## 1ecuea2019002

## Source details

**Canonical URL:** [1ecuea2019002](https://www.imf.org/-/media/files/publications/cr/2019/1ecuea2019002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2019/1ecuea2019002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2019/1ecuea2019002.pdf.json)

---

### Recent Past Performance of Growth Forecasts — stylized facts and oil sector
- Economic activity expanded at an average pace of 3.9 percent per year since 1965.
- Private consumption averaged about 64 percent of GDP over the period.
- Public and private investment showed large decadal swings; since 2007 private investment has been crowded out by the public sector.
- The oil sector:
  - Reached about 15 percent of real GDP at two historical peaks (1973 and 2004).
  - Contributed about 12 percent of real output on average between 1973 and 2004.
  - Declined to 9.9 percent of real output in 2015 (lowest point).
  - The secular decline started before the late-2014 collapse in oil prices; production and exploration lagged regional peers.

### Growth accounting — methodology and historical contributions
- Methodology:
  - Standard Cobb-Douglas production function with contributions from TFP, capital, employment, and human capital.
  - Capital stock via perpetual inventory method with parameters: technological progress rate g = 1.53, population growth n = 2.4, depreciation rate δ = 4.5.
  - Human capital from average years of schooling with parameters θ = 0.32 and φ = 0.58.
- Key historical findings (1970–2017 and related periods):
  - Between 1970 and 2017 the economy grew at an average pace of 4 percent per year; TFP contributed -0.1 percentage points to that average.
  - Over the past 46 years, growth was driven mostly by factor accumulation: roughly 2 percentage points from physical capital and 2 percentage points from labor.
  - Human capital improvements contributed about 0.2 percentage points on average.
- Recent decade (2010–2017) contributions:
  - Capital: 1.8
  - Employment: 2.0
  - Human capital: 0.2
  - TFP: -0.1
  - Total: 3.9

### Oil prices, productivity and investment linkages
- Empirical correlations:
  - Contemporaneous oil price change coefficient with TFP: 0.024* (standard error 0.012).
  - Oil price at t-1 effect on TFP: 0.030** (standard error 0.012).
  - Oil price at t-1 effect on Investment: 0.115*** (standard error 0.038).
  - Regression observations: 50; R-squared: TFP 0.179, Investment 0.161.
- Interpretation:
  - Oil price changes affect TFP contemporaneously and with one-year lag, and affect real investment after one year.
  - The 2014–15 oil price decline transmitted through lower oil-sector investment, reduced government revenues (cutting other investments), and lower export liquidity, contributing to lower productivity growth.

### Long-run potential growth — scenarios and quantitative estimates
- Overall potential growth estimate range: 1¾ to 3 percent.
- Lower estimate (extrapolation of recent trends): 1.8 percent per year.
  - Assumptions: fiscal consolidation via capital expenditure cuts, TFP grows at an average rate of 0.1 percent, human capital improvements continue at recent rates, labor growth aligned with demographic projections and current sluggish labor-market performance.
  - Contribution breakdown (approximate): labor (including human capital) > 50 percent of growth; physical capital accumulation ~32 percent; TFP ~15 percent.
- Higher estimate (reform scenario): about 3 percent per year.
  - Assumptions: policies to increase competitiveness, improve education quality, and create fiscal space to maintain higher public investment; improved business environment and regulatory stability to attract private investment (including FDI); education improvements that boost human capital and TFP.
  - Contribution breakdown: TFP accounts for 17 percent of long-run growth; labor (including human capital) contributes 38 percent; capital accumulation contributes 45 percent.
  - Additional potential: raising women’s labor force participation (current female participation 55.6 percent, 25 points below that of men) would further boost potential growth.

### Business cycle facts, indicators and nowcasting
- Business cycle facts:
  - Expansion peak: 2015Q1; trough: 2016Q1; total decline between peak and trough: 4 percent in output.
  - By 2017Q4 the output gap remained negative at -0.2 percent.
  - Given overvaluation, structural rigidities, and likely fiscal adjustment with capital expenditure cuts, the output gap is not expected to close over the medium term.
- Leading and coincident indicators:
  - Best leading indicators (two quarters lead): exports (merchandise, oil, and non-oil) and private sector deposits.
  - Useful one-quarter lead indicators: international oil prices, imports of merchandise, and fuel.
  - Coincident indicators (no lead): most production sectors (agriculture, oil, communications, etc.).
  - Lags relative to GDP: construction, communications, agriculture, private consumption, gross fixed capital formation, and credit to the private sector tend to lag real GDP growth.
- Nowcasting:
  - Nowcasting models use monthly-frequency variables to improve accuracy and timeliness of real GDP projections.

### Nowcast model setup, selection and performance (Quarterly GDP lag issue)
- Quarterly GDP released with a lag of one quarter; need for preliminary nowcasts based on monthly data to produce a preliminary nowcast by end of second month of each quarter.
- Potential output in one exercise computed using a Hodrick-Prescott filter for period 2000Q1–2023Q4.
- Nowcast model: setup and data
  - Number of monthly-frequency candidate variables: 15.
  - All possible combinations from univariate up to 12 covariates; some excluded to avoid perfect collinearity.
  - Four versions of the general OLS equation estimated (with/without seasonal dummies and with/without lagged covariate terms).
  - Total estimated models: 38,396.
  - Model selection step involved a little over 3.5 million regressions.
  - Sample period for estimation: 2000Q1 to 2016Q3.
  - In-sample nowcasts computed for each quarter between 2009Q1 and 2016Q3 using a rolling window.
  - Each model produces three nowcasts per quarter corresponding to monthly releases; each model produces 93 in-sample nowcasts used to evaluate performance.
  - Nowcast evaluation criterion: lowest root mean square error (RMSE) for the 93 in-sample nowcasts.
- Best single model (model 1) explanatory variables: imports of fuel, imports of capital goods, and income tax collection.
  - Model 1 reduces RMSE by 19.1 percent compared to a “naïve” model using only WTI.
- Ensemble approach:
  - Average nowcast of best four models further reduces RMSE by 2.5 percent compared to model 1.
  - Models one through four form the core average nowcast; models five through eight included when best performing in recent past (2015–16).
  - Alternative versions 5b–8b exclude fiscal variables when fiscal data delayed or unreliable.
- When fiscal data publication delayed for more than 6 months in 2017, models without fiscal variables (models 5b–8b) were prioritized.

### Forecasting GDP with Structural VAR — specification and main results
- Structural VAR endogenous vector: real GDP, real credit to the private sector, real deposits in the banking system, real public gross fixed capital formation, real GDP of advanced economies (AEs), and the oil spot price for WTI. All variables log-transformed.
- Structural identification: Cholesky decomposition with block exogeneity so domestic variables do not affect external variables contemporaneously nor with lags.
- Ordering assumption: external shocks propagate via financial system (deposits and credit) or via government financing and capital investment.
- Seven VAR models explored; Model 4 selected as best for GDP forecasting at multiple horizons.
  - Model 4 includes real deposits and credit along with external variables and GDP.
  - Lag length chosen for Model 4: two lags.
  - Model 4 performance:
    - Very good forecasting 1-step ahead for real GDP and other included variables.
    - 4-steps ahead dynamic forecasts conditional on oil prices and AEs growth show no systematic bias.
    - As early as 2014Q3 forecasts indicated signs of future economic decline that began in 2015Q2; as early as 2015Q4 the model suggested recovery starting in 2016Q3 (actual recovery began a quarter earlier).
    - Actual GDP realization always fell within the 90 percent confidence interval for end-of-sample forecasts.

### RMSEs of real GDP forecasts (2010Q1–2016Q4) — preserved exactly
- Model 1: 0.0066 0.0032 0.0049 0.0065 0.0049
- Model 2: 0.0078 0.0037 0.0055 0.0074 0.0056
- Model 3: 0.0067 0.0032 0.0050 0.0068 0.0050
- Model 4: 0.0065 0.0029 0.0044 0.0060 0.0045
- Model 5: 0.0079 0.0038 0.0058 0.0078 0.0058
- Model 6: 0.0081 0.0036 0.0055 0.0073 0.0055
- Model 7: 0.0067 0.0030 0.0045 0.0062 0.0046
- Average: 0.0072 0.0034 0.0051 0.0069 0.0051

### Uncertainty analysis and fan charts
- Fan charts show 90 percent confidence intervals.
- Parameter and random-error uncertainty intervals computed with 1,000 repetitions.
- Sources of uncertainty examined:
  - Parameter estimation uncertainty and random shocks to projections.
  - Sensitivity to oil price changes using historical deviations in oil price changes to measure uncertainty around WEO oil price projections.
- Findings:
  - Uncertainty from oil prices is lower than that from parameter and random error uncertainties.
  - Large drops in oil prices remain a significant downside risk to growth.
- Practical implication:
  - Best nowcast approach: ensemble average of models one through four, with models five through eight included when outperforming in recent periods.
  - Models excluding fiscal variables (5b–8b) necessary when fiscal data delayed/unreliable.
  - VAR Model 4 provides best GDP forecast performance across horizons and can signal 1–2 year ahead turning points.
  - Fan charts and conditional dynamic forecasts provide quantitative measures of forecast uncertainty and oil-price-related downside risk.

