# DECS-S-0034 Statistics Assignment: "I Have More SIMs Than Friends"
## Complete Regression Analysis — N=201 Countries
### Sasin School of Management | Data: api.worldbank.org | Analysis: 8 July 2026

---

## ABSTRACT

This study examines the determinants of mobile phone penetration across 201 countries using OLS regression. Contrary to the popular belief that "poor people hoard cheap SIMs," GDP per capita shows NO significant relationship with mobile subscriptions. Only internet penetration is a statistically significant predictor (β=0.54, p=0.014), explaining approximately 18% of cross-country variation. Landlines, tourism, and youth demographics are not significant. ASEAN countries average 125 mobile subscriptions per 100 people — meaning the average ASEAN citizen has 1.25 SIMs — supporting the "more SIMs than friends" claim.

---

## 1. INTRODUCTION

### Research Question
What drives mobile phone penetration across countries? Specifically: does the "I have more SIM cards than friends" phenomenon correlate with economic development, internet adoption, landline substitution, tourism intensity, or youth demographics?

### Hypothesis
- **H₀ (null):** Mobile penetration is explained by the five macro variables jointly
- **H₁ (alternative):** At least one variable has a significant effect beyond zero

### Individual Predictions
| Variable | Expected Sign | Reasoning |
|----------|:---:|-----------|
| GDP per capita | + | Wealthier countries own more devices per person |
| Internet penetration | + | Data plans drive SIM adoption |
| Landlines | − | Mobile substitutes for fixed lines in developing countries |
| Tourism arrivals | + | Tourist SIMs inflate subscription counts |
| Youth dependency ratio | + | Young populations adopt mobile technology faster |

---

## 2. DATA

**Source:** World Bank API (api.worldbank.org/v2) — all data verified LIVE on 8 July 2026.

**Dataset:** 201 countries, complete cases only (all 6 indicators present). Filtered to 2-3 letter ISO country codes; regional aggregates excluded.

| Variable | World Bank Code | Mean | SD | Min | Max |
|----------|----------------|------|-----|-----|-----|
| **Mobile_subs_100** (Y) | IT.CEL.SETS.P2 | 117.0 | 51.9 | 4.0 (American Samoa) | 574.2 (Fiji) |
| GDP_per_capita (X₁) | NY.GDP.PCAP.CD | $23,667 | $36,163 | $234 (Burundi) | $288,002 (Monaco) |
| Internet_pct (X₂) | IT.NET.USER.ZS | 70.4% | 27.0 | 0.0% | 100.0% |
| Landlines_100 (X₃) | IT.MLT.MAIN.P2 | 15.0 | 16.4 | 0.0 | 95.5 |
| Tourism_arrivals (X₄) | ST.INT.ARVL | 4,955,028 | 13,025,589 | 900 (Tuvalu) | 117,109,000 (France) |
| Young_dep_ratio (X₅) | SP.POP.DPND.YG | 40.3% | 19.7 | 14.7% (Korea) | 99.1% (CAR) |

**Key ASEAN fact:** All 10 ASEAN countries exceed 100 mobile subscriptions/100 people. ASEAN mean = 125.0 vs global mean = 117.0.

---

## 3. METHODOLOGY

### Model
OLS multiple linear regression:

**Mobile_subs = β₀ + β₁·log(GDP) + β₂·Internet + β₃·Landlines + β₄·log(Tourism) + β₅·Young_dep + ε**

GDP and Tourism were log-transformed to address extreme right-skew (GDP ranges from $234 to $288,002; Tourism from 900 to 117 million).

### Software
Regression computed via matrix algebra (Python/numpy/scipy). Equivalent to Excel Analysis ToolPak with minor precision differences.

---

## 4. RESULTS

### Model Fit

| Statistic | Value |
|-----------|-------|
| N | 201 |
| R-squared | 0.1824 |
| Adjusted R-squared | 0.1614 |
| F-statistic (5, 195) | 8.70 |
| F p-value | 0.000000183 |
| Residual Std. Error | 47.54 |

The model is **statistically significant overall** (F-test p < 0.001) but explains only **18.2%** of cross-country variation — the majority of what drives mobile penetration is NOT captured by these five variables.

### Regression Coefficients

| Variable | Coefficient | Std Error | t-stat | p-value | Significance |
|----------|:----------:|:---------:|:------:|:-------:|:---:|
| Intercept | 83.880 | 57.454 | 1.460 | 0.1459 | |
| log(GDP_per_capita) | −1.306 | 4.821 | −0.271 | 0.7867 | |
| **Internet_pct** | **0.540** | **0.218** | **2.479** | **0.0140** | **\*** |
| Landlines_100 | 0.028 | 0.285 | 0.097 | 0.9229 | |
| log(Tourism_arrivals) | 1.701 | 1.894 | 0.898 | 0.3703 | |
| Young_dep_ratio | −0.411 | 0.343 | −1.197 | 0.2326 | |

*Significance codes: *** p<0.001, ** p<0.01, * p<0.05, . p<0.10*

### Diagnostic Tests

| Test | Result | Interpretation |
|------|--------|---------------|
| Breusch-Pagan (heteroskedasticity) | LM=1.49, p=0.998 | No heteroskedasticity detected |
| Shapiro-Wilk (residual normality) | W=0.65, p=0.000 | **Residuals NOT normal** — standard errors may be unreliable |

