buyers

Value
Cost to to Demand
monopolist buyers (value to buyers)
0 Efficient Quantity
quantity

Value to buyers is greater Value to buyers is less

than cost to seller. than cost to seller.
 Harcourt, Inc. items and derived items copyright © 2001 by Harcourt, Inc.

Profit Maximization for the Competitive Firm...

Costs The firm maximizes

and profit by producing
Revenue the quantity at
which marginal cost MC
equals marginal
revenue.
MC2
ATC
P=MR1 P = AR = MR
AVC

MC1

0 Q1 QMAX Q2 Quantity
 Harcourt, Inc. items and derived items copyright © 2001 by Harcourt, Inc.

The Inefficiency of Monopoly...

Price
Deadweight Marginal cost
loss

Monopoly
price

Marginal
revenue Demand

0 Monopoly Efficient Quantity

quantity quantity
 “Then there was lousy mathematics…”
• “Marshallian” case mathematical nonsense
dP dP
– If atomism applies, then dq  dQ
i
dP dP dQ

dqi dQ dq i
dP  n q j 
  
dQ  j 1 qi 
n q
dP  qi j 
   
dQ  qi j  i qi 
dP  n
 dP
   
1  0
dQ  j i  dQ
• In the literature for
38 years, but ignored!:
 “And Cournot said, ‘let there be strategy’”

• “Grown up” theory of firm is Cournot-Nash game theory

q j
• implicitly sets 0
qi
n q
qi
• Feasible that  j 0
qi j  i qi
• Standard CN analysis game theoretic
– Either “cooperate” or “defect”
q j
• In our terms, discrete values for
qi
q j
• Our innovation: consider variable qi
  j ,i

– Reaction of jth firm to output change by ith

• What is optimal value?

 “And we said, ‘how much?’”
• Profit for ith firm is P  Q  qi  TC  qi 

• Optimal value is where total derivative is zero:

d d
dQ
  qi  
dQ
 P  Q  qi  TC  qi    0
n     n   dq j 
• Expanding, this is  

 P  q j  qi  TC  qi   0

j 1  q j   j 1   dQ 

• In terms of ij, for the ith firm, this is:


n n
 dP n n n
P    j , i  k , j   qi   k , j  MC  qi   j , i  0
j 1  k 1  dQ j 1 k 1 j 1

• Can now compare Marshallian & Cournot analysis

– Marshallian:  j , i  0, i  j
dP
– Substitute: formula reduces to P  nqi  MC  qi   0
dQ
 “Verily, Marshall be bollocks then?”
• Neoclassical “profit-maximisation” rule false
dP
P  nqi  MC  qi   0 can be rearranged to:
dQ
n 1
MR  qi   MC  qi  
n
 P  MC  qi  

• Neoclassical “rule” only maximises profit for n=1

• Multi-firm industry, profit maximisation is MR>MC
• What about when ij non-zero?
– What is optimal value of ij ?
qi
– Consider heuristic case   i  j 1
q j
qi
qi

• Then profit maximum is  n  1 P  P  nqi dP  MC  qi 

dQ

• What is optimal value of ?

d
• Occurs where d   qi   0
 “Do not lead us into the valley of profitless”
d 1 d  dP 
• Optimal  value is zero:   qi   Q  P  nqi  MC  qi  
d n d  dQ 
• Illustration:
– Linear demand curve P  Q   a  bQ
• Constant marginal cost c, fixed cost k
• Profit-maximising output for ith firm as function of
 and n is q   , n     n  1   1 a  c
nb   n  1   2 
  
  n  1   1 a  c  a  c     n  1   1 a  c  
2

• Per firm profit is  max   , n    k

nb   n  1   2   nb  n  1   2 2
  
 

• Maximum value at =0

– Example: a=800, b=1/10,000,000, k=1,000,000 c=100,
20 firm industry…
 “Do not worship false cows before me”
• Cournot-Nash recommended level of strategic interaction
generates 1/5th profit level of no interaction at all
Profit (LHS) and quantit y (RHS) as function of theta for 20 firm indust ry
8
Ratio of maximum equilibrium firm profit
4 10
8 10
10 300
Profit ratio (LHS)
8
3.5 10
7 10
10 Theta value (RHS)

Ratio of Keen to Cournot per firm profits

250

Cournot recommended Theta value

10
3 10
8 0.8
6 10

Profit maximizing quantity

8
2.5 10
Maximum Profit

5 10
10 200

8
0.6
10 2 10
4 10

150
8
10 1.5 10
3 10
0.4
8
2 10
10 1 10 100

7
1 10
10 5 10
0.2
50
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Interaction parameter theta

