Constrained Uni /Multi-
Objective Decision Making in
Economics
Topics
A Consumer’s choice Problem (Utility Maximization)
Life’s Optimization Problem
Air Travel choice Problem
Efficiency vs Equity
Consumer Behaviour
A Consumer’s choice Problem
Two consumable items: Banana & Mango
Price of 1 Banana = Rs. 10
Price of 1 Mango = Rs. 15
Consumer has Rs. 120
How should he/she allocate money between Mangoes & Bananas?
Depends on his taste…what he loves…& How much…Right??
The consumers taste is represented by a Utility Function U: 𝑅2 → 𝑅
max 𝑈 (𝑀, 𝐵) : subject to the constraint 15M + 10B ≤ 120
𝑀,𝐵
Consumer’s Optimization Problem
Decision variables: M ∈ 𝑅+ & B ∈ 𝑅+
Objective function: U(M,B)
Constraint: 15M + 10B ≤ 120
Budget Set
B
(12,0)
Budget Line:
15M + 10B = 120
(8,0) M
All (M,B) ∈ are affordable bundles.
Example
U(M,B) = M.B
max 𝑈 𝑀, 𝐵 = 𝑀. 𝐵 : subject to the constraint 15M + 10B ≤ 120
𝑀,𝐵
i.e max 𝑈 𝑀, 𝐵 = 𝑀. 𝐵 such that (M,B) ∈
𝑀,𝐵
Indifference curves: U(M,B) = M.B
B
. (3,5)
. (1,4) .(5,3)
. (2,2) 𝑈2 = 15
𝑈1 = 4
M
Optimal Choice / Optimal Bundle
Optimization Problem - Revised
max 𝑈 𝑀, 𝐵 = 𝑀. 𝐵 : subject to the constraint 15M + 10B = 120
𝑀,𝐵
≡ max 𝑈 𝑀, 𝐵 = 𝑀. 𝐵 : subject to the constraint B = (24 – 3M)/2
𝑀,𝐵
≡ max M(24 – 3M)
𝑀
𝑀∗ = 4 & therefore 𝐵∗ = 6
(𝑀∗ , 𝐵∗ ) = (4,6) is the optimal bundle.
Practice Problems
U(M,B) = 𝑀2 + 𝐵2
U(M,B) = min {M,B}. (e.g left shoe & right shoe)
U(M,B) =
The Problem
Suppose you will be alive for T+1 years (starting from your Grad. Day)
Let’s say you know that in year ‘n’ you will make income = 𝑦𝑛
Each year you can save = 𝑠𝑛 & consume = 𝑐𝑛
Interest rate = r (remains constant throughout life)
Consumption in youth “feels better” than in Old Age.
World with 100 % inheritance tax … No Dad’s money !!!!
max σ𝑇𝑛=0[𝑓 𝑛 . (𝑙𝑜𝑔𝑒 𝑐𝑛 )] ; 𝑓 𝑛 is the depreciation
𝑐0 ,𝑐1 ,…..𝑐𝑇
function. 𝑓 𝑛 is decreasing in ‘n’
The Periodic Budget Constraints
𝑐0 = 𝑦0 - 𝑠0
𝑐1 = 𝑦1 + 𝑠0 1 + 𝑟 - 𝑠1
𝑐2 = 𝑦2 + 𝑠1 1 + 𝑟 - 𝑠2
𝑐3 = 𝑦3 + 𝑠2 1 + 𝑟 - 𝑠3
…….
𝑐𝑇−1 = 𝑦𝑇−1 + 𝑠𝑇−2 1 + 𝑟 - 𝑠𝑇−1
𝑐𝑇 = 𝑦𝑇 + 𝑠𝑇−1 1 + 𝑟
The Periodic Budget Constraints
𝑐0 = 𝑦0 - 𝑠0 : 𝑐0 = 𝑦0 - 𝑠0
𝑐1 𝑦1 𝑠1
𝑐1 = 𝑦1 + 𝑠0 1 + 𝑟 - 𝑠1 : = + 𝑠0 -
1+𝑟 1+𝑟 1+𝑟
𝑐2 𝑦2 𝑠1 𝑠2
𝑐2 = 𝑦2 + 𝑠1 1 + 𝑟 - 𝑠2 : = + -
(1+𝑟)2 (1+𝑟)2 1+𝑟 (1+𝑟)2
𝑐3 𝑦3 𝑠2 𝑠3
𝑐3 = 𝑦3 + 𝑠2 1 + 𝑟 - 𝑠3 : = + -
(1+𝑟)3 (1+𝑟)3 (1+𝑟)2 (1+𝑟)3
……. ……….
