## CONCLUSION — wpiea2019247-print-pdf ## Source details **Canonical URL:** [CONCLUSION — wpiea2019247-print-pdf](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019247-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2019/wpiea2019247-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019247-print-pdf.pdf.json) --- ### Study scope and methodology - Studied the determinants of growth in the Eastern Caribbean Currency Union (ECCU) using the Growth at Risk (GaR) framework and with a focus on financial variables. - Broadened determinants to include external and domestic financial factors in addition to external and real conditions emphasized in past efforts. - Used the GaR methodology to capture risks to the central growth projection. ### Key findings - Excessive bank credit growth is associated with negative real GDP growth in the medium term, especially on the tail end of growth distribution in the ECCU. - Domestic Financial Soundness Indicators (FSIs), described as mainly backward looking indicators, have limited association with future growth. - The health of the banking system would have a direct effect on credit provision to the economy and therefore on economic growth. - Worsening global financial conditions are associated with lower future growth in the short term. - Worsening external conditions are associated with lower future growth both in the short-term and in the medium-term, especially at the higher end of growth distribution. - Natural disasters are not associated with future growth one way or the other. - Country-specific results are broadly in line with ECCU-wide results, although there was some country-specific variation potentially due to strong Citizenship-By-Investment program inflows and non-inclusion of credit unions from the financial sector due to data unavailability. - Graphical coefficient/value annotations present in the source text (preserved verbatim): - -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.10.25 0.50.750.9 mean — Credit Growth - -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0.1 0.250.50.750.9 mean — Financial Soundness Indicators - -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.10.250.50.750.9 mean — External Real Conditions - -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.10.250.50.750.9 mean — Global Financial Conditions - -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.10.250.50.750.9 mean — Natural Disasters ### Data limitations and country-specific considerations - Non-inclusion of credit unions in the financial sector analysis due to data unavailability limits assessment; credit unions play an important role for financial intermediation in some ECCU countries. - Recommendation that credit union data be included, at least for those countries where they play an important role, when analyzing the forecasting power of credit growth and FSIs. - Analysis highlights that improvement in national account statistics is warranted. ### Policy implications and recommendations - Strengthen banking supervision. - Make progress toward implementing Basel II/III. - Work toward establishing a financial stability and macroprudential policy framework by: - Enhancing institutional arrangements. - Developing indicators to support financial stability assessment. - Include credit unions in financial sector assessments given rapid credit growth in these institutions. - Improve national accounts/statistics to support analysis and forecasting. *CONCLUSION — wpiea2019247-print-pdf* ### Annex III. Country-Specific Results — wpiea2019247-print-pdf ### Common methodological notes across country panels - Regression coefficients represent a percentage point change in the 12 quarters ahead real GDP growth forecast associated with a one standard deviation increase in bank credit growth rate. - Real GDP growth is measured in annualized quarterly compounded growth rate. - Error bars represent 10 percent confidence intervals. - Growth at Risk plotted at 5 percentile, 50 percentile, and 95 percentile. - Figures across country panels consistently present: Time Series of Partitioned Variables; Correlation between Selected Partitions and their Components; 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean); 3-year Ahead Quantile Regression Coefficients (same quantiles); and Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized). - See IMF WP/19/36 for further description of empirical methodology. ### Antigua and Barbuda - Figures provided: - Time Series of Partitioned Variables (series dated: 6/1/2001, 4/1/2002, 2/1/2003, 12/1/2003, 10/1/2004, 8/1/2005, 6/1/2006, 4/1/2007, 2/1/2008, 12/1/2008, 10/1/2009, 8/1/2010, 6/1/2011, 4/1/2012, 2/1/2013, 12/1/2013, 10/1/2014, 8/1/2015, 6/1/2016, 4/1/2017) - Correlation between Selected Partitions and their Components (axes include numeric ranges: -3 to 4, -2.5 to 2.5, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, and Realized plotted across dates: 6/1/2001 … 6/1/2016) - Displayed component numeric labels (examples preserved): -2.5, -2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2, 2.5; -3, -2, -1, 0, 1, 2, 3, 4; -40, -30, -20, -10, 0, 10, 20, 30, 40. ### Dominica - Figures provided: - Time Series of Partitioned Variables (dates include 6/1/2001 … 12/1/2017) - Correlation between Selected Partitions and their Components (axes include numeric ranges: -3 to 4, -2.5 to 4, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized plotted across dates: 6/1/2001 … 6/1/2016) - Country-specific note: "Capital ratios are excluded from FSI given their unavailability prior to 2005, to secure the same length of time series as other countries." - Growth at Risk