Maximum profit as function of theta 0 0
100 200 300 400 500 600 700 800 900 1000
Cournot Profit
Profit maximizing quantity as function of theta
Cournot quantity Number of firms in industry

• Cost of strategic interaction rises with n:

• No interaction 300 times as profitable as Cournot
interaction for 1,000 firms…
• What are real firms likely to do?
 From religion to evolution…
• Consider
F  q .K( 1  a  b  C D  E)
0 0

population of
F  q .K( 1  a  b  C D  E)
1 0

instrumental
for i  firms.min firms.min  firms.st ep s  firms.max

for j  0  rand  1
profit Seed ( j  1)

maximisers  
Q  round runif i  q .K( i  a  b  C D  E)  q .C( i  a  b  C D  E)
0 
– Choose p  P
0

 Q  a  b 
0

initial output dq  500 00sign
 ( runif( i 1  1) )

– Change output for k  1  runs

+/- Q Q
k k 1
 dq

p  P  Q  a  b 
– If profit
k k
 
 
rises, repeat     k    
dq  sign  p  Q  p  Q  t c Q  i C D  E k  t c Q  i C D  E k   dq
 k k k 1 k 1 k 1 

– If falls, Qk  Qk1  Qk2  Qk3

Q.end 
change sign…
j 4
F
i 0  
 mean Q.end

F  stdev  Q.end
i 1
Q.end  0

F
 From reductionism to complexity
• Simple agent design generates striking behavioral
complexity…
Firm Outp uts, 100 firms: 3 samp les
7
8 10

7
7 10

7
6 10

7
5 10

7
4 10

7
3 10

7
2 10
0 50 100 150 200 250 300 350 400 450

Iteratio ns
Cournot
Keen
Mean
Firm 1
Firm 2
Firm 3
 Zero interaction rules…
• Agents converge to Keen equilibrium, not Cournot…
• But curious result
Market Output: Rising Marginal Cost
9
5 10
Cournot

applies when
Keen
9
4.5 10
Average Final Market Output

Average of Simulations
+/- 2 St. Dev.
4 10
9
fixed step size
9
3.5 10 replaced by
3 10
9
normally
9
2.5 10
50 100 150 200 250
distributed one…
300 350 400 450 500

Number of Firms
  q.C(i a  b  C D E)  
Market Output: Constant Marginal Cost dq  round rnorm i 0 
3.6 10
9

Keen
  10 
Average of Simulations
Average Final Market Output

+/- 2 St. Dev.

9
3.55 10

9
3.5 10

9
3.45 10
50 100 150 200 250 300 350 400 450 500

Number of Firms
 Zero interaction rules … sort of
• Convergence to Keen equilibrium not as tight for fixed
marginal cost…
Market Output: Constant Marginal Cost
9
7 10
Cournot
Keen
9
6 10
Average Final Market Output

Average of Simulations
+/- 2 St. Dev.
9
5 10

9
4 10

9
3 10

9
2 10
50 100 150 200 250 300 350 400 450 500

Number of Firms

• But no effect of varying industry structure on result

• However for rising marginal cost…
 Cournot rises from the dead???
• Apparent trend for convergence to Cournot output level
as n rises for normally distributed dq…
Market Output: Rising Marginal Cost
9
6 10
Cournot
Keen
Average Final Market Output

9 Average of Simulations
5 10
+/- 2 St. Dev.

9
4 10

9
3 10

9
2 10
50 100 150 200 250 300 350 400 450 500

Number of Firms

• But looking more closely, consider:

– output for fixed industry structure (value of n)
– rising divergence of dq around mean of zero
 Testing divergence
• Program
F  q .K( 1  a  b  C D  E)
0 0
F  q .K( 1  a  b  C D  E)
iterates over 1 0
for i  0  dispersal .steps  dispersal  1

standard for j  0  rand  1

deviation of Seed ( j  1)
 
Q  round runif firms  q .K( firms  a  b  C D  E)  q .C( firms  a  b  C D  E) 
dq from 1%
0

p  P  Q  a  b 
to dispersal
0 0
 
dq  round rnormfirms  0  
1  i 
% of Cournot  
  q.C( firms  a  b  C D  E)
 100  

firm output for k  1  runs

Q Q  dq
level
k k 1

p  P
k

 Q  a  b 
k


    k    
dq  sign  p  Q  p  Q  tc Q  firms  C D  E k  tc Q  firms  C D  E k   dq
 k k k 1 k 1 k 1 