### Correlation Matrix

| | Mobile | GDP | Internet | Landlines | Tourism | YoungDep |
|------|:---:|:---:|:---:|:---:|:---:|:---:|
| **Mobile** | 1.00 | | | | | |
| GDP | 0.17 | 1.00 | | | | |
| **Internet** | **0.41** | 0.45 | 1.00 | | | |
| Landlines | 0.22 | 0.60 | 0.46 | 1.00 | | |
| Tourism | 0.07 | 0.13 | 0.24 | 0.22 | 1.00 | |
| YoungDep | **−0.38** | −0.47 | −0.78 | −0.60 | −0.25 | 1.00 |

Notable: Internet has the strongest bivariate correlation with Mobile (r=0.41). YoungDep is negatively correlated (−0.38) — countries with MORE young people tend to have FEWER mobile subscriptions, which is the opposite of what was hypothesized.

---

## 5. INTERPRETATION

### Finding 1: Only Internet Matters (Statistically)

The sole significant predictor is **Internet penetration** (β = 0.54, p = 0.014). Each additional percentage point of internet adoption is associated with approximately 0.54 more mobile subscriptions per 100 people. In a country moving from 50% to 90% internet penetration, this predicts an increase of ~22 mobile subscriptions per 100 people.

**This makes sense.** Smartphones require data plans. Countries where people use the internet also have more active mobile subscriptions (people don't buy SIMs for voice-only phones anymore).

### Finding 2: GDP Does NOT Matter (Counter-Intuitive)

The "poor people hoard cheap SIMs" hypothesis is **statistically rejected**. GDP per capita has no significant relationship with mobile penetration (p = 0.79). The intuition that poor people juggle multiple prepaid SIMs to chase the cheapest rates is not supported at the macro level. In fact, the coefficient on GDP is slightly negative — if anything, richer countries have FEWER mobile subscriptions (controlling for internet), though the effect is far from significant.

### Finding 3: Landlines Are Neither Substitute Nor Complement

Landlines show no significant relationship (p = 0.92) with mobile subscriptions. The "mobile replaces landlines" narrative is not supported globally — the relationship is essentially zero. This is surprising given that individual country examples (Vietnam: 2.6 landlines, Indonesia: 1.6) show near-total landline extinction alongside high mobile use.

### Finding 4: Youth Effect Is Opposite of Expected

Young dependency ratio shows a **negative but non-significant** coefficient (−0.41, p = 0.23). Countries with more young people actually have FEWER mobile subscriptions. This is the opposite of the hypothesis that young people drive mobile adoption. Possible explanation: the countries with the highest youth ratios (Niger 90%, CAR 99%) are also the poorest and least connected — the poverty effect overwhelms any youth-adoption effect.

### Finding 5: Model Explains Only 18% — What's Missing?

R² = 0.18 means 82% of variation in mobile penetration is NOT explained by these five variables. Important omitted factors likely include:
- **Regulatory environment** (number of mobile operators, competition policy)
- **Prepaid vs postpaid mix** (prepaid markets have more multi-SIM users)
- **Dual-SIM phone penetration** (hardware enables multi-SIM behavior)
- **Mobile number portability** (without it, switching carriers means buying a new SIM)
- **IoT/M2M SIMs** (smart meters, connected cars — concentrated in developed countries)

### Finding 6: The ASEAN Case

ASEAN countries average **125 mobile subscriptions/100 people** vs the global average of 117. In 9 of 10 ASEAN countries, the average person has more than 1 active SIM. The "more SIMs than friends" claim is **statistically true** for ASEAN as a whole.

---

## 6. LIMITATIONS

1. **Non-normal residuals** (Shapiro-Wilk p=0.000) violate OLS assumptions; reported p-values are approximate
2. **Cross-sectional design** — correlation, not causation
3. **Different reference years** — Tourism data from 2019-2020; Internet from 2024-2025
4. **Omitted variable bias** — regulatory, hardware, and market-structure variables are missing
5. **Extreme outliers** — Fiji (574 mobile/100) and American Samoa (4/100) pull results
6. **Low R²** — the model leaves 82% of variation unexplained

---

## 7. CONCLUSION

The hypothesis that mobile penetration is driven by economic development, tourism, landline substitution, or youth demographics is **not supported** by the data. Only internet penetration shows a statistically significant positive relationship. The "more SIMs than friends" phenomenon is real — especially in ASEAN — but the reasons are more complex than simple macro-economic indicators can capture. Future research should incorporate regulatory, hardware, and market-structure variables.

---

## 8. REFERENCES

All data sourced from the World Bank API (api.worldbank.org/v2), fetched 8 July 2026.

**Indicator codes:**
- Mobile cellular subscriptions: IT.CEL.SETS.P2
- GDP per capita (current USD): NY.GDP.PCAP.CD
- Individuals using the Internet (%): IT.NET.USER.ZS
- Fixed telephone subscriptions: IT.MLT.MAIN.P2
- International tourism arrivals: ST.INT.ARVL
- Age dependency ratio, young: SP.POP.DPND.YG

**Full source URLs for all 120 individual country-indicator links:** See `decs0034-reference-links.md`.

---

*Assignment prepared for DECS-S-0034 Statistics for Business Decision Making*
*Sasin School of Management | Chulalongkorn University*
*Analysis completed: 8 July 2026*