Value Creation & Complementarity

Formalizing Synergistic Value in Strategic Coopetition (TR-1)

This document provides a comprehensive treatment of value creation and complementarity from Technical Report 1 (TR-2025-01), explaining how cooperative surplus emerges and is modeled mathematically.

Executive Summary

For Practitioners: Complementarity explains why cooperation creates value, joint action produces more than the sum of independent efforts. When Samsung and Sony combined manufacturing expertise with brand strength, they created value neither could achieve alone.

For Researchers: We formalize complementarity through value creation functions V(a γ) exhibiting superadditivity. Two specifications (logarithmic, power) are validated, with logarithmic achieving 58/60 accuracy on the S-LCD case study. The complementarity parameter γ controls synergy strength.

Conceptual Foundation

The Value Creation Problem

In coopetition, actors face a fundamental tension:

This is the essence of Brandenburger and Nalebuff’s coopetition framework: actors must balance value creation incentives against value appropriation incentives.

What is Complementarity?

Definition: Complementarity exists when joint action creates superadditive value, the whole exceeds the sum of the parts.

Mathematically:

\[\Large V(\{i, j\}) > V(\{i\}) + V(\{j\})\]

The value created by actors $i$ and $j$ working together exceeds what each could create independently.

Sources of Complementarity

Source Mechanism Example
Resource Combination Heterogeneous assets synergize Manufacturing + Brand
Knowledge Spillovers Learning from partner Technology transfer
Network Effects Combined networks exceed sum User base combination
Risk Sharing Diversification benefits Joint investment
Economies of Scale Combined volume reduces cost Joint purchasing

The Added Value Concept

Following Brandenburger and Nalebuff, we define an actor’s Added Value as:

\[\Large \text{Added Value}_i = V(\text{all actors}) - V(\text{all actors except } i)\]

Actors with high added value have strong bargaining positions because the coalition loses significant value without them. Complementarity increases added value for all participants.

Mathematical Formalization

The Value Creation Function

Equation 2 (TR-1): Total value created by joint action:

\[\Large V(\mathbf{a} \mid \gamma) = \sum_{i=1}^{N} f_i(a_i) + \gamma \cdot g(a_1, \ldots, a_N)\]

Components:

Component Symbol Meaning
Individual Value $f_i(a_i)$ Value actor $i$ creates independently
Synergy Function $g(a_1,\ldots,a_N)$ Value existing only through collaboration
Complementarity γ ∈ [0, 1] Strength of synergistic effects

Individual Value Functions

Individual value represents what each actor contributes independently of collaboration. Two specifications are validated:

Equation 6 (TR-1):

\[\Large f_i(a_i) = \theta \cdot \ln(1 + a_i) \quad \text{where } \theta = 20.0\]

Properties:

Value Function

When to Use: Manufacturing partnerships, technology joint ventures, scenarios where baseline capabilities are highly valuable but incremental improvements have declining impact.

Power Specification (Alternative)

Equation 3 (TR-1):

\[\Large f_i(a_i) = a_i^{\beta} \quad \text{where } \beta = 0.75\]

Properties:

When to Use: General scenarios, platform ecosystems, academic baselines.

Synergy Function

The synergy function captures value that exists only through collaboration, it requires multiple actors contributing.

Equation 4 (TR-1): Geometric Mean

\[\Large g(a_1, \ldots, a_N) = \left(\prod_{i=1}^{N} a_i\right)^{1/N}\]

Synergy Heatmap

For Two Actors:

\[\Large g(a_1, a_2) = \sqrt{a_1 \cdot a_2}\]

Properties:

Property Description Implication
Symmetric Order doesn’t matter Fair contribution accounting
Zero-requiring If any $a_i = 0$, $g = 0$ All must contribute
Balance-rewarding Maximized when $a_i$ equal Discourages free-riding
Smooth Continuous and differentiable Tractable optimization

Why Geometric Mean?