### Appendix I — summary statistics and correlations (selected exact figures)
- Growth rate summary statistics (q-o-q), selected series (Obs / Mean / Std. Dev. / Min / Max):
  - GDP: 660 / 0.97 / 1.1 / -1.9 / 3.3
  - Private Final Consumption: 660 / 0.97 / 1.4 / -3.0 / 5.8
  - General Government Final Consumption: 661 / 1.15 / 2.4 / -7.4 / 10.3
  - Gross Fixed Capital Formation: 661 / 1.69 / 4.3 / -9.6 / 14.0
  - Change in Inventories: 660 / 0.58 / 96.2 / -474.9 / 523.3
  - Exports: 660 / 0.83 / 2.7 / -4.4 / 7.9
  - Imports: 661 / 1.55 / 5.5 / -10.6 / 16.0
  - GDP Oil: 660 / 0.63 / 4.4 / -7.8 / 16.6
  - GDP Non-Oil: 661 / 1.03 / 0.9 / -2.0 / 3.0
  - Index of Economic Activity: 381 / 1.05 / 4.2 / -7.8 / 11.8
  - Consumer Price Index: 661 / 1.90 / 3.2 / -0.1 / 21.2
  - Oil price (WTI spot price): 660 / 0.67 / 16.5 / -70.1 / 32.3
  - Terms of Trade Index: 660 / 0.45 / 9.9 / -47.3 / 22.6
  - Merchandise Exports: 661 / 1.84 / 11.1 / -46.7 / 21.2
  - Total Petroleum Exports: 501 / 1.18 / 22.9 / -80.4 / 50.5
  - Total Non-Petroleum Exports: 502 / 2.3 / 15.5 / -10.2 / 16.6
  - Oil Exports Revenue: 660 / 0.98 / 31.8 / -113.9 / 74.9
  - Value Added Tax (SRI): 663 / 3.1 / 6.6 / -15.7 / 22.3
  - Income Tax (SRI): 665 / 3.2 / 48.2 / -75.3 / 122.0
  - Capital Expenditure: 664 / 4.8 / 37.4 / -97.8 / 87.9
  - Gross Fix Capital Formation (public): 665 / 7.7 / 55.0 / -130.6 / 102.3
  - Credit to the private sector: 423 / 0.72 / 2.8 / -3.5 / 7.6
  - Deposits of the private sector: 423 / 3.2 / 2.7 / -3.1 / 7.8
- Selected correlations of cyclical real GDP growth (t-4 to t+1 highlights):
  - Credit to the private sector: 0.59 (t-4), 0.65 (t-3)
  - Deposits of the private sector: 0.45 (t-4), 0.60 (t-3), 0.52 (t-2)
  - GDP Non-Oil: 0.29 (t-4), 0.77 (t-1), 0.36 (t+1)
  - Petroleum (incl. refining) and Mining: 0.55 (t-4), -0.28 (t-3)
  - Construction: -0.35 (t-4), 0.43 (t-2), 0.51 (t-1), 0.26 (t+1)

### EMBI spreads, fiscal consolidation and reserve adequacy — main findings
- EMBI spreads and fiscal consolidation:
  - An improvement in the cyclically adjusted primary balance (CAPB) of 1 percent of GDP is associated with a decrease in the EMBI spreads on impact of about 17 to 34bps.
  - A reduction of 1 percent of the debt to GDP ratio would help reduce the spread on impact by 4 to 7bps.
  - A country with at most one period of severe debt distress during the sample has on average 140 to 230 higher EMBI spread (lower estimate for oil exporters).
  - Policy implication: credible and prudent fiscal framework can minimize probability of negative credit events and reduce vulnerability to adverse external conditions.
- Quantified scenario for Ecuador:
  - Adoption of discretionary fiscal measures of 5 percentage points of GDP together with a reduction in public debt-to-GDP ratio by 1 percentage points of GDP:
    - The EMBI spread for an oil-producing economy of Ecuador could be permanently lowered by close to 175 basis points.
  - Additional, slower-moving improvements in institutional strength could further reduce the sovereign spread.

### Reserve adequacy in a dollarized economy — risks and metrics
- Dollarization specifics:
  - Reserves correspond only to dollars that represent claims on foreign sources; dollar claims on domestic agents are not part of reserves.
  - Dollarized economies cannot print money to exchange for foreign currency, constraining reserve accumulation.
- Reasons buffers are needed:
  - (i) Balance of Payments shock-absorption,
  - (ii) lender of last resort reasons,
  - (iii) prudential coverage (coverage of private sector claims).
- Liquidity Fund (LF) and BCE balance sheet:
  - LF established in 2009; balance stood at US$2.6 billion at end-2018 (about 2.5 percent of GDP).
  - US$2.6 billion represents 9.3 percent of deposits while the target is 10 percent.
  - LF fully invested abroad (mainly assets issued by BIS and FLAR).
  - LF rule: 30 percent of each bank’s contribution pooled to support small entities; 70 percent reserved for the contributing bank’s potential liquidity needs.
  - LF has not been tapped significantly since creation and could be insufficient in severe system-wide deposit runs.
  - International reserves peaked at US$6.7 billion until September 2014; fell to US$2 billion at end-December 2017, equal to 52 percent of banks deposits at the BCE.
  - TBCs issuance limited (less than US$200 million).
- Banking sector liquidity:
  - Ecuadorian banks’ liquidity ratio initially compared well with dollarized peers but fell below peers beginning in crisis years.
  - Across all exchange rate regimes, Ecuador’s liquidity ratio was near the lower bound.
  - Peer central banks (El Salvador and Kosovo) maintain coverage ratios above 100 percent.
- ARA-EM and findings:
  - A reserve coverage of 100−150 percent of the ARA-EM metric is regarded as adequate.
  - Ecuador’s reserve level has been significantly below most ARA-EM metrics; currently the reserve level in Ecuador is about one-fifth of the ARA metric.
  - Risk Weights in the ARA-EM Metric (Exchange Rate Regimes — Weights in percent):
    - Fixed: Short-term Debt 30, Other Liabilities 20, Broad Money 10, Exports 10
    - Floating: Short-term Debt 30, Other Liabilities 15, Broad Money 5, Exports 5
- Staff’s supplemental absolute reserve-floor metric:
  - Reserves should be sufficient at least to cover:
    - (i) all deposits of the banking system at the central bank;
    - (ii) money issuance (e.g. coins) and issuance of TBCs;
    - (iii) electronic money;
    - (iv) a measure of volatility of public credit; and
    - (v) contingent liabilities.
  - At end-2018, liquid reserves amounted to less than 50 percent of staff’s supplemental metric.
- Policy recommendations:
  - Rebuild reserves and create fiscal mechanisms to protect NIR management and the BCE balance sheet from the fiscal cycle.
  - Maintain adequate reserve cover of private sector claims on the BCE.
  - Reestablish central bank independence and reinstate the four balance sheet accounts at the BCE or implement an equivalent transparent rule to manage reserves.

### Net acquisition of currency and deposits abroad — measurement and interpretation
- BOP observation per authorities:
  - Between 2008 and 2017, residents accumulated US$19 billion in currency and deposits abroad (BOP statistics).
  - Deposits in the domestic banking system increased from 15 to 34 percent of GDP between 2008 and 2017.
  - Errors and omissions averaged 0.26 percent of GDP per year in absolute terms between 2008 and 2017.
- Monetary-side estimates (2008–2017):
  - Private sector deposits held abroad: about US$1.7 billion.
  - Banks’ deposits abroad: net decline of US$0.6 billion.
  - Banks’ vault holdings: added about US$1.1 billion.
  - Increase in U.S. dollars in circulation (cash in circulation): close to US$12 billion (BCE estimate: US$11.6 billion increase over 2008–2017).
- Interpretation and conclusions:
  - Monetary-side evidence suggests the private-sector deposit accumulation recorded in the BOP (US$19 billion) is substantially overstated relative to monetary and BIS data.
  - Sectoral decomposition shows accumulation emanates from the “central bank” sector rather than the private sector; patterns mirror oil trade balance for central bank holdings and non-oil trade balance for private sector holdings.
  - Evidence suggests the accumulation does not indicate private capital flight at material levels, though measurement caveats remain (partial BIS coverage, difficulty measuring currency in circulation, smuggling and under-invoicing).
  - Recommendation: revise authorities’ methodology in line with IMF technical assistance on BOP statistics; such revision would likely result in higher net errors and omissions but provide clearer picture and avoid misleading appearance of large deposit outflows.

