Quantitative Analysis /
Global

External Liquidity Index – Q4 update

  • We update our external liquidity scorecard for 66 emerging and frontier markets to assess the risk of a BOP crisis

  • The most vulnerable (excl. those in default) are Ethiopia, Mozambique, Rwanda, Laos, Mongolia, Bolivia and Tunisia

  • The least vulnerable (excl. Russia) are Kuwait, Saudi Arabia, Iraq, Papua New Guinea, Azerbaijan, Peru and Qatar

External Liquidity Index – Q4 update
Tellimer Data
4 November 2022
Published byTellimer Data

Following the release of the IMF’s updated October 2022 WEO Database earlier this month, we have updated our External Liquidity Index to assess the risk of an external debt or balance of payments crisis across a sample of emerging and frontier markets (following last week’s update to our Debt Sustainability Index). We expand the index from 12 to 17 variables by including both FX and total reserves for all reserve-related variables. We also expand the sample to cover all JP Morgan Emerging Market Bond Index (EMBI) constituents (excluding the CFA franc zone countries with pooled reserves), plus several other countries of interest, increasing our sample from 43 to 66 countries.

We score each variable in terms of standard deviations better (worse) than the sample median and take the simple average across variables to arrive at a composite external liquidity index. The resulting output is a quick and dirty way to quantify the risk of an external debt or balance of payments crisis for each country on a relative basis:

Composite External Liquidity Index

Download the full External Liquidity Index dataset here:

Download .xlsx

Below we present the key takeaways from our latest index update.

Results

After excluding countries that have already defaulted (Sri Lanka, Zambia, and Lebanon), we find that the most vulnerable countries are Ethiopia, Mozambique, Rwanda, Laos, Mongolia, Bolivia and Tunisia. On the other hand, the least vulnerable (excluding Russia) are Kuwait, Saudi Arabia, Iraq, Papua New Guinea, Azerbaijan, Peru and Qatar.

A simple scatter plot shows a negative relationship between the overall external liquidity score and the country risk premium (measured by the Bloomberg EM Sovereign OAS), as we should expect. A linear line of best fit is statistically significant at the 99% confidence level with an R2 of 33%, which rises to 43% with a log-linear line of best fit. Excluding the countries that are in the process of restructuring (Ethiopia, Lebanon, Sri Lanka and Zambia), the relationship is still statistically significant at a 99% confidence level with a smaller R2 of 14% for the linear model and 24% for the log-linear model.

External liquidity index versus OAS

The better fit of the log-linear model and the drop in explanatory power when extreme cases are excluded indicate that the model is best at flagging cases of extreme stress (which also holds for our debt sustainability index) and that vulnerability (as measured by spreads) doesn’t change much when liquidity variables weaken from a strong base but increase exponentially as liquidity indicators deteriorate further.

We also backtest our model by regressing the external liquidity score from the May update by the change in spread for each country since the last WEO was published on 19 April (notably this is before the index was expanded from 43 to 66 countries). We find that the external liquidity score explained 36% of the change in spreads over the subsequent 6-month period at a 99% confidence level, with each standard deviation difference in score translating to an 801bps difference in performance. When Sri Lanka and Ethiopia are excluded, however, the R2 drops to 11% and the confidence level to 90%, with each standard deviation difference in score translating to a 471bps difference in performance.

Spread change versus previous external liquidity score

Notably, the relationship between the external liquidity index and spreads is much stronger than it is for the debt sustainability index (which had an R2 of 12-31% versus 24-43% for the log-linear external liquidity model). Likewise, the external liquidity score in May was a much better indicator of performance over the next six months, with the debt sustainability index explaining a much smaller proportion of the spread change over that period (16% versus 36% for the external liquidity index) and the relationship breaking down entirely when extreme cases were excluded. This suggests that external liquidity variables are a better predictor of near-term vulnerability (as measured by credit spreads) and asset price performance (as measured by spread changes) than debt sustainability variables.

That said, the two indexes can be combined to identify countries at the greatest risk of debt distress, with the debt sustainability risk proxying for “solvency” risk and the external liquidity index proxying for “liquidity” risk. The chart below displays the 48 countries that have scores for both indexes, with those in the lower left quadrant at the greatest risk of distress.

External liquidity and debt sustainability scatterplot

Methodological issues

While we appreciate our model is highly stylised, the simplified and transparent approach is part of its appeal and we find it mostly offers intuitive results (and where it does not, this can be a signal for further investigation). Of course, we urge our readers to take this data and its conclusions with a pinch of salt, and caution that it should be used in conjunction with traditional country risk analysis.

We also recognise some drawbacks of this approach. It fails to account for potential non-linearities and threshold effects within variables, while equal weighting may ignore potential differences in importance. In addition, more timely and thorough data can be found for many countries using official sources, providing a more complete snapshot (but making cross-country comparison more difficult).

Data availability and the vintage of the available data is another challenge, though most of our indicators were deliberately chosen for their forward-looking or high-frequency nature. The longer the lag, and less contemporaneous the data is, the less useful it is as an early warning indicator.