The geometric mean captures the intuition that:

  1. Everyone must contribute: A single defector ($a_i = 0$) destroys all synergy
  2. Balance matters: 50-50 contribution creates more synergy than 90-10
  3. Scale invariance: Synergy scales appropriately with contribution size

Alternative Synergy Functions (not implemented, for reference):

Function Formula Properties
Arithmetic Mean $\sum a_i / N$ Weak complementarity, tolerates defection
Minimum $\min(a_i)$ Leontief production, bottleneck-limited
Cobb-Douglas $\prod a_i^{\alpha_i}$ Asymmetric weights possible

The Complementarity Parameter ($\gamma$)

Range: $\gamma \in [0, 1]$

Interpretation:

γ Value Interpretation Environment Behavior
0.0 No complementarity Purely additive value; no synergy benefit
0.3 Weak complementarity Modest cooperation incentive
0.5 Moderate complementarity Balanced individual/joint value
0.65 Validated default S-LCD case study calibration
0.8 Strong complementarity Substantial cooperation incentive
1.0 Maximum complementarity Synergy dominates individual value

Validated Value: $\gamma = 0.65$ achieves optimal multi-criteria performance across experimental validation (TR-1 §7.2).

Superadditivity Verification

Superadditivity

Proving Complementarity Creates Value

To verify that $V(\mathbf{a} \gamma)$ exhibits superadditivity, consider two actors choosing actions $(a_1, a_2)$.

Joint Value (power specification):

\[\Large V(\{a_1, a_2\}) = a_1^{\beta} + a_2^{\beta} + \gamma \sqrt{a_1 \cdot a_2}\]

Independent Values:

\[V(\{a_1\}) = a_1^{\beta} \quad \text{and} \quad V(\{a_2\}) = a_2^{\beta}\]

Superadditivity Condition:

\[V(\{a_1, a_2\}) > V(\{a_1\}) + V(\{a_2\})\] \[\Leftrightarrow a_1^{\beta} + a_2^{\beta} + \gamma \sqrt{a_1 \cdot a_2} > a_1^{\beta} + a_2^{\beta}\] \[\Leftrightarrow \gamma \sqrt{a_1 \cdot a_2} > 0\]

This holds for any $\gamma > 0$ and positive actions, confirming superadditivity. The synergy term $\gamma \sqrt{a_1 \cdot a_2}$ represents Added Value from collaboration.

Quantifying Added Value

Example: Both actors invest 50 units, $\gamma = 0.65$, $\theta = 20$

Logarithmic Specification:

Component Formula Result
Individual values $f_1(50) = 20 \cdot \ln(51)$ 78.64
  $f_2(50) = 20 \cdot \ln(51)$ 78.64
Synergy $g(50, 50) = \sqrt{50 \times 50}$ 50
Synergy value $0.65 \times 50$ 32.50
Total value $V = 78.64 + 78.64 + 32.50$ 189.78
Added Value $189.78 - (78.64 + 78.64)$ 32.50 (17% increase)

Value Appropriation

The Private Payoff Function

Value creation determines how much total value exists. Value appropriation determines who gets it.

Equation 11 (TR-1):

\[\Large \pi_i(\mathbf{a}) = e_i - a_i + f_i(a_i) + \alpha_i \left[V(\mathbf{a}) - \sum_{j=1}^{N} f_j(a_j)\right]\]

Components:

Term Formula Meaning
Endowment e_i Initial resources before interaction
Investment Cost -a_i Resources committed to partnership
Individual Return f_i(a_i) Return from own contribution
Synergy Share $\alpha_i$ × Synergy Share of collaborative surplus

Synergy = Collaborative Surplus

The synergy being divided is:

\[\Large \text{Synergy} = V(\mathbf{a}) - \sum_{j=1}^{N} f_j(a_j) = \gamma \cdot g(a_1, \ldots, a_N)\]

This is the Added Value from collaboration, value that exists only because actors worked together.