𝑐𝑇−1 𝑦𝑇−1 𝑠𝑇−2 𝑠𝑇−1
𝑐𝑇−1 = 𝑦𝑇−1 + 𝑠𝑇−2 1 + 𝑟 - 𝑠𝑇−1 : = + -
(1+𝑟)𝑇−1 (1+𝑟)𝑇−1 (1+𝑟)𝑇−2 (1+𝑟)𝑇−1
𝑐𝑇 𝑦𝑇 𝑠𝑇−1
𝑐𝑇 = 𝑦𝑇 + 𝑠𝑇−1 1 + 𝑟 : = +
(1+𝑟)𝑇 (1+𝑟)𝑇 (1+𝑟)𝑇−1
The Consolidated Budget Constraint
𝑐1 𝑐2 𝑐3 𝑐𝑇−1 𝑐𝑇
𝑐0 + + + …… + +
1+𝑟 (1+𝑟)2 (1+𝑟)3 (1+𝑟)𝑇−1 (1+𝑟)𝑇
𝑦1 𝑦2 𝑦3 𝑦𝑇−1 𝑦𝑇
= 𝑦0 + + + …… + +
1+𝑟 (1+𝑟)2 (1+𝑟)3 (1+𝑟)𝑇−1 (1+𝑟)𝑇
Life’s Optimization Problem
max σ𝑇𝑛=0[𝑓 𝑛 . (𝑙𝑜𝑔𝑒 𝑐𝑛 )]
𝑐0 ,𝑐1 ,…..𝑐𝑇
Subject to the constraints:
𝑇 𝑐𝑛 𝑇 𝑦𝑛
σ𝑛=0 = σ𝑛=0
(1+𝑟)𝑛 (1+𝑟)𝑛
𝑐𝑛 ≥ 0 ∀ n = 0,1,2,……T
Let’s get real !!
You don’t know what my lifetime income stream {𝑦0 , 𝑦1 , 𝑦2 …… 𝑦𝑇 }
What is the depreciation function 𝑓 𝑛 for you??
Will the interest rate remain same all the while?
Won’t you invest in other asset classes?
Depreciation function
Typically 𝑓 𝑛 = 𝛽 𝑛 ∀ n = 0,1…T & 𝛽 ∈ (0,1)
Lower the 𝛽 lesser value you assign to future: MYOPIC
Higher the 𝛽 higher value you assign to the future: FAR - SIGHTED
Of course you can have other functional forms too for 𝑓 𝑛
Depends on your personality type !!
Income Estimation
Let’s say life starts at 35
From your grad day to the time you turn 35 let’s say your annual incomes are given
by {𝑦0 , 𝑦1 , 𝑦2 …… 𝑦14 }
Fit the model on the data: {𝑦0 , 𝑦1 , 𝑦2 …… 𝑦14 }: * Use MLE.
Get estimates: 𝑦ෞ
15 , 𝑦ෞ 𝑦𝑇
16 …….ෞ
Define 𝑦
෦𝑛 = 𝑦ෟ
𝑛+15 (* 𝑦
෦0 = 𝑦ෞ Ӗ Ӗ
15 + 𝑆 ; 𝑆 = savings till age 35)
So your annual income flow from age 35 is given by:
′
𝑦0 , 𝑦
{෦ ෦1 ,…… 𝑦෦
𝑇 ′ }; 𝑇 = T - 15
Sensex
Nominal Int. rate (savings)
NIFTY 50
Investment Returns
CAGR of Sensex (1979 – 2019) -> 16.1%
CAGR of NIFTY 50 (2001 – 2021) -> 14%
Avg. interest rate (going forward) -> 6%
Constant Portfolio: 𝛾1 % in Sensex fund
𝛾2 % in NIFTY 50
( 1- 𝛾1 − 𝛾2 ) % in FD / savings A/c
r = 16.1 𝛾1 + 14 𝛾2 + 6 ( 1- 𝛾1 − 𝛾2 )
Life’s Optimization Problem (at 35 !)