### Data gaps, BOP-side distortions and ISD tax as cross-check
- Monetary-side coverage gaps:
  - BIS locational banking statistics limited in Latin America coverage; important partners (China, Vietnam, Colombia, Peru) not fully covered.
  - Under- or over-estimation of U.S. dollars in circulation is possible; BCE estimates assume currency leaving financial system returns.
  - Currency in circulation increased from 6.4 percent of GDP at end-2007 to 14.2 percent of GDP at end-2017, much of the increase since 2014.
- BOP-side distortions:
  - Currency and deposits category used as a residual/contra-entry; inaccurate reporting of trade (exports/imports) can distort the category.
  - Press reports suggest smuggling and under-invoicing may have biased trade reporting, particularly 2014–mid-2017.
- ISD tax as information source:
  - ISD levied on outward foreign transactions; tax base of gross outflows approximated using tax rate and ISD collections corresponds closely to imports.
  - If large legal outflows occurred, inferred tax base would be much larger than imports — not observed.

### Cooperatives, electronic money (EM) and financial system features
- Cooperative sector key facts:
  - Cooperatives represent around 16 percent of assets and 23 percent of deposits in the consolidated financial system.
  - Total assets of cooperative sector ~12 percent of GDP.
  - Currently 618 COACs, down from almost 1,000 in 2012.
  - Number of members has more than doubled since 2013 and stands at about 7.7 million.
  - Segment 1 has US$9.4 billion in assets, representing 69 percent of assets in the coop system.
  - Segment 1 average NPLs of 3.6 percent in September 2018 (loans classified non-performing after 30 days vs 15 days for private banks).
- Safety nets and liquidity funds:
  - Deposit insurance for COACs has about US$331 million available; at five percent of insured deposits in the sector this is insufficient.
  - Between April 2014 and December 2018, deposit insurance paid out US$47.5 million due to 166 liquidations, benefiting 443,114 members.
  - Liquidity fund for cooperatives established in 2016 with initial capital US$40 million; as of end-September 2018 it had US$155 million available, covering about 1.5 percent of deposits in the coop system.
- Electronic money (EM) system (introduced December 2014):
  - EM defined as an electronic payment means managed and regulated by the central bank, denominated in U.S. dollars.
  - Peak stock of EM in the financial system: US$11.8 million in approximately 400,000 registered accounts.
  - Government tax incentives to boost EM diffusion: VAT refund of two percent of purchases using EM; rule that revenue, costs, and expenses incurred with EM could be excluded from income tax base during 2017–19.
  - Adoption stalled amid perceptions EM might be first step to de-dollarize and privacy concerns.
  - Moreno administration shifted EM to private sector; Economic Reactivation Law (December 2017) decreed central bank EM system phased out; by mid-April EM account balances down to zero.
  - Private banks adopted a different mobile payments platform; as of February 2019 still awaiting regulatory approval.
- Policy-relevant points on EM:
  - EM advantages: convenience, time savings, money savings (lower fees), transparency.
  - EM disadvantages: safety risks, privacy concerns, need for internet access.
  - Macroeconomic implications: EM may increase money demand, improve monetary policy effectiveness, and reduce seigniorage if fully adopted.
  - Financial inclusion potential: EM can extend access to finance in remote areas; extension from transfers to savings, credit, insurance could provide additional benefits.

### Debt ceiling calibration and sensitivity analysis — key findings
- Sensitivity of debt ceiling and safety margin:
  - Debt ceiling smaller and safety margin larger when MDL and risk tolerance are smaller; volatility of macro shocks is higher; and government’s fiscal response to debt increases is weaker.
- Calibration parameters and scenarios:
  - MDL range estimate: (32, 47, and 57 percent of GDP).
  - Risk tolerance levels: 5 to 15.
  - Two fiscal reaction function (FRF) alternatives: “past fiscal behavior” and “market-led adjustment”.
  - “Past fiscal behavior” FRF: primary balance (t) = .335* primary balance (t-1) + 0.324 terms of trade gap (t) - .313*external disbursements + 2.194
- Key quantitative thresholds:
  - A prudent debt ceiling for Ecuador would be at most 30 percent of GDP.
  - Ecuador’s current debt ceiling of 40 percent of GDP could be deemed prudent only under risk tolerance levels and debt distress probabilities above 10 percent and if fiscal policy allows for greater adjustment of primary balances in response to debt increases than observed in the past.
- Policy implications:
  - Recent changes to fiscal framework and proposed enhancements expected to promote fiscal savings in good times and use savings to help close market financing gaps in bad times, supporting stronger fiscal behavior consistent with “market-led adjustment” scenario.

*Source: POTENTIAL OUTPUT, GDP NOWCASTING AND FORECASTING (Selected Issues Paper), IMF, March 1, 2019; IMF staff analysis as presented in the source content.*

### 1. Recent Past Performance of Growth Forecasts _________________________________________ 18

### 1. Recent Past Performance of Growth Forecasts

### Stylized facts and recent performance
- Economic activity expanded at an average pace of 3.9 percent per year since 1965.
- Private consumption averaged about 64 percent of GDP over the period.
- Public and private investment showed large decadal swings; since 2007 private investment has been crowded out by the public sector.
- The oil sector:
  - Reached about 15 percent of real GDP at two historical peaks (1973 and 2004).
  - Contributed about 12 percent of real output on average between 1973 and 2004.
  - Declined to 9.9 percent of real output in 2015 (lowest point).
  - The secular decline started before the late-2014 collapse in oil prices; production and exploration lagged regional peers.

### Growth accounting—drivers of long-run growth
- Methodology:
  - Standard Cobb-Douglas production function with contributions from TFP, capital, employment, and human capital.
  - Capital stock constructed via the perpetual inventory method with parameters: technological progress rate g = 1.53, population growth n = 2.4, depreciation rate δ = 4.5. (As used from Ferreira et al (2013) and data sources.)
  - Human capital constructed from average years of schooling using parameters θ = 0.32 and φ = 0.58 (Bils and Klenow (2000) approach).
- Key historical findings (1970–2017 and related periods):
  - Between 1970 and 2017 the economy grew at an average pace of 4 percent per year; TFP contributed -0.1 percentage points to that average.
  - Over the past 46 years, growth was driven mostly by factor accumulation: roughly 2 percentage points from physical capital and 2 percentage points from labor.
  - Human capital improvements contributed about 0.2 percentage points on average.
- Recent decade (2010–2017) contributions (from Table/figure reporting):
  - Capital: 1.8 (for 2010-17 row shown)
  - Employment: 2.0
  - Human capital: 0.2
  - TFP: -0.1
  - Total: 3.9

### Oil prices, productivity and investment linkages
- Empirical correlations show oil prices affect both TFP and investment:
  - Contemporaneous oil price change coefficient with TFP: 0.024* (standard error 0.012).
  - Oil price at t-1 effect on TFP: 0.030** (standard error 0.012).
  - Oil price at t-1 effect on Investment: 0.115*** (standard error 0.038).
  - Regression observations: 50; R-squared: TFP 0.179, Investment 0.161.
- Interpretation:
  - Changes in oil prices affect TFP contemporaneously and with one-year lag, and affect real investment after one year.
  - The 2014–15 oil price decline transmitted through lower oil-sector investment, reduced government revenues (cutting other investments), and lower export liquidity, contributing to lower productivity growth.

### Long-run potential growth—scenarios and quantitative estimates
- Overall potential growth estimate range: 1¾ to 3 percent.
  - Lower estimate (extrapolation of recent trends): 1.8 percent per year.
    - Assumptions: fiscal consolidation via capital expenditure cuts, TFP grows at an average rate of 0.1 percent, human capital improvements continue at recent rates, labor growth aligned with demographic projections and current sluggish labor-market performance.
    - Contribution breakdown (approximate): labor (including human capital) > 50 percent of growth; physical capital accumulation ~32 percent; TFP ~15 percent.
  - Higher estimate (reform scenario): about 3 percent per year.
    - Assumptions: policies to increase competitiveness, improve education quality, and create fiscal space to maintain higher public investment; improved business environment and regulatory stability to attract private investment (including FDI); education improvements that boost human capital and TFP.
    - Contribution breakdown: TFP accounts for 17 percent of long-run growth; labor (including human capital) contributes 38 percent; capital accumulation contributes 45 percent.
    - Additional potential: raising women’s labor force participation (current female participation 55.6 percent, 25 points below that of men) would further boost potential growth.

### Business cycle, nowcasting, and indicators
- Business cycle facts:
  - The economy experienced an expansion peak in 2015Q1 and a trough in 2016Q1, resulting in a total decline of 4 percent in output between peak and trough.
  - By 2017Q4 the output gap remained negative at -0.2 percent (economy still below potential).
  - Given overvaluation, structural rigidities, and likely fiscal adjustment with capital expenditure cuts, the output gap is not expected to close over the medium term.
- Leading and coincident indicators:
  - Best leading indicators (two quarters lead): exports (merchandise, oil, and non-oil) and private sector deposits.
  - Useful one-quarter lead indicators: international oil prices, imports of merchandise, and fuel.
  - Coincident indicators (no lead): most production sectors (agriculture, oil, communications, etc.).
  - Lags relative to GDP: construction, communications, agriculture, private consumption, gross fixed capital formation, and credit to the private sector tend to lag real GDP growth.
- Nowcasting:
  - The paper develops nowcasting models using monthly-frequency variables identified above to improve the accuracy and timeliness of real GDP growth projections.

