Le texte de l’ouvrage est en anglais, tel qu’il a été écrit ; seule l’interface du site est en français.

Annex G — Protocol and empirical diagnostic

Annexe 7 sur 17 · ≈ 4 min de lecture · 7 paragraphes · 29ᵉ des 40 sections de lecture · rechercher dans l’ouvrage

This annex establishes the reproducible protocol and an identification diagnostic; it reports no causal estimate. Such an estimate remains conditional on access to the sovereign debt register — foreign-currency share and contractual flows — and to high-frequency market series, not yet assembled at this stage.¶ 1

G.1 Purpose and linkage. The body of the report (§9.5, §9.6, §9.7) explicitly defers the empirical demonstration: monetary policy spillover is established there as the least priced link of the accountability chain, and its eventual correction can be grounded only after attributing the incidence to an identified shock, separating the transfer from the welfare loss, and neutralising the offsets. The present annex is that vehicle. Its contribution is not a global figure — which the report explicitly withdraws — but a disciplined attribution procedure, applied to a case where the contractual data may, in time, exist: Tunisia, treated as a case nested within a regional panel of sixteen countries (Ugazio & Xin, 2024). The full international extension — comparative panel and census of the 44 least developed countries — is the subject of Annex J of this report; the present annex constitutes its identification core, executed on the regional panel.¶ 2

G.2 Data and feasibility audit. The panel brings together sixteen countries of the Middle East and Central Asia (Ugazio & Xin, 2024, IMF WP 24/14), including Tunisia by name. Public coverage (World Bank interfaces, IDS and WDI) reaches 95 to 100% for Tunisia on structural exposure — debt, debt service, reserves, investment, remittances, exports, direct investment — but at annual frequency. Two structural gaps remain: the foreign-currency share of the debt (the weight of mechanisms A1 and A2) and concessional or multilateral debt (the development finance channel) are not public, and fall under the Ministry of Finance register and OECD-DAC data; the future service tables (QEDS Tables 3-4) are unavailable for Tunisia. The stratification takes account of the fact that the six Gulf exporters report no debt under IDS: the panel is stratified — sixteen countries for the causal estimation, ten debtors for the cash-flow mechanisms A1 to A7, the Gulf being null by construction on those mechanisms. One framing fact, descriptive and non-causal, emerges from this audit: Tunisia's Guidotti ratio stands at 0.59, the lowest in the panel — its reserves cover only 59% of its short-term external debt, a recipient profile of high refinancing vulnerability, consistent with the premise of the report.¶ 3

G.3 Identification strategy. Monetary policy shocks are identified at high frequency: for the Federal Reserve, the decomposition of Jarociński & Karadi (2020), which separates the pure policy surprise from the information component; for the European Central Bank, the EA-MPD event database (Altavilla et al., 2019), whose factor rotation (target, timing, forward guidance, balance-sheet measures) remains to be executed. Exposure is measured in a predetermined manner (lagged one period) and winsorised. Estimation rests on panel local projections (Jordà, 2005; inference from Montiel Olea & Plagborg-Møller, 2021), in two equations — average response, then differential response interacting the shock with the exposure. The three registers of the report (cash flow; balance sheet as memo item; real and welfare) are never added together. Inference combines clustering and wild bootstrap, simultaneous confidence bands, false discovery rate control (Benjamini & Hochberg, 1995) and placebo tests.¶ 4

G.4 Identification diagnostic. The annual reduced-form identification is run as a stress test. On a grid of thirty-six tests — two central banks, three exposure measures, two outcome variables, three horizons — seven pass false discovery rate control (at the 0.10 threshold). But the placebo test fails: a subsequent Federal Reserve shock (period t+1) “explains” growth in the present period, with a statistic of the order of −2.7. In annual data, the persistence of policy cycles and that of exposure generate a spurious differential correlation that fixed effects do not absorb; the seven survivors are artefacts. The conclusion is clear, and it is demonstrated rather than postulated: the annual reduced-form shortcut is invalid for causal inference, and high-frequency event identification — narrow windows around announcements, on the shocks already ingested — is not optional. The only regular raw signal, non-causal, is refinancing exposure measured by the short-term debt share. The value of this phase lies in the fact that a complete and disciplined pipeline — false discovery control, then placebo — rejects false positives instead of publishing them.¶ 5

G.5 What remains to be established. Three dependencies condition the full causal estimation: high-frequency outcome variables (exchange rate and monthly sovereign spreads around announcements); access to the Ministry of Finance debt register for the foreign-currency share and the contractual flows (mechanisms A1 and A2 in cash flow); and finally the factor rotation of the EA-MPD database, the additional global controls (the oil price is already incorporated; the volatility index remains to be connected), the neutralisation of transfers and the bridge to welfare loss (Register III).¶ 6

G.6 Reproducibility. The whole processing chain is version-controlled — from the availability audit to estimation, by way of shock ingestion, the exposure matrix, panel assembly and the local projections — and accompanied by a provenance register (eighteen dated series). Raw and confidential data remain outside the repository; the derived series and the outputs are preserved. The detail is set out in the research record of the report.¶ 7

@incollection{yaiche_hc_annexe_g,
  author    = {Yaïche, Nizar},
  title     = {Annex G — Protocol and empirical diagnostic},
  booktitle = {The Humanity Compact},
  year      = {2026},
  version   = {2.3},
  doi       = {10.5281/zenodo.22746533},
  publisher = {Zenodo},
  url       = {https://humanitycompact.org/ouvrage/annexe-g/}
}

Ouvrage de Nizar Yaïche, ancien ministre des Finances de la République tunisienne. Qui est l’auteur ?