Decoupling (the signature finding)

This page is the core of the project.

It shows, with data, that Mauritania’s inflation system did not merely cool down after 2022 - it changed its mechanics.

We are testing a behavioral claim:

In 2022-2023, inflation behaved like an amplifier of shocks. In 2024-2025, inflation behaved like a shock absorber.

To prove that, we need two ingredients:

  • Transmission (β): does FX move inflation?
  • Memory (ρ): does inflation carry over?

1) Definitions (symbol explanation)

1.1 Monthly inflation (MoM)

Let \(CPI_t\) be the CPI index in month \(t\).

\[ \pi_t = 100 \times \left(\frac{CPI_t}{CPI_{t-1}} - 1\right) \]

  • \(\pi_t\): monthly inflation in percent
  • If \(\pi_t = 1\), prices rose ~1% from last month

1.2 Monthly FX change (USD/MRU)

Let \(FX_t\) be the monthly average USD/MRU in month \(t\).

\[ \Delta fx_t = 100 \times \left(\frac{FX_t}{FX_{t-1}} - 1\right) \]

  • If \(\Delta fx_t > 0\): the Ouguiya weakened vs USD (depreciation)
  • If \(\Delta fx_t < 0\): the Ouguiya strengthened vs USD (appreciation)

2) The behavioral model

We estimate a simple equation:

\[ \pi_t = \alpha + \beta \cdot \Delta fx_t + \rho \cdot \pi_{t-1} + \varepsilon_t \]

Every term has a real-world meaning:

  • \(\alpha\) (alpha): baseline inflation drift (what inflation tends to be even if FX is flat)
  • \(\beta\) (beta): FX pass-through
    • if \(\beta = 0.12\), then a 1% USD/MRU depreciation is associated with ~0.12% higher monthly inflation
  • \(\rho\) (rho): inflation persistence / inertia / memory
    • if \(\rho = 0.74\), then 74% of last month’s inflation carries into this month
  • \(\varepsilon_t\) (epsilon): everything else (supply shocks, policy, commodity prices, etc.)

Important: this is not a claim of perfect causality. It is a disciplined way to describe system behavior.

3) Why “decoupling” is the right word

A classic problem in small open economies is:

  • FX shocks enter import prices
  • import prices enter CPI
  • CPI becomes persistent via expectations and pricing behavior

In an “amplifier” regime: - \(\beta\) is meaningful (FX transmits) - \(\rho\) is high (inflation persists) - a shock today influences next month and the month after

In an “absorber” regime: - \(\beta\) weakens (FX transmits less) - \(\rho\) weakens (inflation forgets faster) - shocks dampen instead of propagating

So the signature test is not “was inflation low?” It’s: did β and ρ fall together?

4) The signature overlay chart (β + ρ behavior)

This chart is the visual proof of decoupling: - pass-through weakens - persistence weakens

If the inflation system changed, it should show up here.

Structural Overlay (β and ρ)

5) The regime split

We split the sample into two equal-length windows:

  • Amplifier regime: 2022-2023 (commodity shock period)
  • Absorber regime: 2024-2025 (post-2023 modernization + post-2024 transition)

Then we re-estimated:

  • \(\beta\) for headline CPI and food CPI
  • \(\rho\) for headline CPI and food CPI

The result table was produced by your analysis pipeline and saved as:

  • analysis/outputs/10_regime_table.csv

Regime Table (measured)

Regime β (Headline CPI) ρ (Headline CPI) β (Food CPI) ρ (Food CPI) Months
Amplifier (2022-2023) 0.121 0.738 (measured in your output) 0.738 (food persistence) 24
Absorber (2024-2025) -0.022 0.399 (measured in your output) 0.399 (food persistence) 24

Interpretation:

  • In 2022-2023, FX depreciation shows up in inflation (β positive).
  • In 2024-2025, the FX-to-inflation relationship collapses (β ~ 0 or negative).
  • Inflation also becomes less self-reinforcing (ρ falls from ~0.74 to ~0.40).

That is not “inflation went down.” That is a regime change in inflation dynamics.

6) Memory: how long does a shock stay alive?

Persistence gives a clean intuition.

If inflation follows:

\[ \pi_t \approx \rho \cdot \pi_{t-1} \]

Then after \(h\) months, the remaining portion of a shock is:

\[ \rho^h \]

You computed:

  • \(\rho \approx 0.573\) in the full sample
  • half-life ≈ 1.24 months
  • shock remaining after 3 months: \(\rho^3 \approx 0.188\)

That means:

  • after 3 months, only ~19% of the shock remains (on average)
  • but in 2022-2023, persistence was much higher than average
  • by 2024-2025, persistence drops sharply

This is exactly how an “amplifier” becomes an “absorber.”

7) Volatility did not collapse - so what changed?

A key sanity check:

If inflation became stable only because the world calmed down, then volatility would collapse too.

But your volatility results show:

  • FX volatility pre-2023: 0.4903

  • FX volatility 2024+: 0.5036

  • inflation volatility pre-2023: 0.3088

  • inflation volatility 2024+: 0.3222

So volatility is not the main explanation.

That strengthens the interpretation:

The transmission mechanism weakened. The system is dampening shocks instead of propagating them.

8) Candidate breakpoint markers (what happened when)

We do not claim politics “caused” the decoupling. We treat events as candidate breakpoints because they can change: - credibility - policy coordination - market design - expectations

2022: Commodity shock (global)

  • Food and energy price shock pushed inflation into the “amplifier” zone.
  • In behavioral terms: price-setters become defensive, and persistence rises.

Late 2023: FX market modernization (institutional)

  • The FX market moved toward more transparent price discovery.
  • In microstructure terms: less bottleneck, better matching, less stale pricing.

Aug 2024: New Prime Minister (coordination signal)

  • A government transition can shift administrative execution and expectations.
  • Even with the same formal objectives, implementation intensity can change.

This is why we annotate these dates later in the Results charts with vertical markers.

9) What this finding does and does not say

It does say:

  • inflation persistence fell sharply after 2023
  • FX pass-through weakened dramatically after 2023
  • Mauritania’s inflation stopped behaving like an amplifier

It does not say:

  • “FX is irrelevant”
  • “inflation is solved forever”
  • “politics alone caused the change”

It says the system’s mechanics changed in a measurable way.

Explore the Evidence

This analysis proceeds through three mechanical lenses:

  1. Transmission — How FX shocks propagate (or don’t)
  2. Memory — How inflation persists (or fades)
  3. Structure — How different CPI categories behave

The Appendix contains full regression tables, robustness checks, and the forward-looking synthesis of what this decoupling means for Mauritania’s gas export era.