*Source: POTENTIAL OUTPUT, GDP NOWCASTING AND FORECASTING (Selected Issues Paper), IMF, March 1, 2019.*

### 12.      Quarterly GDP numbers in Ecuador are released with a considerable lag of one quarter

### 12.      Quarterly GDP numbers in Ecuador are released with a considerable lag of one quarter

### Overview
- Quarterly GDP data are released with a lag of one quarter, creating a need for preliminary estimates or “nowcasts” based only on monthly-frequency data so a preliminary nowcast can be available by the end of the second month of each quarter and updated with each monthly release until the official GDP estimate is announced.
- The potential output in one exercise was computed using a Hodrick-Prescott filter for the period 2000Q1–2023Q4.

### Nowcast model: setup and data
- Number of monthly-frequency candidate variables: 15.
- All possible combinations of these variables were created from univariate models up to models with 12 covariates; some combinations were excluded to avoid perfect collinearity.
- General estimated OLS model form (equation (6)) includes quarter-on-quarter (q-o-q) growth rate of real GDP as dependent variable, q-o-q growth rates of candidate indicators as covariates, and seasonal quarterly dummy variables. Four versions of equation (6) were estimated (with/without seasonal dummies and with/without lagged covariate terms).
- Total estimated models based on combinations and the four equation versions: 38,396 models.
- Model selection step involved a little over 3.5 million regressions.
- Sample period for estimation: 2000Q1 to 2016Q3.
- In-sample nowcasts calculated for each quarter between 2009Q1 and 2016Q3 estimating models on a rolling window.
- Each model produces three nowcasts per quarter corresponding to the three monthly data releases; each model therefore produces 93 in-sample nowcasts used to evaluate nowcasting performance.
- Nowcast evaluation criterion: lowest root mean square error (RMSE) for the 93 in-sample nowcasts.

### Nowcast model: best specifications and performance
- Best single model (model 1) explanatory variables: imports of fuel, imports of capital goods, and income tax collection.
- Model 1 reduces RMSE by 19.1 percent compared to a “naïve” model that only uses the WTI oil price as an explanatory variable.
- The best model is not uniformly superior across every year, quarter, or month; hence an ensemble approach is useful:
  - The average nowcast of the best four models further reduces the RMSE by 2.5 percent compared to model 1.
  - Models one through four form the core average nowcast.
  - Models five through eight are included in the average when they are the best performing in the recent past (2015–16).
  - Because of delays and unreliability in fiscal data reporting (e.g., oil revenues, non-financial public-sector capital expenditure), alternative versions 5b–8b (excluding fiscal variables) are considered when fiscal data are not available.
- When fiscal data publication was delayed for more than 6 months in 2017 (and continued delays thereafter), models without fiscal variables (models 5b–8b) were prioritized for periods lacking fiscal data.

### Forecasting GDP (VAR approach)
- Structural VAR model (equation (7)) used for forecasting with endogenous vector including: real GDP, real credit to the private sector, real deposits in the banking system, real public gross fixed capital formation, real GDP of advanced economies (AEs), and the oil spot price for WTI. All variables enter log-transformed.
- Structural identification: Cholesky decomposition with block exogeneity restrictions such that domestic variables do not affect external variables (oil price and AEs GDP) contemporaneously nor with lags.
- Ordering assumption: external shocks propagate via financial system (deposits and credit) or via government financing and capital investment.

### VAR model selection and forecasting performance
- Seven different VAR models were explored; all include external variables (WTI and AEs GDP) and Ecuador’s GDP, differing in domestic variables included.
- Model selected as best for GDP forecasting at multiple horizons: model 4. Model 4 includes real deposits and credit along with the external variables and GDP.
- Lag length chosen for Model 4: two lags (from order selection tests).
- Model 4 performance:
  - Does a very good job forecasting 1-step ahead for real GDP and other included variables (see Figure 8 in source).
  - 4-steps ahead dynamic forecasts are conditional on available information on oil prices and AEs growth and show no systematic bias.
  - As early as 2014Q3 the forecasts indicated signs of the future economic decline that began in 2015Q2; as early as 2015Q4 the model suggested recovery starting in 2016Q3 (actual recovery began a quarter earlier).
  - The actual GDP realization always fell within the 90 percent confidence interval for the end-of-sample forecasts.

### RMSEs of real GDP forecasts (2010Q1–2016Q4)
- Table 2 RMSEs of real GDP forecasts (static 1-step and h-step ahead): numbers preserved exactly as presented.

  - Model 1: 0.0066 0.0032 0.0049 0.0065 0.0049
  - Model 2: 0.0078 0.0037 0.0055 0.0074 0.0056
  - Model 3: 0.0067 0.0032 0.0050 0.0068 0.0050
  - Model 4: 0.0065 0.0029 0.0044 0.0060 0.0045
  - Model 5: 0.0079 0.0038 0.0058 0.0078 0.0058
  - Model 6: 0.0081 0.0036 0.0055 0.0073 0.0055
  - Model 7: 0.0067 0.0030 0.0045 0.0062 0.0046
  - Average: 0.0072 0.0034 0.0051 0.0069 0.0051

- Notes associated with RMSEs:
  - RMSEs reported correspond to static (1-step) and h-step ahead forecasts.
  - Numbers in bold in the original table indicate the model with the lowest RMSE for each horizon (Model 4 displays the lowest RMSEs for several horizons).

### Uncertainty analysis and fan charts
- Uncertainty around forecasts examined via fan charts:
  - One source of uncertainty: parameter estimation uncertainty and random shocks to projections (left chart of Figure 10).
  - Another source: sensitivity to oil price changes (right chart of Figure 10), using historical deviations in oil price changes to measure uncertainty around WEO oil price projections and their impact on Ecuador’s growth forecasts.
- Findings on uncertainty:
  - The uncertainty stemming from oil prices is lower than the one from parameter and random error uncertainties.
  - Large drops in oil prices would still pose a significant downside risk to growth.
- Technical details:
  - Fan charts show 90 percent confidence intervals.
  - Parameter and random-error uncertainty intervals were computed with 1,000 repetitions.
  - Dynamic forecasts are conditional on WEO projections of oil price and advanced economies growth.

### Summary and practical implications
- Best nowcast approach for Ecuador: ensemble (average) nowcast of models one through four, with inclusion of models five through eight when they perform better in recent periods (2015–16).
- Models excluding fiscal variables (5b–8b) are necessary when fiscal data are delayed or unreliable.
- Model 4 in the VAR framework provides the best GDP forecast performance across horizons examined and can signal 1–2 year ahead turning points in economic activity.
- Fan charts and conditional dynamic forecasts provide quantitative measures of forecast uncertainty and stresses, including an explicit assessment of oil-price-related downside risk.

*International Monetary Fund staff analysis as presented in the source content.*

### Appendix I. Forecasting and Nowcasting GDP Growth

### Appendix I. Forecasting and Nowcasting GDP Growth

### Summary statistics: growth rates (q-o-q)
- Table A1 reports Obs, Mean, Std. Dev., Min, Max for multiple series (growth rates calculated as log differences).
- Selected series (Obs / Mean / Std. Dev. / Min / Max):
  - GDP: 660 / 0.97 / 1.1 / -1.9 / 3.3
  - Private Final Consumption: 660 / 0.97 / 1.4 / -3.0 / 5.8
  - General Government Final Consumption: 661 / 1.15 / 2.4 / -7.4 / 10.3
  - Gross Fixed Capital Formation: 661 / 1.69 / 4.3 / -9.6 / 14.0
  - Change in Inventories: 660 / 0.58 / 96.2 / -474.9 / 523.3
  - Exports: 660 / 0.83 / 2.7 / -4.4 / 7.9
  - Imports: 661 / 1.55 / 5.5 / -10.6 / 16.0
  - GDP Oil: 660 / 0.63 / 4.4 / -7.8 / 16.6
  - GDP Non-Oil: 661 / 1.03 / 0.9 / -2.0 / 3.0
  - Index of Economic Activity: 381 / 1.05 / 4.2 / -7.8 / 11.8
  - Consumer Price Index: 661 / 1.90 / 3.2 / -0.1 / 21.2
  - Oil price (WTI spot price): 660 / 0.67 / 16.5 / -70.1 / 32.3
  - Terms of Trade Index: 660 / 0.45 / 9.9 / -47.3 / 22.6
  - Merchandise Exports: 661 / 1.84 / 11.1 / -46.7 / 21.2
  - Total Petroleum Exports: 501 / 1.18 / 22.9 / -80.4 / 50.5
  - Total Non-Petroleum Exports: 502 / 2.3 / 15.5 / -10.2 / 16.6
  - Oil Exports Revenue: 660 / 0.98 / 31.8 / -113.9 / 74.9
  - Value Added Tax (SRI): 663 / 3.1 / 6.6 / -15.7 / 22.3
  - Income Tax (SRI): 665 / 3.2 / 48.2 / -75.3 / 122.0
  - Capital Expenditure: 664 / 4.8 / 37.4 / -97.8 / 87.9
  - Gross Fix Capital Formation (public): 665 / 7.7 / 55.0 / -130.6 / 102.3
  - Credit to the private sector: 423 / 0.72 / 2.8 / -3.5 / 7.6
  - Deposits of the private sector: 423 / 3.2 / 2.7 / -3.1 / 7.8
- Sources: Haver; SRI; and IMF staff estimations.