vertical axis examples include -10 to 25. ### Grenada - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -2 to 2.5, -2 to 3, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized plotted across dates: 6/1/2001 … 6/1/2016) - Growth at Risk vertical axis examples include a span from -15 to 25. ### St. Kitts and Nevis - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -3 to 2.5, -2 to 2.5, -2.5 to 3, -2 to 6, -0.6 to 0.6 for some component correlations) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Country-specific notes: - "Natural disaster dimension is not considered because there are no recorded natural disasters in EM-DAT between 2001-17 in St. Kitts and Nevis." - Variant analysis: "St. Kitts and Nevis, with CBI revenue as additional growth predictor" — figures include 1-year and 3-year Ahead Quantile Regression Coefficients with CBI Inflows as a predictor. - Variant Growth at Risk axis examples include -15 to 20. ### St. Lucia - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -2.5 to 3, -2 to 2.5, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Growth at Risk axes examples include a span from -20 to 15. ### St. Vincent and the Grenadines - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges similar to other countries) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Growth at Risk axis examples include a vertical span from -20 to 25. *CONCLUSION — wpiea2019247-print-pdf* ### CONCLUSION ### CONCLUSION ### Study scope and methodology - Studied the determinants of growth in the Eastern Caribbean Currency Union (ECCU) using the Growth at Risk (GaR) framework and with a focus on financial variables. - Broadened determinants to include external and domestic financial factors in addition to external and real conditions emphasized in past efforts. - Used the GaR methodology to capture risks to the central growth projection. ### Key findings - Excessive bank credit growth is associated with negative real GDP growth in the medium term, especially on the tail end of growth distribution in the ECCU. - Domestic Financial Soundness Indicators (FSIs), described as mainly backward looking indicators, have limited association with future growth. - The health of the banking system would have a direct effect on credit provision to the economy and therefore on economic growth. - Worsening global financial conditions are associated with lower future growth in the short term. - Worsening external conditions are associated with lower future growth both in the short-term and in the medium-term, especially at the higher end of growth distribution. - Natural disasters are not associated with future growth one way or the other. - Country-specific results are broadly in line with ECCU-wide results, although there was some country-specific variation potentially due to strong Citizenship-By-Investment program inflows and non-inclusion of credit unions from the financial sector due to data unavailability. - Graphical coefficient/value annotations present in the source text include the following numeric sequences and labels (preserved verbatim): - -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.10.25 0.50.750.9 mean — Credit Growth - -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 0.1 0.250.50.750.9 mean — Financial Soundness Indicators - -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.10.250.50.750.9 mean — External Real Conditions - -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.10.250.50.750.9 mean — Global Financial Conditions - -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.10.250.50.750.9 mean — Natural Disasters ### Data limitations and country-specific considerations - Non-inclusion of credit unions in the financial sector analysis due to data unavailability limits assessment; credit unions play an important role for financial intermediation in some ECCU countries. - Recommendation that credit union data be included, at least for those countries where they play an important role, when analyzing the forecasting power of credit growth and FSIs. - Analysis highlights that improvement in national account statistics is warranted. ### Policy implications and recommendations - Strengthen banking supervision. - Make progress toward implementing Basel II/III. - Work toward establishing a financial stability and macroprudential policy framework by: - Enhancing institutional arrangements. - Developing indicators to support financial stability assessment. - Include credit unions in financial sector assessments given rapid credit growth in these institutions. - Improve national accounts/statistics to support analysis and forecasting. *CONCLUSION — wpiea2019247-print-pdf* ### Annex III. Country-Specific Results ### Annex III. Country-Specific Results ### Antigua and Barbuda - Figures provided: - Time Series of Partitioned Variables (series dated: 6/1/2001, 4/1/2002, 2/1/2003, 12/1/2003, 10/1/2004, 8/1/2005, 6/1/2006, 4/1/2007, 2/1/2008, 12/1/2008, 10/1/2009, 8/1/2010, 6/1/2011, 4/1/2012, 2/1/2013, 12/1/2013, 10/1/2014, 8/1/2015, 6/1/2016, 4/1/2017) - Correlation between Selected Partitions and their Components (axes include numeric ranges: -3 to 4, -2.5 to 2.5, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients (quantiles shown: 0.1, 0.25, 0.5, 0.75, 0.9, and "mean") - 3-year Ahead Quantile Regression Coefficients (same quantiles) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, and Realized plotted across dates: 6/1/2001 … 6/1/2016) - Key methodological notes and numeric elements: - Regression coefficients represent a percentage point change in the 12 quarters ahead real GDP growth forecast associated with a one standard