Bargaining and Shares ($\alpha_i$)

Constraint: $\Sigma\alpha_i = 1$ (all synergy must be allocated)

Determination Methods: 1. Equal Shares: $\alpha_i = 1/N$ (symmetric bargaining)

  1. Shapley Value: $\alpha_i$ based on marginal contribution
  2. Nash Bargaining: $\alpha_i$ reflects relative bargaining power
  3. Contractual: Pre-negotiated based on relationship structure

Connection to Interdependence: Actors with strong bargaining positions (low dependency, high alternatives) typically secure larger $\alpha_i$. See Interdependence Framework.

Specification Comparison

Experimental Validation

Both specifications were validated against the Samsung-Sony S-LCD joint venture (TR-1 §7-8):

Criterion Logarithmic (θ=20) Power (β=0.75) Winner
Overall Validation 58/60 (96.7%) 46/60 (76.7%) Logarithmic
Historical Alignment 16/16 12/16 Logarithmic
Cooperation Prediction 41% increase 166% increase Logarithmic
Bounded Predictions Yes No Logarithmic
Mathematical Tractability Moderate High Power

Why Logarithmic Wins Empirically

The logarithmic specification produces cooperation increases (41%) within the documented S-LCD range (15-50%), while the power specification produces increases (166%) exceeding realistic bounds.

Key Insight: The logarithmic function’s bounded growth prevents runaway predictions that don’t match real-world partnership dynamics.

When to Use Each

Scenario Recommended Rationale
Manufacturing JV Logarithmic Bounded returns, validated
Technology partnership Logarithmic Diminishing returns realistic
Platform ecosystem Either Power may be simpler
Academic baseline Power Cobb-Douglas tradition
Sensitivity analysis Both Compare robustness

Implementation Details

Code Correspondence

The value functions are implemented in coopetition_gym/core/value_functions.py:

# Logarithmic individual value
def logarithmic_individual_value(action, theta=20.0): return theta * np.log(1 + action)

# Power individual value
def power_individual_value(action, beta=0.75): return action ** beta

# Geometric mean synergy
def geometric_mean_synergy(actions): return np.prod(actions) ** (1 / len(actions))

# Total value
def total_value(actions, gamma=0.65, theta=20.0): individual = sum(logarithmic_individual_value(a, theta) for a in actions)
    synergy = gamma * geometric_mean_synergy(actions)
    return individual + synergy

Parameter Configuration

import coopetition_gym

# Using logarithmic specification (default)
env = coopetition_gym.make("TrustDilemma-v0",
    theta=20.0,      # Logarithmic scale
    gamma=0.65,      # Complementarity
)

# Using power specification
env = coopetition_gym.make("TrustDilemma-v0",
    value_spec="power",
    beta=0.75,       # Power exponent
    gamma=0.65,      # Complementarity
)

Equilibrium Implications

How Complementarity Affects Equilibrium

Higher γ (more complementarity) shifts equilibrium toward:

Complementarity and Trust

Complementarity interacts with trust dynamics (TR-2): 1. High γ creates incentive to cooperate → builds trust

  1. Built trust enables more cooperation → realizes synergy
  2. Realized synergy reinforces cooperative equilibrium

This creates a virtuous cycle when γ is high and a vicious cycle when γ is low.

Benchmark Evidence

From 760 experiments (76,000 episodes):

γ Level Mean Cooperation Mean Trust Mean Return
0.50 42.3% 41.8% 34,521
0.65 52.1% 54.3% 47,832
0.80 61.4% 67.2% 58,947

Insight: Higher complementarity produces more cooperative outcomes, higher trust, and higher returns.

Practical Applications

For Partnership Design

For Environment Customization

# High-complementarity scenario (technology partnership)
high_comp_env = coopetition_gym.make("TrustDilemma-v0",
    gamma=0.80,  # Strong synergy
    theta=25.0,  # Higher value scale
)

# Low-complementarity scenario (commodity market)
low_comp_env = coopetition_gym.make("TrustDilemma-v0",
    gamma=0.35,  # Weak synergy
    theta=15.0,  # Lower value scale
)

For Research

Further Reading

Primary Source

Background

Technical Reports