Currency Crises in the
Post-Bretton Woods Era

A D A T A S E T O F L A R G E D E P R E C I A T I O N S

The Culiuc & Park (2025) Dataset of Large Depreciations identifies and characterizes 562 large exchange rate depreciations across 152 countries between 1971 and 2024.

DAtaset construction

Large depreciation events are identified based on whether the depreciation in the bilateral end-of-month exchange rate against the US dollar is large and accelerating on both quarterly and annual bases. The precise month of the large depreciation event is pinpointed based on the size of the monthly depreciation and whether the abandonment of a peg is detected.

A large depreciation episode covers 24 months starting from the month of a large depreciation event, treating any subsequent events within the 24-month window as aftershocks rather than beginnings of new episodes.

Key concepts

Laree depreciation episodes are characterized across three exchange rate measures: bilateral vis-à-vis the US dollar, nominal effective (NEER), and real effective (REER). For each of these measures, we compute the size of maximum depreciation and its timing, equilibrium depreciation (using various approaches, see Annex I in the paper), overshooting, and number of aftershocks. All exchange rates—bilateral, NEER, REER—are normalized to 100 at T=0 and defined such that an increase denotes a depreciation. The figure summarizes the key concepts using the 1997 Thailand episode as example, focusing on the REER.

Episodes are also classified across multiple dimensions: income level of the country, trajectory of the real exchange rate, presence of aftershocks, exchange rate flexibility before/after initial exchange rate shock, and IMF-supported program status.

Download the dataset

The dataset comprises two Stata files:

Episodes.dta contains one observation per episode, recording country and date (year-month) of the onset of the episode, episode characteristics (size of overshooting, etc.) and classification across multiple dimensions (income group, etc.).
Panel.dta contains 48 monthly observations per episode (from T=−23 to T=24, where T=1 marks the onset of the large depreciation episode), tracking exchange rate and CPI dynamics, exchange rate flexibility before/after the onset of the event, IMF-supported program status, and timing of aftershocks (if any).

Compared to Other Currency Crises Datasets

Many existing datasets are based on annual exchange rate data (Frankel and Rose 1996, Laeven and Valencia 2013/2020, Reinhart and Rogoff 2014, IMF 2015).
Our dataset improves upon other datasets based on monthly data (Cavallo and others 2005, De Gregorio 2016) by (i) filtering out catch-ups to a previous trend and rebounds after short-lived appreciations and (ii) more accurately dating the large depreciations.

Recommended Citation

Culiuc, A. and Park, H., 2025. Currency Crises in the Post-Bretton Woods Era: A New Dataset of Large Depreciations. IMF Working Paper No. 25/221.

Inquiries and Suggestions

Alexander Culiuc (ACuliuc@imf.org) and Hyunmin Park (HPark3@imf.org).

Page updated

Google Sites

Report abuse