The core of our project rests on estimating a common cointegrating vector across the six CEMAC countries and calibrating a reaction parameter κ. The interest rate applied to digital wallets follows this rule:
rate = 3% + κ × ECT
where ECT is the Error-Correction Term — the deviation of real money supply from its long-run equilibrium with GDP and inflation. When κ is calibrated via Bellman/LQR optimal control, the resulting rate is not arbitrary: it is the mathematically optimal response of a central bank with a quadratic loss function over monetary stability and rate smoothness.
Use the slider below to see how κ affects the trajectory of the ECT and the digital wallet rate.
Bellman/DARE optimal control solution for a central bank with λ = 0.1 rate-smoothness penalty and β = 0.99 quarterly discount factor.
Bootstrapped calibration on CEMAC's empirical shock distribution. Nearly identical to the LQR result, proving robustness.
The rule damps monetary imbalances by a third compared to a constant 3% neutral rate.
The semi-elasticity of money growth to the policy rate is estimated by instrumental variables (euro area rate as instrument), addressing the endogeneity critique directly.
Our system follows the champion/challenger framework used by the Federal Reserve, the ECB, and the Bank of England:
Linear DOLS + LQR — transparent, causally validated, mathematically optimal. This model is deployed in production.
NARDL asymmetric (κ⁺ for reserve defence) and Threshold crisis mode (higher κ when |ECT| exceeds a pre-approved threshold).
CEMAC-adapted NN and Kolmogorov-Arnold Network (KAN) — run offline to validate the champion. They never affect production decisions without MPC approval.
Bridge equations and dynamic factor models detect monetary shocks within days, not months, using high-frequency indicators (oil prices, mobile money, PAPSS flows).