A4: Network Effects in Payment Systems

L04 Payment Systems

Assignment Brief

Network Effects: When Does a Payment System Take Off?

Introduction

Network effects (the phenomenon where a product becomes more valuable as more people use it) determine when a payment system reaches critical mass (the point where enough people use it that growth becomes self-sustaining). Three theories model network value V as a function of n users:

Metcalfe's Law: V = n²/1000 (every pair can connect)
Odlyzko-Tilly: V = n·ln(n)/10 (only close contacts matter)
Linear: V = n (no network effects)

The divisors (1000 and 10) normalize raw connection counts to comparable value units.

Key Concepts:

Switching cost: the time, effort, and money needed to change from one payment system to another
Critical mass: the number of users at which network value exceeds switching costs, making adoption self-sustaining

Your Task

You will explore how different network value models predict when payment systems reach critical mass under various switching cost scenarios.

Variations to Implement

VARIATION 1 - Lower switching cost threshold from 500 to 100
- How does critical mass change for each model?
- Which model reaches critical mass fastest?
VARIATION 2 - Raise switching cost to 2,000
- How does critical mass change?
- Which model never reaches critical mass within 1000 users?
VARIATION 3 - Add Reed's Law
- Implement Reed's Law: V = 2^(n/10)
- Where does it cross the switching cost threshold (500)?
- How does it compare to the other models?

Open Extension (Advanced)

Remove the scaling divisors and use raw formulas:

Metcalfe: V = n²
Odlyzko-Tilly: V = n·ln(n)
Linear: V = n

Questions:

How do critical mass points change?
What do the divisors represent economically?
Why are they needed for realistic modeling?

How to Run

Use Google Colab or your local Python environment with matplotlib and numpy installed.

Time Allocation

45 minutes: Implementation and analysis
10 minutes: Presentation preparation

Deliverables

Modified Python code implementing all three variations
7-slide presentation (use Marp or PowerPoint) summarizing your findings
Brief written analysis of what the scaling factors mean economically

References

Metcalfe, B. (2013). "Metcalfe's Law after 40 Years of Ethernet"
Odlyzko, A., & Tilly, B. (2005). "A refutation of Metcalfe's Law and a better estimate for the value of networks and network interconnections"

Grading Criteria

Correctness (40%): All three variations implemented correctly
Analysis (30%): Clear explanation of critical mass changes
Presentation (20%): Clear visualization and communication
Extension (10%): Insight into scaling factors (bonus)

Model Answer

Show Model Answer Presentation

Slide 1

Network Effects: When Does a Payment System Take Off?

Assignment A4 Model Answer

References:

Metcalfe, B. (2013). "Metcalfe's Law after 40 Years of Ethernet"
Odlyzko, A., & Tilly, B. (2005). "A refutation of Metcalfe's Law..."

Slide 2

The Model: Three Network Value Functions

Network value V as a function of n users:

Metcalfe's Law: $V = \frac{n^2}{1000}$
- Every user can connect to every other user
- Total possible connections = n(n-1)/2 ≈ n²/2
Odlyzko-Tilly: $V = \frac{n \ln(n)}{10}$
- Only close contacts matter (logarithmic decay)
- More realistic for large networks
Linear: $V = n$
- No network effects (baseline)

Scaling factors (1000, 10): Normalize raw connection counts to comparable value units

Slide 3

Baseline Results (Threshold = 500)

Critical mass points where V > 500:

Linear: n = 500 (V = n, so n = 500)
Metcalfe: n = 708 (n²/1000 = 500 → n = √500,000 ≈ 707.1)
Odlyzko: n ≈ 750 (solve n·ln(n)/10 = 500 numerically)

Key insight: Linear reaches critical mass FIRST (at n=500), while Metcalfe requires 42% more users (n=708). The n²/1000 scaling divisor means Metcalfe's quadratic advantage doesn't overcome the denominator until larger network sizes

Slide 4

Variation 1: Lower Threshold (100)

Critical mass points where V > 100:

Linear: n = 100
Metcalfe: n = 317 (√100,000 ≈ 316.2)
Odlyzko: n ≈ 278

Key insight: Lower switching costs favor rapid adoption even with weak network effects

Slide 5

Variation 2: Higher Threshold (2,000)

Critical mass points where V > 2,000:

Linear: n = 2,000 (beyond our 1000-user range)
Metcalfe: n ≈ 1,414 (√2,000,000 ≈ 1,414.2, beyond range)
Odlyzko: n > 1,000 (beyond range)

Key insight: High switching costs prevent any model from reaching critical mass within 1000 users. Payment systems need to lower barriers to adoption.

Slide 6

Variation 3: Reed's Law (Exponential Growth)

Reed's Law: $V = 2^{n/10}$ (value from forming groups/coalitions)

Critical mass for threshold = 500:

Reed's Law: n ≈ 90 (since 2^9 = 512 > 500)

Comparison:

Reed: n = 90
Metcalfe: n = 708
Odlyzko: n = 750
Linear: n = 500

Key insight: If group formation drives value (e.g., merchant networks), critical mass arrives much earlier

Slide 7

The Role of Scaling Divisors

Without divisors (raw formulas):

Metcalfe V = n²: Critical mass at n = 23 (√500 ≈ 22.4)
Odlyzko V = n·ln(n): Critical mass at n ≈ 75

With realistic divisors (/1000, /10):

Metcalfe: n = 708
Odlyzko: n = 750

What divisors represent:

Economic interpretation: Not all connections generate equal value
Divisors calibrate theoretical models to empirical network value data
Example: Facebook has 3 billion users, but value isn't 9×10¹⁸ dollars

Key insight: Scaling factors are essential for realistic predictions of when payment systems become viable

← Prev: CBDC Disintermediation Next: Winner-Take-All →