deviation increase in bank credit growth rate. - Real GDP growth is measured in annualized quarterly compounded growth rate. - Error bars represent 10 percent confidence intervals. - Growth at Risk plotted at 5 percentile, 50 percentile, and 95 percentile. - Reference to IMF WP/19/36 for further description of empirical methodology. - Displayed component numeric labels (examples preserved as in figures): -2.5, -2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2, 2.5; -3, -2, -1, 0, 1, 2, 3, 4; -40, -30, -20, -10, 0, 10, 20, 30, 40. ### Dominica - Figures provided: - Time Series of Partitioned Variables (dates include 6/1/2001 … 12/1/2017) - Correlation between Selected Partitions and their Components (axes include numeric ranges: -3 to 4, -2.5 to 4, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - 3-year Ahead Quantile Regression Coefficients (same quantiles) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized plotted across dates: 6/1/2001 … 6/1/2016) - Country-specific numeric/methodological notes: - "Capital ratios are excluded from FSI given their unavailability prior to 2005, to secure the same length of time series as other countries." - Regression coefficient and Growth at Risk definitions identical to those above (12 quarters ahead, one standard deviation, annualized quarterly compounded growth rate, error bars = 10 percent confidence intervals). - Displayed axis numeric examples: -2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2, 2.5, 3; -10 to 25 on Growth at Risk vertical axis. ### Grenada - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -2 to 2.5, -2 to 3, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - 3-year Ahead Quantile Regression Coefficients (same quantiles) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized plotted across dates: 6/1/2001 … 6/1/2016) - Methodological numeric notes repeated: - Coefficients = percentage point change in the 12 quarters ahead real GDP growth forecast per one standard deviation increase in bank credit growth rate. - Real GDP growth measured in annualized quarterly compounded growth rate. - Error bars = 10 percent confidence intervals; Growth at Risk plotted at 5 percentile, 50 percentile, 95 percentile. - Displayed axis numeric examples include ranges: -1.4 to 0.9 for some components; Growth at Risk vertical axis spanning -15 to 25. ### St. Kitts and Nevis - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -3 to 2.5, -2 to 2.5, -2.5 to 3, -2 to 6, -0.6 to 0.6 for some component correlations) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Country-specific numeric/methodological notes: - "Natural disaster dimension is not considered because there are no recorded natural disasters in EM-DAT between 2001-17 in St. Kitts and Nevis." - A variant analysis: "St. Kitts and Nevis, with CBI revenue as additional growth predictor" — figures include: - 1-year Ahead Quantile Regression Coefficients (quantiles as above) - 3-year Ahead Quantile Regression Coefficients including CBI Inflows as a predictor (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Regression/coefficient definitions and measurement identical to other countries (12 quarters ahead, one standard deviation, annualized quarterly compounded growth rate, error bars = 10 percent confidence intervals). - Displayed numeric examples include Growth at Risk axis from -15 to 20 in variant plot. ### St. Lucia - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges: -2.5 to 3, -2 to 2.5, -2.5 to 3, -2 to 6, -1 to 9) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Methodological numeric notes consistent with other country panels: - Coefficients correspond to percentage point change in the 12 quarters ahead real GDP growth forecast for a one standard deviation credit growth shock. - Real GDP growth measured in annualized quarterly compounded growth rate; error bars = 10 percent confidence intervals. - Growth at Risk axes examples include a span from -20 to 15. ### St. Vincent and the Grenadines - Figures provided: - Time Series of Partitioned Variables (dates 6/1/2001 … 4/1/2017) - Correlation between Selected Partitions and their Components (axes numeric ranges similar to other countries) - 1-year Ahead Quantile Regression Coefficients and 3-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean) - Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized) - Methodological numeric notes: - Same regression interpretation: coefficients = percentage point change in the 12 quarters ahead real GDP growth forecast associated with a one standard deviation increase in bank credit growth rate. - Real GDP growth measured in annualized quarterly compounded growth rate; error bars represent 10 percent confidence intervals. - Displayed Growth at Risk axis examples include a vertical span from -20 to 25. *Figures across country panels consistently present: Time Series of Partitioned Variables, Correlation between Selected Partitions and their Components, 1-year Ahead Quantile Regression Coefficients (quantiles: 0.1, 0.25, 0.5, 0.75, 0.9, mean), 3-year Ahead Quantile Regression Coefficients (same quantiles), and Historical 1-year Ahead Growth at Risk and Realized Growth (var5%, var50%, var95%, Realized). Regression coefficients represent percentage point changes in the 12 quarters ahead real GDP growth forecast per one standard deviation increase in bank credit growth; real GDP growth is measured in annualized quarterly compounded growth rate; error bars are 10 percent confidence intervals. See IMF WP/19/36 for further description of empirical methodology.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019247-print-pdf.pdf_