Predicting Macroeconomic Trends Through Real-Time Mobile Data Collection
Jul 12, 2015
Predicting Macroeconomic Trends Through Real-Time Mobile Data Collection
Author: Jon Gosier D8A Group, LLC
Conducted on Behalf of Market Atlas, LLC Telephone: (+1) 520-301-7906; [email protected]
Acknowledgements: Ahmed Maawy, Research Assistant; Akin Sawyerr, CSO, Market Atlas; Justin Mahwikizi, CEO, Market Atlas;
Appfrica and D8A Group LLC; and the John S. and James L. Knight Foundation
Does real-time consumer spending data predict
macro-economic trends?
Jon D. Gosier Lead Researcher
Justin Mahwikizi Market Atlas, CEO
Akin Sawyerr Market Atlas, CSO
Ahmed M. Maawy Assistant Researcher
Project Team
The ultimate goal of this project is to see if there are strong correlations that can be found between real-time consumer spending patterns and macro-economic trends and market fluctuations in African countries.
If successful, our methodology will present a new way investment decisions can be made as it relates to Africa and other emerging market countries which suffer from poor private sector visibility and financial infrastructure.
Methodology
Real-Time Consumer Spending (RTCS) is the blanket term we created to refer to all statistically relevant data collected from a sample of data providers to represent larger population trends.
This first half of the experiment was limited to collecting RTCS data and using it to test our hypotheses about macroeconomic trends. The second half will focus on finding strong correlations.
Methodology
Mombasa, Kenya
* Surveys via Mobile Phone and SMS
* Accounted for pricing bias
* Collected sales data from vendors instead of consumers
* Tested correlations against macro-economic indicators like Gross Domestic Product, Purchasing Power Parity, Inflation, Foreign Direct Investment, Debt and others.
Methodology
* Cellphone Vendors
* SIM Card Resellers
* Fruit Vendors
* Meat Vendors
* Grain/Rice Vendors
* General Store
* Clothing Vendors
Data Providers
Findings
Vendor Type NOV-14 DEC-14 Trend Change % Over Inflation
Clothes 200 211 ↑ 5.5% ↓
Everything 8.16 7.54 ↓ -7.59% ↓
Grains/Rice 900 910 ↑ 1.11% ↓
Meat 450 390 ↓ -13.33% ↓
Fruit 19000 21774 ↑ 14.6% ↑
Cell Phones 44 27 ↓ -38.64 ↓
SIM Cards 300 100 ↓ -66.66% ↓
All 2986.02 3345.65 ↓ -15.00% ↓
Kenya Microeconomic RTCS Data (Monthly)
Kenya Macroeconomic Data (Annual)
2005 2007 2009 2011 2013 Trend
GDP ($ billions) $18.7 $31.9 $37.0 $41.9 $55.2 ↑
GDP (2-yr change) 25.5% 70.59% 15.99% 13.24% 31.74% ↑
GDP (growth rate) 5.91% 6.99% 2.74% 4.42% 4.69% ↓
GDP (per capita) $523.61 $721.46 $771.29 $816.44 $994.31 ↑
Real Interest Rate 7.6% 5.0% 2.8% 3.8% 10.9% ↑
Consumer Price Index
72.57 80.24 102.09 121.17 140.11 ↑
Inflation (consumer prices annual %)
10.3% 9.8% 9.2% 14.0% 5.7% ↓
Kenya Macroeconomic Data (Monthly)
NOV-14 DEC-14 Trend
Inflation Rate 6.43% 6.09% ↓
Food Inflation 8.16% 7.54% ↓
Consumer Price Index
151.92 pts 151.85 pts ↓
CPI (% change) -0.21% -0.05% ↓
Findings
1- Because interest rates were up (↑) for the year, and inflation was down (↓) month on month between November and December, traditional economic patterns would suggest consumer spending should also be down.
Our observations showed that it was.
2- Some products seemed to defy expectations. Clothing, Grains/Rice, and Fruit all moved volumes at rates higher than the percentage change in inflation.
One might conclude that this is because these are ‘essentials’ that people will buy regardless of economic trends.
Findings
3- Historic performance of the Kenyan Stock Market and rates of inflation seem to follow similar patterns.
Findings
Kenya Inflation Rate Kenya Stock Market
Future Research
* If RTCS and Inflation show strong a correlation to one another, and Inflation and the stock market show correlated patterns, can RTCS serve as leading indicator for market trends?
* If there is a causal link between RTCS and Inflation, is there a link with other macroeconomic indicators?
* Why do fruit, clothing, and grains/rice defy trends that other industries do not?
Predicting Macroeconomic Trends Through Real-Time Mobile Data Collection