### Correlations of cyclical real GDP growth (q-o-q)
- Table A2 reports correlations (5% significance level) across leads and lags (t-4 to t+4) for many series with Real GDP growth.
- Selected correlation coefficients reported:
  - GDP: 0.30 (t-4), 1.00 (t-3), 0.30 (t-2)
  - Private Final Consumption: 0.51 (t-4), 0.25 (t-3), -0.26 (t-2)
  - Gross Fixed Capital Formation: 0.46 (t-4), 0.36 (t-3)
  - Exports: 0.27 (t-4), 0.36 (t-3)
  - Imports: 0.30 (t-4)
  - Petroleum (incl. refining) and Mining: 0.55 (t-4), -0.28 (t-3)
  - Manufacturing (excl. petroleum refining): 0.40 (t-4), 0.51 (t-2)
  - Electricity & Water: -0.40 (t-4)
  - Construction: -0.35 (t-4), 0.43 (t-2), 0.51 (t-1), 0.26 (t+1)
  - Trade: 0.56 (t-4)
  - Post and Communications: 0.39 (t-4), 0.35 (t-2)
  - Net Taxes & Other Elements of GDP: 0.62 (t-4)
  - GDP Oil: 0.53 (t-4), -0.28 (t-3)
  - GDP Non-Oil: 0.29 (t-4), 0.77 (t-1), 0.36 (t+1)
  - Credit to the private sector: 0.59 (t-4), 0.65 (t-3)
  - Deposits of the private sector: 0.45 (t-4), 0.60 (t-3), 0.52 (t-2)
- Sources: Haver; SRI; and IMF staff estimations.

### Forecasting performance: RMSE of selected models (1-step ahead nowcast)
- Table A3 reports RMSEs for a variety of models across years 2009–2016 and for different forecast horizons (1-step ahead).
- Average model RMSE series (as reported): 0.0080, 0.0080, 0.0107, 0.0077, 0.0064, 0.0073, 0.0055, 0.0087, 0.0091, 0.0056, 0.0094, 0.0062, 0.0102, 0.0087, 0.0077
- Models identified:
  - Models with lowest RMSE overall: labelled Mo Mk Inc_sri; Mn Rev_o Gov_cap; Mo Mk Gov_gfcf Inc_sri; Mn Rev_o Gov_cap VAT_sri (specific yearly RMSEs listed in table).
  - Models with lowest RMSE in 2015-2016 include Xn Mk Gov_gfcf TOT Cre (5a), Xn Mk Rev_o Gov_gfcf TOT Cre (6a), Xn Mo Mk Rev_o Gov_cap TOT Inc_sri Cre (7a), Xn Mo Mk Gov_gfcf TOT Cre (8a) — marked "Yes" under certain columns in table.
  - Models without fiscal variables and their RMSEs (examples): Xo Mo Mk WTI Cre: 0.0101; Xo Mk WTI Cre: 0.0100; Xo Mk TOT VAT_sri Cre: 0.0102; Xo Mo Mk WTI Cre Dep: 0.0099.
  - Model with just the oil price (WTI) reported RMSEs across years: 0.0102, 0.0157, 0.0101, 0.0082, 0.0047, 0.0080, 0.0059, 0.0141, 0.0096, 0.0105, 0.0125, 0.0061, 0.0105, 0.0101, 0.0101
- Note: Numbers in bold for models 1-4 indicate the model with the lowest RMSE in each period (table formatting in source).

### Variable list and codes
- Table A4 maps variables to short codes used in models:
  - GDP = GDP
  - Private Final Consumption = Cpri
  - General Government Final Consumption = Cpub
  - Gross Fixed Capital Formation = GFCF
  - Change in Inventories = Inv
  - Exports = Xr
  - Imports = Mr
  - Agriculture and Fishing = Agri
  - Petroleum (incl. refining) and Mining = Oil
  - Manufacturing (excl. petroleum refining) = Man
  - Electricity & Water = Elec
  - Construction = Const
  - Trade = Trade
  - Accomodations & Restaurants = Tour
  - Transportation = Trans
  - Post and Communications = Comm
  - Financial Services Activities = Fin
  - Administrative, Tech & Prof Activities = Admin
  - Education and Social & Health Services = Edu
  - Public Administration, Defense = PubAd
  - Domestic Services = Dom
  - All Other Service Activities = Other
  - Net Taxes & Other Elements of GDP = NetT
  - GDP Oil = GDP_oil
  - GDP Non-Oil = GDP_noil
  - Index of Economic Activity = IEA
  - Consumer Price Index = CPI
  - Oil price (WTI spot price) = WTI
  - Terms of Trade Index = TOT
  - Merchandise Exports = Xn
  - Total Petroleum Exports = Xo
  - Total Non-Petroleum Exports = Xno
  - Merchandise Imports = Mn
  - Imports of Fuel and Lubricants = Mo
  - Imports of Capital Goods = Mk
  - Oil Exports Revenue = Rev_o
  - Value Added Tax (SRI) = VAT
  - Income Tax (SRI) = Inc
  - Capital Expenditure = Gov_cap
  - Gross Fix Capital Formation (public) = Gov_gfcf
  - Credit to the private sector = Cre
  - Deposits of the private sector = Dep

### VAR models and lag selection
- Tables A5 and A6 present VAR model specifications and order selection tests.
- Models listed include combinations with WTI and AEgdp treated as exogenous and various permutations of AEgdp, Gov_cap, Dep, Cred, GDP with different lag orders.
- Information criteria and tests reported: LLR, FPE, df, p-value, AIC, HQIC, SBIC. Example reported values:
  - Model (one reported line): LLR = 1535.557; FPE = 505.240; df = 160.000; p-value = 4.80E-12; AIC = -26.4111; HQIC = -26.2056; SBIC = -25.8931
  - Another line: LLR = 2554.181; FPE = 37.246; df = 160.002; p-value = 4.50E-12; AIC = -26.4872; HQIC = -26.0761; SBIC = -25.4511
- Note: Lag order selection treated WTI oil price and AE GDP as exogenous, using Lütkepohl version of information criteria. Numbers in bold in source indicate selected lag order.

### EMBI spreads: external factors, and the impact of fiscal consolidation
- Context and stylized facts:
  - Ecuador's borrowing costs declined substantially since the 2015–16 peak but have been on an increasing trend and remain elevated.
  - Since end-2015, Ecuador EMBI spreads declined by 819 bps, reaching a new low following the oil price collapse of 447 bps in December 2017.
  - During the first half of 2018 the sovereign spread trended upward back to the neighborhood of 700 basis points before a subsequent decline after announcement of program negotiations with the IMF.
- Drivers and recent dynamics:
  - Oil prices and global risk appetite are important drivers of Ecuador’s EMBI spreads.
  - A twelve-month rolling correlation between year-on-year change in EMBI spread and percentage change in oil prices (daily data) is largely negative in most periods; it declined notably in 2011, 2013, 2015, and early 2018 coinciding with episodes of increased global risk aversion.
  - The correlation pattern suggests that analysis must consider external and domestic factors alongside oil prices.
- Empirical approach:
  - Method: Local Projections (LP) per Jorda (2005) applied to a panel of 60 emerging market economies over 1990–2017 (unbalanced).
  - Dependent variable: EMBI sovereign spread.
  - Key domestic explanatory variables: cyclically adjusted primary balance (lagged), gross public debt to GDP ratio (lagged).
  - Additional domestic variables: institutional quality (lagged), default history (lagged).
  - External variables: oil price (WTI), investors’ risk appetite proxied by VIX, and an oil-exporter dummy interacting with oil price.
  - Default history proxied by a 500-bps increase in average annual EMBI spread; institutional quality calculated as average of bureaucracy quality and law and order from ICRG.
  - IRFs constructed by plotting coefficient β_h from regressions with h ranging 0 to 3; results for h = 0 reported in Table 1 of source.
- Main empirical findings and interpretation:
  - Historically, oil prices were important drivers of Ecuador’s EMBI spreads, but reduction in global risk appetite has become an important factor more recently.
  - Decompositions using estimated coefficients explain changes in EMBI spreads in 2015, 2016, and 2018 relatively well; large unexplained residuals exist for changes in 2014 and 2017, implying additional unmodeled factors.
  - Possible drivers for unexplained moves:
    - 2014: appreciation of the U.S. dollar and deterioration in external competitiveness.
    - 2017: political changes (new administration) likely contributed to reduced borrowing costs.
  - In 2018: negative contribution from recovering oil prices and remnants of fiscal balance improvement in 2017 were offset by increased global risk aversion and other unmodeled factors, yielding only a minor decrease in EMBI spread.
  - Conclusion: Unfavorable external conditions are likely to persist given risks from U.S. monetary policy normalization and geopolitical tensions; Ecuador remains vulnerable to reductions in global risk appetite and consequent increase in borrowing costs.
- Policy implication emphasized:
  - Domestic fiscal policy can have a notable impact on borrowing costs via reduction in fiscal deficit and public debt stock over time.
  - Building fiscal buffers can help reduce the EMBI spread for Ecuador in the face of unfavorable global developments.

