Malawi International
Merchandise Trade Statistics

Data Mining Project · KNIME Analytics Platform · 2022–2025

Philip Katema · SPU 3,000 Records 25 Import Commodities 25 Export Commodities 4 Mining Modules Source: Malawi NSO

Total Imports 2022–2025

$13.1B

USD · 48 months

Total Exports 2022–2025

$3.8B

USD · 48 months

Cumulative Trade Deficit

$9.3B

Exports cover only 28.7% of imports

Tobacco Export Share

50.7%

Single commodity dominance

Time Series Forecasting

Linear Regression vs Random Forest · Imports & Exports

Model Performance — Imports

Metric	Linear Regression	Random Forest
R²	0.546	0.721
MAE	$23.4M	$20.6M
RMSE	$29.0M	$22.7M
MAPE	8.7%	7.2%
Bias (MSD)	-$3.6M	-$9.2M

Imports — R² Comparison

Key Finding — Imports

Random Forest explains 72.1% of monthly import variance — a 17.5 percentage point improvement over Linear Regression. The gap reflects non-linear fuel and fertiliser import spikes that linear models cannot capture from lag features alone.

Model Performance — Exports

Metric	Linear Regression	Random Forest
R²	0.513	0.638
MAE	$19.5M	$15.0M
RMSE	$21.1M	$18.2M
MAPE	32.0%	24.6%
Bias (MSD)	$3.7M	$1.0M

Exports — R² Comparison

Key Finding — Exports

Export forecasting is harder than import forecasting across all models — R² values are consistently lower. The high MAPE (24.6% even for RF) reflects Tobacco's auction season spike which dominates export revenue and resists pattern-based prediction without commodity-level price data.

Commodity Clustering

k-Means + Hierarchical · 25 Export Commodities · PCA Visualisation

PCA Scatter — Both Algorithms Agree

Cluster — Dominant Outlier

Tobacco Other Exports

PCA dim 0: −5.4 to −3.3 · Behaviourally isolated by both algorithms

Cluster — Mid-Tier Agricultural

TeaPulses GroundnutsMacadamia Nuts Soya Beans

Best diversification candidates · Distinct seasonal patterns

Cluster — Low-Value Mixed

Natural RubberPlastics Scrap MetalSugar

Low value · moderate variability

Cluster — Marginal Mass

CottonCoffee FishRice SpicesApparel ×2 +8 more

14 undifferentiated commodities · near-zero PCA separation

Key Finding — Clustering

Both k-Means and Hierarchical Clustering assign every commodity to identical clusters — 100% algorithm agreement without being provided value data. Tobacco's behavioural uniqueness is confirmed algorithmically: its seasonal auction cycle and scale create a trade signature no other commodity replicates.

Anomaly Detection

Rolling Z-Score · 12-Month Window · 4-Tier Classification

Anomaly Count by Category

Z-Score Legend

ANOMALY — Z > 2.0 or Z < −2.0 · Statistically extreme

ELEVATED HIGH — 1.0 < Z ≤ 2.0

ELEVATED LOW — −2.0 ≤ Z < −1.0

NORMAL — −1.0 ≤ Z ≤ 1.0

Four Driving Shocks

Cyclone FreddyMar–Apr 2023

44% Kwacha DevaluationNov 2023

El Niño DroughtJan–Sep 2024

Tanzania Trade DisputeMar–Apr 2025

Selected Anomaly Timeline — Imports & Exports

IMPORTS · Apr 2023

Anomaly Low · Z = −2.394

Acute forex shortage post-Cyclone Freddy. Fuel imports collapsed due to dollar scarcity. Foreign reserves at crisis level — import compression at most severe point in dataset.

EXPORTS · Aug–Nov 2023

Elevated High × 4 months

Pre-devaluation export surge. Buyers locking in contracts before 44% kwacha adjustment (Nov 9 2023). Export values elevated in USD terms as kwacha pressure made Malawi more price-competitive.

IMPORTS · Jan–Apr 2024

Elevated Low × 3 months

Post-devaluation import compression. El Niño dry spells damaged 44% of maize crop. State of disaster declared in 23 of 28 districts (Mar 25 2024).

EXPORTS · Aug 2024

Anomaly High · Z = 2.159

2024 tobacco season record: 40% revenue increase to $396M from $284M. Sales volume up 10% to 133M kg. Average price surged 26.5% to $2.98/kg. Season close in August.