Qk  Qk1  Qk2  Qk3

Q.end 
j 4
F
i 0  
 mean Q.end

F  stdev  Q.end 
i 1
Q.end  0

F
 “Nope, he’s still dead!”
• Convergence to Cournot a function not of number of
firms, but of dispersal of dq!
Market Output vs Dispersal (Rising Marginal Cost)
9
6 10
Cournot
Keen
Average Final Market Output

9 Average of Simulations
5 10
+/- 2 St. Dev.

9
4 10

9
3 10

9
2 10
0 2 4 6 8 10 12 14 16 18 20

Per cent of qC

• Conjecture: result due to stochastic differential effects

– ij a stochastic variable…
 Goodbye to the ‘totem of the econ’
• Neoclassical 1200
Market Functions, Predictions, Outcomes: 50 firms

religion teaches Price Function

Demand, Marginal Revenue, Marginal Cost

Marginal Revenue
“perfect 1000
Marginal Cost

competition good,
Cournot 50 Firms
800 Keen
monopoly bad” 1% dispersal
10% dispersal
– But Marshallian
600
20% dispersal

theory bollocks 400

– Cournot results
depend on
200

dispersal, not n… 0
9 9 9 9 10
0 2 10 4 10 6 10 8 10 1 10

Market Quantity

• Worse still, neoclassical assumptions counter-factual

• Not simplifications of reality, but contradictions of it…
 “Just the facts, Ma’am”
• Diminishing marginal productivity/Rising marginal cost?
– “The overwhelmingly bad news here (for economic
theory) is that, apparently, only 11 percent of GDP is
produced under conditions of rising marginal cost.
Almost half is produced under constant MC… But that
leaves a stunning 40 percent of GDP in firms that
report declining MC functions” (Blinder 1998: 102)
• “Applied” economic concepts like Elasticity of demand?
– “Can it really be true that firms that sell 40 percent
of GDP believe that their demand is totally insensitive
to price, and that only about one-sixth of GDP is sold
under conditions of elastic demand?” ([Blinder]: 101)
 Blinder’s economic facts of the firm
Summary of Selected Factual Results
• Economic facts of the firm conflict Price Policy

strongly with assumptions of

Median number of price changes in a year 1.4
Mean lag before adjusting price months following
(neoclassical) economics Demand Increase
Demand Decrease
2.9
2.9
– Infrequent price adjustments Cost Increase
Cost Decrease
2.8
3.3
– Fixed price contracts common Percent of firms which
Report annual price reviews 45
– Most sales to other businesses, Change prices all at once
Change prices in small steps
74
16
not “utility maximizing” consumers Have nontrivial costs of adjusting prices of 43
which related primarily to

– Fixed costs very important, large

the frequency of price changes 69
the size of price changes 14
percentage of product costs Sales
Estimated percent of GDP sold under contracts

– Marginal costs fall for most which fix prices

Percent of firms which report implicit contracts
28
65
businesses, not rise Percent of sales which are made to
Consumers 21
• Neoclassical theory irredeemably bad Businesses 70
Other (principally government) 9
• Internally flawed Regular customers 85
Percent of firms whose sales are
• Factually irrelevant Relatively sensitive to the state of the economy 43
Relatively Insensitive to the state of the economy 39
• We need a new “theory of the firm” Costs
Percent of firms which can estimate costs at least
moderately well 87
Mean percentage of costs which are fixed 44
Percentage of firms for which marginal costs are
Increasing 11
Constant 48
Decreasing 41
 Econophysicists to the rescue?
• Neoclassical economics a dead end
• Alternative evolutionary theory of firm needed
– Schumpeter’s “creative destruction” best verbal model
– Economists not up to the task of formalising it
• ‘great expectations … have only become substantiated
slowly and partially… First, … there are significant
barriers to entry for students and researchers who want
to explore and extend evolutionary models by simulation;
you simply have to master a great many skills to be able
to combine evolutionary theorising and simulation in a
fruitful way.
way Second, there is a lack of cumulativeness …
many entrants to the field seem to build their efforts
from scratch.’ (Anderson & Valente: 44)
 Econophysicists to the rescue?
• Physics skill set much more up to the job than economics
• Please… having reinvented finance…

• And help move economics

– from a religion
– to a science…