木瓜影院

The DSF standardized stress tests apply temporary shocks to a set of macroeconomic variables. The magnitude of these shocks is based on the historical behavior of these variables in each country. Specifically, most stress test shocks set the relevant variable (GDP growth, export growth, primary balance, and current transfers and FDI flows) at one standard deviation below its historical average for two consecutive years. Both the averages and the standard deviations are calculated using each country鈥檚 historical annual data over the most recent ten-year period. The exchange rate shock is the exception, which is defined as a one-off 30 percent nominal depreciation, or the size needed to close the estimated real exchange rate overvaluation gap, (on top of the baseline projections) of the domestic currency for all countries regardless of their historical exchange rate movements.

However, to address possible bias from historical averages, there is an alternative approach to derive post-shock values from baseline projections. The baseline projections turned out to have better predictive power for future performance than historical averages, having smaller median forecast errors as well as smaller standard deviations in forecast errors. However, basing the shock scenario solely on baseline projections may have shortcomings. In particular, it would not work as envisaged when baseline projections are too optimistic. An effective approach is to set post-shock values at one standard deviation below the historical average, or the projected value under the baseline, whichever is lower, so to minimize possible biases from historical averages, while at the same time limiting the potential bias from overly optimistic baseline projections.

An event analysis was also conducted to evaluate the appropriateness of the assumed duration of shocks. Using the WEO database, the study compares the shocks used in the current DSF with the distribution of outturns over the 1995-2015 period to assess whether the duration of stress test shocks was adequate. Specifically, the study first identified 鈥渆vents鈥� where an outturn has fallen below a shock as used in the existing DSF (i.e., 10-year historical average minus one standard deviation). Results indicate that a shock duration of about 2 years remains broadly appropriate.

If you would like to find more information check out the DSF Policy Paper

Interactions of key macro variables under the standardized stress tests were identified using event studies and complemented with the literature. The event study analyzed the behavior of key variables 鈥� including FX nominal depreciation, real GDP growth, GDP deflator, and export growth 鈥� under the relevant economic shocks. To remove the impact of outliers and appropriately capture the characteristics of the interactions, events are defined as cases where an outturn falls near the relevant cutoff for shocks (i.e. its historical average minus one standard deviation). For example, events for the growth shock are defined as those where actual GDP growth falls between a historical average of minus 1.25 and 0.75 standard deviations. Based on this sample of events, elasticities were calculated. These calculations suggest that inflation decreases with an elasticity to real growth of 0.6 and real GDP growth falls with an elasticity to exports of 0.8. Exchange rate pass-through to the GDP deflator is estimated at 0.3 in the year of the shock (Table AI.5). Other interactions are taken from the literature (e.g. Aisen and Hauner (2008) on the domestic borrowing costs-primary balance interaction).

If you would like to find more information check out the DSF Policy Paper

Bias in historical averages appears to be the main driver for the underestimated primary balance shock. As post-shock values in DSA stress tests are set at one standard deviation below historical averages, a key assumption is that the 10-year historical averages and standard deviations are good predictors for the performance of the relevant indicators in the projection years. This assumption did not hold for the primary balance: outturns of primary balances turned out to be lower than historical averages in about 2/3 of all the observations (Table AI.2). On average, the actual primary balance to GDP ratio was 1.2 percentage points lower than historical averages.

To correct for the possible bias from historical averages, an alternative approach would be to derive post-shock values from baseline projections. The baseline projections turned out to have better predictive power for future performance than historical averages, having smaller median forecast errors as well as smaller standard deviations in forecast errors (Table AI.3). However, basing the shock scenario solely on baseline projections may have shortcomings. In particular, it would not work as envisaged when baseline projections are too optimistic. An effective approach would be to set post-shock values at one standard deviation below the historical average, or the projected value under the baseline, whichever is lower, so to minimize possible biases from historical averages, while at the same time limiting the potential bias from overly optimistic baseline projections. Indeed, calculations using the same sample show that once this approach is applied, the post-shock value of primary balance would be at about the 16th percentile of the distribution of actual outturn, in line with other shocks.

If you would like to find more information check out the DSF Policy Paper