*Prepared by IMF staff; sources within chapter: Haver; SRI; IMF staff estimations.*

### 7.      Building fiscal buffers can help reduce EMBI spreads in Ecuador and mitigate the

### 1ecuea2019002 - 7.      Building fiscal buffers can help reduce EMBI spreads in Ecuador and mitigate the

### Impact of fiscal consolidation on EMBI spreads
- An improvement in the cyclically adjusted primary balance (CAPB) of 1 percent of GDP is associated with a decrease in the EMBI spreads on impact of about 17 to 34bps.
- A reduction of 1 percent of the debt to GDP ratio would help reduce the spread on impact by 4 to 7bps.
- The CAPB effect is stronger for oil exporters than for other countries; the lower bound estimate refers to all EMs in the sample, the upper bound corresponds to oil exporting EMs (the latter is used for Ecuador calculations).
- A country that had at most one period of severe debt distress during the sample period has on average 140 to 230 higher EMBI spread, with the lower estimate for oil exporters.
- Policy implication: a credible and prudent fiscal framework can minimize the probability of negative credit events and reduce vulnerability to adverse external conditions.

### Quantified scenario for Ecuador
- Adoption of a package of fiscal measures of 5 percent of GDP (discretionary fiscal measures of 5 percentage points of GDP) together with a corresponding reduction in public debt-to-GDP ratio by 1 percentage points of GDP:
  - The EMBI spread for an oil-producing economy of Ecuador could be permanently lowered by close to 175 basis points.
- Additional, slower-moving improvements in the strength of institutions could further reduce the sovereign spread.

### Reserve adequacy in a dollarized economy (motivation and risks)
- Dollarization specifics:
  - Reserves correspond only to dollars that represent claims on foreign sources; dollar claims on domestic agents (including the government) are not part of reserves.
  - Any dollar in the economy could be used for international trade or cross-border financial transactions; any dollar on the asset side of the BCE balance sheet could support banking liquidity or finance government activities if liquid.
  - Dollarized economies cannot print money to exchange for foreign currency, constraining reserve accumulation.
- Three broad reasons buffers are needed in dollarized economies:
  - (i) Balance of Payments shock-absorption,
  - (ii) lender of last resort reasons,
  - (iii) prudential coverage (coverage of private sector claims, e.g., deposit runs translate into pressure on foreign reserves).
- Specific prudential considerations for Ecuador:
  - Even expected higher demand for cash during end-of-year holidays negatively impacts liquid reserves at the BCE.
  - A prudential reserve coverage of banks’ deposits implies a liquid reserve floor; voluntary bank deposits are more volatile than mandatory tranches but, in crisis, all deposits should be accessible and need coverage.
  - Deposit-receiving public banks should be included in a precautionary reserve floor.
  - A prudent reserve level should encompass buffers to mitigate fiscal shocks (e.g., a measure of volatility of government credit needs or one month of government spending could be appropriate benchmarks).

### Liquidity Fund (LF) and BCE balance sheet details
- Liquidity Fund (LF):
  - Established in 2009; accumulated resources from the banking system.
  - Balance of the LF stood at US$2.6 billion at end-2018 (about 2.5 percent of GDP).
  - US$2.6 billion represents 9.3 percent of deposits while the target is 10 percent.
  - LF is fully invested abroad (mainly assets issued by BIS and FLAR).
  - LF rule: 30 percent of each bank’s contribution treated as part of a pooled fund to support small entities; remaining 70 percent reserved for the contributing bank’s potential liquidity needs.
  - The fund has not been tapped in any significant way since creation and could be insufficient in a scenario of severe system-wide deposit runs.
- BCE balance sheet and trends:
  - Until September 2014 international reserve buildup peaked at US$6.7 billion.
  - Reserves fell to 2 billion at end-December 2017, equal to 52 percent of banks deposits at the BCE.
  - After 2014 fiscal deterioration and drying external financing turned the central government into a net borrower from the central bank.
  - Composition shifted toward domestic assets (e.g., lending to the central government), lowering coverage of banks’ deposits with reserves.
  - Banks have become the main funding source of the central bank; TBCs issuance has been limited (less than US$200 million).

### Banking sector liquidity and peer comparison
- Ecuadorian banks’ liquidity ratio (liquid assets to short term liabilities):
  - Initially compared well with dollarized peers, but beginning in the crisis years the average liquid assets ratio gradually fell below other dollarized economies’ levels.
  - More recently liquidity was severely affected by the sudden drop in oil prices and concurrent fiscal deterioration, but has recovered somewhat.
  - Comparing across all exchange rate regimes, Ecuador’s liquidity ratio was near the lower bound not only for dollarized economies but for all.
- Peer central banks:
  - Central banks in similar constrained economies tend to hold better reserve coverage of banks’ deposits; example: El Salvador and Kosovo maintain coverage ratios above 100 percent.

### Reserve adequacy metrics (ARA-EM and findings)
- The Fund’s operational methodology for reserve adequacy (ARA) developed in 2011; ARA-EM incorporates relative risk measures from potential BoP pressure and estimates reserve cover needed relative to risk-weighted measure.
- ARA-EM contemplates: i) losses in export earnings, ii) rollover risk of short-term debt, iii) portfolio outflows, iv) changes in broad money.
- A reserve coverage of 100−150 percent of the ARA-EM metric is regarded as adequate.
- Ecuador’s reserve level has been significantly below most of the ARA-EM metrics and traditional benchmarks; currently the reserve level in Ecuador is about one-fifth of the ARA metric.
- Risk Weights in the ARA-EM Metric (Exchange Rate Regimes — Weights in percent):
  - Fixed: Short-term Debt 30, Other Liabilities 20, Broad Money 10, Exports 10
  - Floating: Short-term Debt 30, Other Liabilities 15, Broad Money 5, Exports 5

*Source: IMF staff chapter: “Building fiscal buffers can help reduce EMBI spreads in Ecuador and mitigate the impact of unfavorable global external conditions,” and “Reserve Adequacy” (excerpts provided).*

### 11.      The Fund has evaluated the relevance of the ARA metric for fully and partially

### 11.      The Fund has evaluated the relevance of the ARA metric for fully and partially dollarized economies

### Relevance of the ARA metric
- IMF, 2013 and IMF, 2015 found that, in assessing the performance of the ARA metric for dollarized economies versus other emerging economies, dollarized economies do not seem to be statistically more vulnerable than others to balance of payments crises.
- IMF (2016) states that “the ARA EM metric may provide a conservative starting point as an adequate liquidity buffer to support domestic financial institutions in a fully dollarized economy”.
- Conclusion from the Fund: the standard ARA metric would in principle apply for dollarized economies, but specific circumstances of each case should be taken into account when discussing reserve adequacy.

### Ecuador: limitations in applying the ARA metric
- Ecuador’s interpretation of the ARA metric is complicated by poor external sector statistics, particularly financial account data.
- The largest component of the ARA-metric for Ecuador is the 20 percent of “Other Liabilities”, designed to cover the risk of non-resident equity and medium and long-term capital outflows (IMF, 2016), and based on IIP data for portfolio and other investment.
- Ecuador’s IIP is calculated based on an initial stock plus the cumulative flow from the balance of payments; there are significant shortcomings in determining the initial stock and in the methodology for compiling “Other Investment” in the balance of payments statistics (used mainly as a counter-entry for other BoP items).

### Supplemental absolute reserve-floor metric (staff’s proposed supplemental metric)
- Rationale: supplement ARA with an absolute reserve floor to capture minimum liquidity buffers needed against potential reserve drains from the banking and fiscal sectors.
- Under this criterion, reserves should be sufficient at least to cover:
  - (i) all deposits of the banking system at the central bank;
  - (ii) money issuance (e.g. coins) and issuance of TBCs;
  - (iii) electronic money;
  - (iv) a measure of volatility of public credit; and
  - (v) contingent liabilities.
- Notes:
  - This metric does not include a buffer to supplement the ability of the liquidity fund to supply emergency liquidity assistance to banks.
  - Any encumbrance on liquid reserves should also be subtracted to obtain a better measure of NIR.
- Historical coverage:
  - Up until 2015, those potential claims were comfortably covered by NIR, including a fiscal buffer based on the volatility of credit to the public sector.
  - After 2015, as public financing needs ballooned and external financing to the government became more scarce, the margin of coverage for these concepts faded while NDA turned positive, crowding out NFA in the BCE’s balance sheet.
  - At end-2018, liquid reserves amounted to less than 50 percent of staff’s supplemental metric, although reserves recovered since then on the basis of sovereign debt issuance.
- Authorities’ operational concept:
  - Authorities adopted an operational concept akin to the proposed floor, but their metric does not cover required reserves of private banks (only excess deposits), or any amount for fiscal buffers.