IMPORTS · Jul–Sep 2024

Elevated High × 3 months

Peak El Niño food import response. Government sourcing maize from Argentina and US as regional Zambia/Zimbabwe supply exhausted. WFP Flash Appeal activated for 4.2M food-insecure people.

IMPORTS · Jan 2026

Anomaly High · Z = 2.031 · $390.7M

Structural import surge against already-elevated 2025 baseline ($314.8M rolling mean). Not an isolated spike — culmination of a worsening structural pattern. World Bank: current account deficit near 20% of GDP, forex reserves below 1 month of imports.

Key Finding — Anomaly Detection

All 29 flagged periods across imports and exports trace to four macro shocks. The January 2026 Anomaly High is not an isolated event but the structural capstone of the entire dataset — confirming that Malawi's trade vulnerability is accelerating, not stabilising.

Association Rule Mining

Apriori · Min Support 0.3 · Min Confidence 0.6 · Min Lift 1.2

Top Import Association Rules (by Lift)

Antecedent → Consequent	Support	Conf.	Lift
Nuclear Reactors → Glass and Glassware	0.375	0.750	1.500
Electrical Machinery → Plastics	0.375	0.750	1.500
Fats and Oils → Vehicles	0.375	0.750	1.500
Plastics → Nuclear Reactors	0.354	0.708	1.417
Cereals → Electrical Machinery	0.354	0.708	1.417
Ceramic Products → Plastics	0.354	0.708	1.417
Fertilizer → Diesel and Others	0.354	0.708	1.417
Soaps → Nuclear Reactors	0.354	0.708	1.417
Tobacco Substitutes → Electrical Machinery	0.354	0.708	1.417
Other Made-Up Textiles → Lime	0.354	0.708	1.417

Import Cluster Groups

Construction Cluster · Lift 1.5

Nuclear Reactors/Boilers ↔ Glass and Glassware. Infrastructure investment and post-disaster reconstruction procurement cycles peak together.

Manufacturing Cluster · Lift 1.5

Electrical Machinery ↔ Plastics. Coordinated intermediate goods procurement by Malawi's manufacturing sector — arrive in same months.

Discretionary Cluster · Lift 1.5

Fats/Oils ↔ Vehicles. Apparently unrelated but share one driver: forex availability. When forex is available, both spike simultaneously.

Top Export Association Rules (by Lift)

Antecedent → Consequent	Support	Conf.	Lift
Spices → Cotton ⭐ Strongest rule	0.396	0.792	1.583
Cotton → Spices	0.396	0.792	1.583
Wooden Furniture → Tea	0.375	0.750	1.500
Tobacco → Spices	0.375	0.750	1.500
Macadamia Nuts → Tea	0.375	0.750	1.500
Skins and Hides → Soya Beans	0.354	0.708	1.417
Natural Rubber → Macadamia Nuts	0.354	0.708	1.417
Pulses → Spices	0.354	0.708	1.417
Beverages → Tobacco	0.354	0.708	1.417
Scrap Metal → Apparel (Not Knitted)	0.333	0.667	1.333

Export Cluster Drivers

Harvest Calendar Effect

Tobacco, Tea, Spices, Macadamia Nuts all peak together in Q3 (Jul–Sep). Co-movement reflects Malawi's unified agricultural season — not production interdependencies.

Agro-Ecological Co-Location

Spices ↔ Cotton (Lift 1.583, strongest rule overall) — grown in same southern/central regions by same smallholder communities. Skins/Hides ↔ Soya Beans — livestock and rotation crops from same farms.

Import vs Export — Co-Movement Drivers

The Central Asymmetry

Export co-movement is governed by climate and agricultural geography — commodities co-move because they share growing regions, seasonal calendars, and farmer communities. Import co-movement is governed by a single binding financial constraint: foreign exchange availability. When forex is available, a broad basket of unrelated commodities spikes simultaneously; when scarce, they all drop together. This asymmetry — nature governing exports, finance governing imports — is the most succinct summary of Malawi's structural trade challenge.

Malawi InternationalMerchandise Trade Statistics

Time Series Forecasting

Commodity Clustering

Anomaly Detection

Association Rule Mining

Malawi International
Merchandise Trade Statistics