We also omit other indicators which might be a cause of or signal distress, including political and institutional factors. However, we have chosen a more general approach to allow for cross-country comparison, and think our scorecard serves as a useful warning light for external stress.

We welcome feedback from our readers on methodology and coverage and remain available to answer any questions. 

Appendix: Data explanations and sources

Gross FX reserves (months of import): Common international reserve benchmark, calculated in this instance by dividing gross foreign exchange reserve holdings by the trailing 12-month average of monthly imports (rather than forward looking 12-month estimate of goods and services imports, which is more complete but for which data was lacking).

Source: IMF International Financial Statistics (via Bloomberg) for reserves and IMF Direction of Trade Statistics (via Bloomberg) for imports. Frequency: Monthly (ranges from March 2022 to September 2022 for reserves and June 2022 for imports)

Total reserves (months of import): Same as above, but with total reserves (FX reserves + gold + other reserve assets) in numerator.

Gross FX reserves (% of ARA metric): Gross FX reserves as % of the IMF’s Assessing Reserve Adequacy metric (see here for details). IMF benchmark is 100-150% of ARA metric.

Source: IMF Assessing Reserve Adequacy DataMapper. Frequency: Annual (2022) for ARA metric and monthly (ranges from March 2022 to September 2022) for FX reserves.

Total reserves (% of ARA metric): Same as above, but with total reserves (FX reserves + gold + other reserve assets) in numerator.

External principal payments / FX reserves: Public and public guaranteed principal payments due to external creditors in 2022 relative to gross FX reserves.

Source: World Bank International Debt Statistics. Frequency: Annual (2023).

External principal payments / Total reserves: Same as above, but with total reserves (FX reserves + gold + other reserve assets) in denominator.

Bond principal payments / FX reserves: Public and public guaranteed principal payments due to commercial external bondholders (incl. Eurobonds) in 2023 relative to gross FX reserves.

Source: World Bank International Debt Statistics. Frequency: Annual (2023).

Bond principal payments / Total reserves: Same as above, but with total reserves (FX reserves + gold + other reserve assets) in denominator.

External debt service / exports: Total public and public guaranteed debt service (principal + interest) relative to 12-month trailing exports of goods. Exports of goods, services, and remittances may be a more useful denominator, subject to availability.

Source: World Bank International Debt Statistics for external debt service and IMF Direction of Trade Statistics (via Bloomberg) for goods exports. Frequency: Annual (2023) for external debt service and monthly (June 2022) for exports.

External debt service / revenue: Total public and public guaranteed debt service (principal + interest) relative to projected government revenue.

Source: World Bank International Debt Statistics for external debt service and IMF WEO for government revenue. Frequency: Annual (2023).

Option-adjusted spread (bps): Option-adjusted spread from Bloomberg EM Sovereign Index. Proxy for refinancing risk, with spreads above 1,000bps implying limited market access and potential difficulties refinancing external obligations.

Source: Bloomberg EM Sovereign Index. Frequency: Daily (31 October 2022).

Real effective exchange rate (vs 10-year average): Proxies exchange rate over/undervaluation in lieu of detailed assessment of equilibrium REER (see here and here for detailed methodology and estimates).

Source: Bruegel (via Haver). Frequency: Monthly (September 2022). Note: Based on CPI differential with 51 largest trading partners.

Current account balance / GDP: Projected external funding needs (sources) arising from the current account deficit (surplus) as % of GDP.

Source: IMF WEO. Frequency: Annual (2023).

Current account gap / GDP: Difference between “cyclically adjusted CA” (proxied by the projected 2023-27 average, without cyclical adjustment) and the “CA norm” (proxied by 2010-19 average). Large deviations between the cyclically adjusted CA and CA norm point to external imbalances that must be resolved through BoP consolidation or exchange rate devaluation (see here and here for detailed methodology and estimates).

Source: IMF WEO. Frequency: Annual (2010-27).

GEFR / FX reserves: Projected gross external financing requirement (GEFR) for 2023 derived by adding (subtracting) the current account deficit (surplus) to (from) estimated amortizations of medium and long-term (MLT) external debt, divided by gross FX reserves. Some sources exclude short-term external debt in their GEFR calculation and some net out FDI, but we exclude both due to data limitations. 

Sources: IMF WEO for current account and World Bank IDS DataBank for amortizations. Frequency: Annual (2023).

GEFR / Total reserves: Same as above, but with total reserves (FX reserves + gold + other reserve assets) in denominator.

NIIP / GDP: Net international investment position (foreign assets less foreign liabilities) as % of GDP. Proxy for risk of capital flight. We prefer non-resident holdings of domestic government debt or portfolio investment liabilities, but use this as a proxy due to data limitations.

Source: IMF International Investment Position for NIIP and IMF WEO for GDP. Frequency: Annually or quarterly for NIIP (2021 or Q2 2022) and annual for GDP (2021 or 2022).