### Conclusion (Ecuador reserve position and policy implications)
- The difficult macroeconomic situation of the last five years in Ecuador significantly weakened its NIR position leaving buffers well below prudent levels.
- Considering a set of standard and non-standard criteria, the current stock of reserves is insufficient to mitigate potential shocks to the economy.
- Policy objectives and actions should incorporate the need to rebuild reserves and consider creating fiscal mechanisms to protect NIR management and the BCE balance sheet from the fiscal cycle.
- Priority actions to support confidence in the banking system and the dollarized regime:
  - Maintain adequate reserve cover of private sector claims on the BCE.
  - Reestablish central bank independence and reinstate the four balance sheet accounts at the BCE or implement an equivalent tool imposing a transparent rule to manage reserves.

### Interpreting apparent deposit outflows — balance of payments (BOP) perspective
- BOP observation: Ecuador’s financial account appears to show a significant net accumulation of currency and deposits abroad by residents in recent years; this trend in Other Investment, Net Acquisition of Assets, Currency and Deposits has been particularly pronounced since 2014.
- Magnitudes recorded: according to Ecuador’s BOP statistics, between 2008 and 2017, residents accumulated US$19 billion in currency and deposits abroad.
- Concurrent indicators:
  - International reserves have fallen well below the standard metrics and exhibit high volatility.
  - Deposits in the domestic banking system increased from 15 to 34 percent of GDP between 2008 and 2017.
- BOP mechanics and Ecuador’s specifics:
  - In Ecuador, the currency and deposits category is used to register the contra-entry of transactions which do not have a counterpart elsewhere in the BOP; it is a residual category rather than being based on observed flows in the financial system.
  - Errors and omissions therefore appear artificially low, averaging 0.26 percent of GDP per year in absolute terms between 2008 and 2017.
  - Consequence: the large magnitude of apparent accumulation of currency and deposits warrants further investigation into measurement and recording practices.

### Sectoral decomposition of currency and deposits (findings)
- Sectoral breakdown shows the accumulation emanates from the “central bank” sector rather than the private sector.
- Patterns:
  - Central bank category shows large acquisition of currency and deposits abroad.
  - General government is mostly zero.
  - Other sectors have tended to draw down currency and deposits abroad, with the exception of 2016.
- Correlations:
  - Net acquisition of currency and deposits by the central bank mirrors closely the oil trade balance (a proxy for the public sector trade balance).
  - Private sector currency and deposits track the non-oil trade balance.
- Implication: these patterns are consistent with the use of the currency and deposits category as a contra-entry for other BOP transactions and cast doubt on the assertion that private outflows are the primary driver of the recorded accumulation.
- Comparative observation: these sectoral patterns are unusual relative to other dollarized economies (e.g., El Salvador and Panama), where major movements typically occur between deposit-taking corporations and/or other sectors, with little movement under “central bank”.

### Monetary-side analysis of currency and deposits
- Alternative data sources: monetary statistics, supervisory data, and BIS locational banking statistics can be used to estimate net accumulation of currency and deposits abroad.
- Results from monetary-side estimates (2008–2017):
  - Private sector deposits held abroad: accumulation of about US$1.7 billion.
  - Banks’ deposits abroad: net decline of US$0.6 billion.
  - Banks’ vault holdings: added about US$1.1 billion.
  - Increase in U.S. dollars in circulation (cash in circulation): close to US$12 billion (BCE estimate: US$11.6 billion increase over 2008–2017).
- Interpretation:
  - Currency flows (increase in cash in circulation) are much more significant than deposit flows in explaining the BOP’s large currency and deposits figures.
  - The monetary-side evidence suggests the private-sector deposit accumulation recorded in the BOP (US$19 billion) is substantially overstated relative to what monetary and BIS data indicate.

### Key statistics (preserved exactly as in source)
- “20 percent” — component of the ARA metric labeled “Other Liabilities”.
- US$19 billion — residents accumulated in currency and deposits abroad between 2008 and 2017 (BOP statistics).
- US$11.6 billion — BCE estimate of increase in cash in circulation over 2008–2017.
- Deposits in the domestic banking system increased from 15 to 34 percent of GDP between 2008 and 2017.
- Errors and omissions averaged 0.26 percent of GDP per year in absolute terms between 2008 and 2017.
- At end-2018, liquid reserves amounted to less than 50 percent of staff’s supplemental metric.
- Private sector deposits held abroad (monetary-side estimate): about US$1.7 billion (2008–2017).
- Banks’ net decline in deposits abroad: US$0.6 billion (2008–2017).
- Banks’ vault additions: about US$1.1 billion (2008–2017).

*Source: https://www.imf.org/-/media/files/publications/cr/2019/1ecuea2019002.pdf (Content unit: 1ecuea2019002).*

### 10.      However, these estimates of net acquisition of currency and deposit from the

### 10.      However, these estimates of net acquisition of currency and deposit from the

### Data gaps and measurement differences between monetary side and BOP
- Monetary-side estimates of net acquisition of currency and deposit differ substantially from BOP estimates in many years.
- Two potential sources of data gaps in monetary estimations:
  - Incomplete coverage of BIS data:
    - BIS locational banking statistics cover banks in most major financial centers, including the United States.
    - Coverage in Latin America is limited to Brazil, Chile, Mexico and Panama.
    - Important trade and investment partners of Ecuador not covered by BIS reporting include China, Vietnam, Colombia and Peru.
    - Surveys of private companies can help remedy these coverage gaps.
  - Under- or over-estimation of U.S. dollars in circulation:
    - BCE uses the flow of physical banknotes between the central bank and the rest of the financial system as the basis for its estimations of currency in circulation (Vera, 2007).
    - Some currency flows may not pass through the financial system (remittances, tourism, illicit activities) and are therefore more difficult to track.
    - Asociación de Bancos del Ecuador (2016) posits that the BCE’s estimate of notes and coin in circulation is inflated because it assumes cash dollars leaving the financial system will eventually return.
    - Currency in circulation increased from 6.4 percent of GDP at end-2007 to 14.2 percent of GDP at the end-2017, with much of the increase occurring since 2014.
    - The increase corresponds to the oil price shock and consequent liquidity crunch, falling deposits, and increased preference for cash.
    - It is unclear how much of the additional cash stayed in Ecuador or was spent abroad; evidence suggests much did not return to the financial system.
    - Incentives to use cash abroad included an overvalued exchange rate and avoidance of the five percent ISD tax (levied on credit card transactions).

### BOP-side distortions and trade reporting problems
- On the BOP side, accumulation of currency and deposits could be distorted by inaccurate trade data.
- Currency and deposits category in Ecuador’s BOP is used to register counter-entries of other BOP transactions; inaccurate reporting of other components causes distortions.
- Trade (exports and imports) is the largest BOP component and may be over- or under-estimated.
  - Press reports suggest smuggling (underestimation of imports) may be a problem, particularly from 2014 through mid-2017 due to an overvalued real exchange rate and punitive “safeguard” tariffs.
  - Customs administration has highlighted under-invoicing to reduce tax burden of imports.

### Tax data and the ISD as an additional information source
- ISD is levied on all outward foreign transactions, including payments for imports of goods and services; some exemptions and exporter rebates apply.
- The ISD’s tax base of gross outflows can be approximated using the tax rate and total ISD collections.
- Figure 3 (in source) shows the tax base of gross outflows corresponds closely to imports.
- If large legal outflows were occurring, the inferred tax base would be much larger than imports.

### Conclusions on accumulation of currency and deposits and capital flight
- Evidence suggests the accumulation of currency and deposits over the period does not appear to indicate private capital flight at material levels.
- Sectoral disaggregation suggests the public sector, not the private sector, has been apparently accumulating currency and deposits abroad, as the oil trade balance is typically in surplus and external loan disbursements have exceeded debt service payments.
- Important caveats and measurement difficulties:
  - Magnitude of deposit accumulation abroad by Ecuadorian residents is much lower than suggested by BOP figures, but BIS data have only partial coverage.
  - Acquisition of foreign currency likely an important component, but measuring currency in circulation in a dollarized economy is difficult and current measures may be over-estimates.
  - Illicit activities such as smuggling and under-invoicing also contribute to BOP measurement difficulties.
- It is unclear to what extent the “missing dollars” can be explained by accumulation by corporations, hoarding by households, or spending on imports rather than saving.
- A revision of authorities’ methodology in line with recommendations of an IMF technical assistance mission on BOP statistics is appropriate.
  - Such a revision would likely result in higher net errors and omissions but would provide a clearer picture and avoid the misleading appearance of large deposit outflows.

