All projects

Time series research

Forecasting Ghana’s GDP.

How do three forecasting methods handle the same economic history?

KNUST2025

The question

For my final-year statistics thesis at KNUST, I compared three ways of forecasting Ghana’s GDP. Each model had to contend with the same changes in the country’s economic history from 1960 to 2023.

My contribution

I benchmarked ARIMA, exponential smoothing and Prophet. The findings were submitted as a policy reference to the Bank of Ghana and the Ministry of Finance.

Three models, one series

ARIMA
Uses relationships between earlier values and forecast errors, with differencing where appropriate.
Exponential smoothing
Gives more weight to recent observations as it updates the estimated level and trend.
Prophet
Represents a time series through components such as trend and changes in its slope.

Historical data

Ghana’s GDP, 1960–2023

Annual GDP in billions of current US dollars. Select a year to inspect the series.

Annual GDP rises from around 1.22 billion US dollars in 1960 to 80.55 billion in 2023.

Series from the World Bank, via FRED. The chart shows historical observations.

Next projectMaggie