𝑇 ′
max σ𝑛=0[𝑓 𝑛 . (𝑙𝑜𝑔𝑒 𝑐𝑛 )]
𝑐0 ,𝑐1 ,…..𝑐𝑇′
Subject to the constraints:
𝑇 ′ 𝑐𝑛 𝑇 ′ ෦𝑛
𝑦
σ𝑛=0 = σ𝑛=0
(1+𝑟)𝑛 (1+𝑟)𝑛
𝑐𝑛 ≥ 0 ∀ n = 0,1,2,…… 𝑇 ′
Solving the optimization problem yields optimal periodic
consumptions (𝑐0 ∗ , 𝑐1 ∗ , 𝑐2 ∗ ……. 𝑐 𝑇 ′ ∗ )
𝑇 ′
σ𝑛=0[𝑓 𝑛 . (𝑙𝑜𝑔𝑒 𝑐𝑛 ∗ ) = 𝑉 ∗ (The optimal utility)
𝑉 ∗ (𝛾1 , 𝛾2 , 𝑇 ′ )
Given that you know 𝑇 ′ , maximize 𝑉 ∗ (𝛾1 , 𝛾2 , 𝑇 ′ ) w.r.t 𝛾1 & 𝛾2
Let’s Fly !! (Multi – Objective Choice)
Kolkata –> Delhi
AIRLINE TIME (Hrs) COST (Rs. ‘000)
INDIGO 2 8
SPICE JET 3 6
Go AIR 3 8
AIR INDIA 4 5
VISTARA 4 7
AIR ASIA 5 5
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Optimization Problem
Decision Variable: Airline ∈ {Indigo , SpiceJet, GoAir, Air India, Vistara, Air
Asia, Jet Airways, Akasa, Kingfisher}
Here the consumer / buyer has two objectives.
i. Minimize cost
ii. Minimize Time (of travel)
Let’s assume for the time being that your bank a/c balance > Rs. 8000
i.e no budget constraint.
Let’s NOT be DUMB !
Indigo vs GoAir
AIRLINE TIME (Hrs) COST (Rs. ‘000)
INDIGO 2 8
SPICE JET 3 6
Go AIR 3 8
AIR INDIA 4 5
VISTARA 4 7
AIR ASIA 5 4
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Air India vs Vistara
AIRLINE TIME (Hrs) COST (Rs. ‘000)
INDIGO 2 8
SPICE JET 3 6
Go AIR 3 8
AIR INDIA 4 5
VISTARA 4 7
AIR ASIA 5 4
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Air Asia vs {Jet Airways, Akasa, Kingfisher}
AIRLINE TIME (Hrs) COST (Rs. ‘000)
INDIGO 2 8
SPICE JET 3 6
Go AIR 3 8
AIR INDIA 4 5
VISTARA 4 7
AIR ASIA 5 4
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Re-Inspecting the “Smart” Choices
AIRLINE TIME (Hrs) COST (Rs. ‘000)
INDIGO 2 8
SPICE JET 3 6
AIR INDIA 4 5
AIR ASIA 5 4
Pareto Optimal Set
{INDIGO , SPICEJET , AIR INDIA , AIR ASIA} form a pareto optimal set.
In this set moving from one Airline to another will NOT lead to an
improvement in one objective function without incurring a loss in the
other objective function.
Pareto Optimal Set is also called Dominant Set of Rank -1.
Re-Inspecting the “DUMB” choices
AIRLINE TIME (Hrs) COST (Rs. ‘000)
Go AIR 3 8
VISTARA 4 7
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Vistara vs {Akasa, Kingfisher}
AIRLINE TIME (Hrs) COST (Rs. ‘000)
Go AIR 3 8
VISTARA 4 7
JET AIRWAYS 5 6
AKASA 5 7
KINGFISHER 6 7
Dominant Set of Rank-2
{Go AIR, VISTARA, JET AIRWAYS} form a dominant set of Rank -2
With in this set moving from one Airline to another will NOT lead to
an improvement in one objective function without incurring a loss in
the other objective function.
{Akasa, Kingfisher}
Akasa dominates Kingfisher.
Dominant set of Rank -3 = {Akasa}
Dominant set of Rank -4 = {Kingfisher}
The Hierarchy of Choice
DSR-1 -> “SMART” (pareto optimal)
DSR-2 -> “DUMB”
DSR - 3 -> “DUMBER”
DSR – 4 -> “DUMBEST”
Constraints
The consumer might face two kinds of constraints:
i. Budget constraint
ii. Time constraint
A consumer might have either one of them or both.
the constraints determine the set of feasible alternatives.
Constraint Example 1: Budget = Rs. 6500
The feasible set of alternatives are:
AIRLINE TIME (Hrs) COST (Rs. ‘000)
SPICE JET 3 6
AIR INDIA 4 5
AIR ASIA 5 4
JET AIRWAYS 5 6
The Pareto Optimal (DSR -1) set: {SPICE JET, AIR INDIA, AIR ASIA}
DSR -2: { JET AIRWAYS}
Movie Time !!
Two Competing Theatres
Two Movie Theatres (of same quality) in a neighbourhood.
Total market size (per annum) = Rs. 1 Cr.
Marketing budget of theatre 1 = Rs. 1 L & that of Theatre – 2 is Rs 2 L.
Then Theatre 1 grabs 1/3 of the market size & Theatre 2 grabs 2/3.
Market share is proportional to the marketing budget.