*Sources: Bank for International Settlements (BIS), IMF Balance of Payments Statistics, Superintendency of Banks, BCE, and IMF staff calculations.*

### The cooperative sector in Ecuador — key findings
- Importance and size:
  - Cooperatives represent around 16 percent of assets and 23 percent of deposits in the consolidated financial system.
  - Total assets of the cooperative sector constitute around 12 percent of GDP, slightly more than the assets of Ecuador’s largest private bank, Banco Pichincha.
  - There are currently 618 COACs, down from a peak of almost 1000 in 2012.
  - Number of members (depositors/shareholders) has more than doubled since 2013 and stands at about 7.7 million.
- Structure and regulation:
  - Lending concentrated in consumer and microfinance sectors.
  - Coops are divided into 5 “segments” by asset size; strictest requirements apply to the largest coops (Segment 1).
  - Regulations resemble those of banks but with less strict rules, lower requirements, or longer compliance periods.
  - Cooperatives are regulated by the Superintendent of the Popular and Solidarity Economy (SEPS), a different entity from private banks’ regulator.
  - Continued consolidation and alignment of supervision with banking sector best practices recommended to avoid regulatory arbitrage.
- Reported soundness indicators:
  - Segment 1 has US$9.4 billion in assets, representing 69 percent of assets in the coop system.
  - Segment 1 had an average level of NPLs of 3.6 percent in September 2018, close to the average for private banks.
  - Caveat: loans in cooperatives are typically classified as non-performing after 30 days rather than 15 days for private banks, so NPLs are not strictly comparable.
  - Segment 1 experienced rapid growth to September 2018, with deposit and credit growth outpacing private banking system growth (partly due to consolidation).
- Interconnectedness and safety nets:
  - Interconnectedness mainly linked to common safety net funds and access to liquidity financing from the BCE and from a tier 2 bank (Conafips).
  - Some cooperatives structurally linked to the banking system through their operations; no recent network analysis quantifies linkages.
  - Funds from the private bank deposit insurance scheme may be lent to coops to facilitate consolidation, at 20 years maturity with a two-year grace period.
- Deposit insurance and liquidity funds:
  - Deposit insurance for COACs currently has about US$331 million available, which, at five percent of insured deposits in the sector, is still insufficient.
  - Between April 2014 and December 2018, the deposit insurance paid out US$47.5 million due to 166 liquidations, benefiting 443,114 members.
  - Liquidity fund for cooperatives:
    - Creation mandated by the Organic Financial and Monetary Code of 2014; separate liquidity funds for private banks and cooperatives.
    - Fund for cooperatives established in 2016 with initial capital of US$40 million.
    - As of end-September 2018, it had US$155 million available, covering about 1.5 percent of deposits in the coop system.
    - Central bank may provide liquidity credits to cooperatives in Segment 1.
- Economic role:
  - Coops operate mostly in rural areas, concentrate on personal credit and SMEs.
  - Healthy consolidation and alignment with best supervisory practices would benefit rural employment, financial deepening, and financial inclusion.

### Electronic money in Ecuador — summary of features and policy-relevant points
- Ecuador was the first country to introduce an electronic money system managed by its central bank; adoption failed to take off.
- Distinction between electronic money (EM) and virtual currencies (VCs):
  - VCs are digital representations of value issued by private developers and denominated in their own unit of account.
  - EM is a digital payment mechanism denominated in fiat currency and generally originates within the banking system; regulated by the central bank and government agencies.
- Advantages of EM include:
  - Convenience (anytime, anywhere transfers; solves “exact change” problem)
  - Time savings (short transfer times)
  - Money savings (lower fees and commissions than regular wire transfers)
  - Transparency (real-time expense checks)
- Disadvantages include:
  - Safety risks (physical electronic wallets can be lost, stolen or damaged; risk of hacking)
  - Privacy concerns (transactions cannot be anonymous; information stored in ledger)
  - Need for internet access
- Macroeconomic implications:
  - EM may increase money demand and monetary policy effectiveness by increasing the opportunity cost of holding physical cash.
  - EM can facilitate conduct of monetary policy and help avoid lower-bound limitations on nominal interest rates typical of physical cash.
  - If EM became the only currency in circulation (illustrative), costs associated with paper money and seigniorage revenue would be zero.
- Financial inclusion potential:
  - EM can extend access to finance in remote areas where bank branches are scarce but mobile phone ownership is common.
  - EM services often start with money transfers; expanding to savings, credit, and insurance could provide additional benefits.

### 5. In December 2014, Ecuador introduced its public EM system.

### 5. In December 2014, Ecuador introduced its public EM system.

### Legal definition and technical features
- EM was defined as an electronic payment means managed and regulated by the central bank, denominated in U.S. dollars.
- Opening EM accounts was voluntary, free of charge, and bore zero maintenance cost.
- Account holders could move funds between bank accounts and electronic wallets through electronic devices, including mobiles, smart cards, and computers.
- Services available included: sending and receiving money orders (including to/from people or entities without EM accounts), making and receiving transfers to and from other bank accounts, paying taxes, and sending payments to other EM accounts.
- Some services required the payment of fees and commissions set by the law.
- Recent EM systems use biometric identification (e.g., thumbprints and retina images), providing an additional layer of security.
- Resolution N. 005-2014-M and Resolution N. 109-2015-M are cited as regulatory instruments.
- The law specified that every day the amount of EM in the system was to be registered in the liabilities of the central bank’s balance sheet, and that the amount had to be backed by U.S. dollar assets.
- At any point, holders of EM could request physical money in exchange for EM from macro agents—entities supervised by the central bank that (among other criteria) had been operating in Ecuador for a minimum of two years and had a minimum capital of US$100,000 and provided customer service.

### Stock, diffusion, and fiscal incentives
- Peak stock of EM in the financial system: US$11.8 million in approximately 400,000 registered accounts.
- Government tax incentives to boost EM diffusion included:
  - a VAT refund of two percent of purchases using EM;
  - a rule that revenue, costs, and expenses incurred with EM could be excluded from the calculation of the tax base for the income tax during 2017–19.
- The government announced that users could pay taxi services with EM.
- Slow diffusion was likely due to the private sector’s perception that EM was a first step to de-dollarize the system and general concerns about privacy.

### Policy shift and transition
- The Moreno administration decided to shift the EM system to the private sector, with the stated purpose of increasing use of e-money to reduce the use of dollar bills and promote financial inclusion.
- The Economic Reactivation Law, passed in December 2017, decreed that the central bank EM system would be phased out; by mid-April the balance on the EM accounts was down to zero.
- Private banks opted to adopt a different platform to facilitate mobile payments; as of February 2019 the system was still awaiting regulatory approval.

### Potential benefits and outlook
- A wider diffusion of EM could:
  - reduce transaction costs,
  - enhance transparency,
  - spur financial inclusion,
  - increase financial services availability.

*Source: 1ecuea2019002 - 5. In December 2014, Ecuador introduced its public EM system.*

### 9. The debt ceiling and the size of the safety margin will be sensitive to a number of key

### 9. The debt ceiling and the size of the safety margin will be sensitive to a number of key parameters

### Sensitivity of the debt ceiling and safety margin
- The debt ceiling will be smaller and the safety margin will be larger when:
  - the MDL and risk tolerance level are smaller (the initial debt must be lower to reduce the probability of breaching the limit);
  - the volatility of macroeconomic shocks is higher (because larger shocks can generate larger increases in debt, a larger safety margin is required);
  - the response of the primary balance to changes in debt following negative shocks is weaker, as reflected in the parameters of the FRF (a larger safety margin is required when the government is not acting strongly enough to offset the impact on debt of negative shocks).

### Quantitative calibration and scenarios analyzed
- A sensitivity analysis was performed across:
  - MDL range estimate: (32, 47, and 57 percent of GDP);
  - risk tolerance levels: 5 to 15;
  - two fiscal reaction function (FRF) alternatives: “past fiscal behavior” and “market-led adjustment”.
- The FRF used for the “past fiscal behavior” scenario is:
  - primary balance (t) = .335* primary balance (t-1) + 0.324 terms of trade gap (t) - .313*external disbursements + 2.194
- Ceilings are lower under the “past fiscal behavior” FRF than under the “market-led adjustment” FRF because:
  - the historical FRF for commodity-exporting EMs is quite insensitive to debt and quite procyclical with respect to commodity prices;
  - the “market-led adjustment” FRF implies stronger adjustment of fiscal policy to debt increases, particularly when oil prices decline and market financing is more costly.

### Key findings and policy-relevant thresholds
- A prudent debt ceiling for Ecuador would be at most 30 percent of GDP.
- Ecuador’s current debt ceiling of 40 percent of GDP could be deemed prudent only:
  - under risk tolerance levels and debt distress probabilities above 10 percent; and
  - if fiscal policy allows for a greater adjustment of primary balances in response to debt increases than those observed in the past.
- Recent changes in Ecuador’s fiscal framework, together with proposed options to further enhance the framework, are expected to:
  - promote fiscal savings in good times; and
  - use those savings to help close market financing gaps in bad times, thereby supporting a shift toward the stronger fiscal behavior implicit in the “market-led adjustment” scenario.

### Supporting analysis notes
- The calibration exercise and illustrative scenario (Figure 2) use IMF staff calculations of Ecuador: Public Debt-to-GDP Ratio 6-year Forecast from Debt Ceiling Quintile of Public Debt Empirical Distribution (In percent of GDP).
- Footnote: See Analytical Note 1 for evidence of the strong correlation between oil prices and market financing costs proxied by EMBI spreads.

*Source: IMF staff analysis as presented in the Ecuador staff report chapter on debt ceiling calibration.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2019/1ecuea2019002.pdf_