Predicting tax morale in Nigeria
A machine learning model of tax morale — the intrinsic motivation to pay taxes honestly — trained on a nationally representative survey of 15,043 Nigerian households.
- Best model
- 81.9%Gradient Boosting accuracy
- F1-score
- 0.881ROC-AUC 0.842
- Respondents
- 15,04362 engineered features
- High morale
- 73.7%believe evading tax is wrong
What predicts tax morale
- 1Expected penalty for tax evasion47.8%
- 2Perceived obligation to pay taxes5.3%
- 3Expected living condition in a year4.1%
- 4Satisfaction with state government services3.4%
- 5Perceived tax burden3.1%
Deterrence dominates: the expected penalty for evasion carries more weight than every other factor combined. Demographic characteristics such as gender and education do not appear in the top ten at all. Geopolitical zone does, at tenth — and not along the usual north–south lines, with the South-West recording the lowest morale (64.0%) and the South-East the highest (79.1%).
What this model does not tell you
It predicts a self-reported attitude toward tax evasion, not observed compliance behaviour. Its probabilities are uncalibrated, and it has not been audited for fairness across regions, gender, or education. It is exploratory research output and must not be used to make decisions about individuals.
Data and citation
Built on the NESG Nigeria Tax & Subsidy Perception Survey (2018–2019), openly available for academic research from the International Centre for Tax and Development, and funded by the Bill & Melinda Gates Foundation. Reusing the data requires citing its creators:
McCulloch, N., & Moerenhout, T. (2019). Nigeria Tax and Subsidy Perception Survey Dataset [Data set]. Nigerian Economic Summit Group (NESG) and International Centre for Tax and Development (ICTD), funded by the Bill & Melinda Gates Foundation. ictd.ac/datasets