If the marketing budget of firm j is = 𝑋𝑗 where j ∈ 1,2
Then the market share of firm j is given by
𝑋𝑗
𝜋𝑗 = 𝑋1 +𝑋2
if 𝑋𝑗 > 0 & 𝑋𝑖 > 0
= 1 if 𝑋𝑗 > 0 & 𝑋𝑖 = 0
= 0 if 𝑋𝑗 = 0 & 𝑋𝑖 > 0
𝜋𝑖 = 𝜋𝑗 = 0 if 𝑋𝑗 = 0 & 𝑋𝑖 = 0
Herfindahl Hirschman Index (HHI) – Inequality
Measure
Given that the total market size is = 1, if firm j ∈ 1,2, … . 𝑁 grabs a
market share = 𝜋𝑗 then HHI = σ𝑁 𝜋
𝑗=1 𝑗 .
2
Clearly HHI is minimum if 𝜋𝑗 = 1/N (all firms grab equal share)
HHI is maximum (=1) if 𝜋𝑗 = 1 for some j (monopoly)
In the example: HHI = (1/3)2 + (2/3)2 = 5/9
A Network of Contests
Graphical Representation
Objective of every Theatre (Agent)
Maximize their own market share – the “prize”
They allocate – the marketing budget.
Distribution mechanism – Tullock Contest (1980)
Model
S: set of agents (firms), who are engaged in a network of conflicts (for market share).
Model
S: set of agents (firms), who are engaged in a network of conflicts (for market share).
This is represented by a weighted undirected graph G = <V,E,W> , where V, the set of vertices denote
the agents (firms) and the edge 𝑒𝑖𝑗 ∈ 𝐸 denotes that agent ‘i’ is contesting with agent j.
Model
S: set of agents (firms), who are engaged in a network of conflicts (for market share).
This is represented by a weighted undirected graph G = <V,E,W> , where V, the set of vertices denote
the agents (firms) and the edge 𝑒𝑖𝑗 ∈ 𝐸 denotes that agent ‘i’ is contesting with agent j.
W(𝑒𝑖𝑗 ) = 𝑤𝑖𝑗 i.e the weight on the edge connecting agent i & j signifies the common valuation of the
prize which the two agents are fighting over. 𝑤𝑖𝑗 is the market size firms i & j are contesting over.
Of course W(𝑒𝑖𝑗 ) = 0 if 𝑒𝑖𝑗 ∉ 𝐸.
Model
S: set of agents (firms), who are engaged in a network of conflicts (for market share).
This is represented by a weighted undirected graph G = <V,E,W> , where V, the set of vertices denote
the agents (firms) and the edge 𝑒𝑖𝑗 ∈ 𝐸 denotes that agent ‘i’ is contesting with agent j.
W(𝑒𝑖𝑗 ) = 𝑤𝑖𝑗 i.e the weight on the edge connecting agent i & j signifies the common valuation of the
prize which the two agents are fighting over. 𝑤𝑖𝑗 is the market size firms i & j are contesting over.
Of course W(𝑒𝑖𝑗 ) = 0 if 𝑒𝑖𝑗 ∉ 𝐸.
Also each agent i ∈ 𝑆 has “arms” endowment 𝑋ഥ𝑖 (marketing budget of firm i).
Theatre j’s Optimization Problem
Theatre ‘j of course wants to maximize his total market share.
Let 𝑋𝑗𝑘 is the marketing allocation by Theatre ‘j’ against theatre ‘k’
𝑋𝑗𝑘
max σ𝑘∈𝑁(𝑗) . 𝑊𝑗𝑘
{𝑋𝑗𝑘 }𝑘∈𝑁(𝑗) 𝑋𝑗𝑘 +𝑋𝑘𝑗
s.t σ𝑘∈𝑁(𝑗) 𝑋𝑗𝑘 = 𝑋ഥ𝑗
The above optimization problem is solved simultaneously by all j ∈ V
Solving the optimization problem will yield the optimal marketing
∗
allocations in all markets {𝑋𝑗𝑘 }𝑘∈𝑁(𝑗) ∀ j ∈ V
Market Shares
∗
𝑋𝑗𝑘
The market share of theatre ‘j’ , 𝜋𝑗 ∗ = σ𝑘∈𝑁(𝑗)[ ∗ +𝑋 ∗ ]. 𝑊𝑗𝑘
𝑋𝑗𝑘 𝑘𝑗
∗ ∗ 2
The resulting 𝐻𝐻𝐼 = σ𝑗∈𝑉 [𝜋𝑗 ]
Clearly 𝐻𝐻𝐼 ∗ is a function of the marketing budget vector (𝑋ഥ𝑖 )𝑖∈𝑉
The Policy Maker’s Problem
How to allocate (or re-allocate with taxes / subsidies) the marketing
budgets.
Objective: Minimize HHI
