The dataset contain 36 devaluation episodes, all of which featured a minimum 15% nominal depreciation over six months less-than-fully offset by inflation (i.e. a real depreciation). They are all pre-2008 so do not include the flight-to-liquidity event associated with the Lehman Brothers collapse.
These events struck developed as well as developing economies; the 1990s capital-account crises are all there. As it happens, real per capita GDP (log terms) makes no difference to the results.
In the language of econometrics, what I’m doing is estimating the following equation:
devaluation = constant + real appreciation + ca deficit + per-capita GDP in constant PPP dollars
where all terms are logs or log-change except ca deficit, which is in percent.
Here are the descriptive statistics for the variables:
| nominal depreciation | prior real appreciation | prior ca deficit | gdp per cap (2005 PPP$) | |
| mean | 0.37 | 0.21 | -2.88 | 13,696 |
| sd | 0.28 | 0.27 | 3.32 | 9,815 |
| min | 0.16 | 0.00 | -9.73 | 793 |
| max | 1.30 | 1.52 | 3.97 | 34,233 |
| N | 36.00 | 36.00 | 36.00 | 36.00 |