## cr17139

## Source details

**Canonical URL:** [cr17139](https://www.imf.org/-/media/files/publications/cr/2017/cr17139.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2017/cr17139.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2017/cr17139.pdf.json)

---

### Interest Rates and Monetary Stance
- Monetary policy tightening cycle: policy rates raised by 325 bps to 7.75 percent since September 2015.
- December 2016 MPC action: policy rate cut by 25 bps citing:
  - transitory shocks dissipating;
  - real GDP growth during 2016Q3 lower than expected;
  - ex-ante real interest rates above their average since 2005.
- Neutral interest rate evidence:
  - Traditional estimate (Magud and Tsounta (2012) methods): real neutral interest rate = 1.7 percent.
  - DSGE (small open-economy, Bayesian priors from Gonzalez and Hamann (2011) and Hamann et al (2014)):
    - real neutral rate has declined since 2007;
    - DSGE estimate of current real neutral rate ≈ 0.7 percent;
    - neutral rate level depends on model specification and shock assumptions.
  - Central Bank range median = 1.4 percent (BanRep 2015).
- Monetary stance:
  - After recent cuts and with inflation expectations at 4 percent, ex-ante real policy rate ≈ 3 percent.
  - Staff assessment neutral range = 1−2 percent.
  - Conclusion: policy rate is above the neutral rate; stance is tight.

### Financial Conditions Index (FCI): Methodology and Findings
- Purpose: capture overall financial conditions via credit, leverage, and risk channels while purging direct effects of monetary policy and business cycle.
- Methodology:
  - Factor-augmented VAR (FAVAR) with time-varying parameters (TVP-FAVAR) per Koop and Korobilis (2014).
  - Additional specifications: PCA-FAVAR (Stock and Watson 2002; Bernanke et al 2005) and KF-FAVAR (Doz, Giannone, and Reichlin 2011).
  - Empirical model uses X_t (20 financial variables) and Y_t (inflation, real GDP growth, REER, policy rate); unobservable factor f_t interpreted as FCI.
- Composition: risk (spreads, asset volatilities), collateral/leverage (bank leverage, housing and stock prices), and credit (commercial, consumption, mortgage).
- Normalization and controls:
  - FCI normalized around 0 over 2001–16: >0 = “loose”, <0 = “tight.”
  - Purged of cyclical influence by controlling for GDP growth and for the policy rate.
- Key empirical findings:
  - Financial conditions deteriorated and remained tight during 2016.
  - Impulse response: a 1 standard deviation tightening of financial conditions lowers GDP growth by about 0.25 percentage points within 12 months.
  - Estimated FCI tightened by one standard deviation in 2015 and by one fifth of one standard deviation in 2016.
  - Cumulative tightening implies a drag on growth in 2017 subtracting around 0.2 ppts.
- Policy implication:
  - Central bank has room to continue easing given inflation trends and expectations, but timing of cuts should depend on consolidation of trends.
  - Monitor FCI closely because a 1 standard deviation tightening reduces GDP growth by about 0.25 percentage points within 12 months.

### Housing Market: Recent Developments
- House price evolution (2005 to mid-2016, main three cities Bogota, Cali, Medellin):
  - nominal rise ≈ 200 percent; real rise = 110 percent.
  - Average home-price index real annual growth = 5.18 percent (2005–2016).
  - Bucaramanga: highest average annual real growth = 8 percent.
  - Armenia: lowest average annual real growth = 3.4 percent.
- Ownership and deficit:
  - 2015 households = 14 million: 46 percent own homes, 38 percent rent, 16 percent free usufruct.
  - Total housing deficit = 25 percent of households in 2014 (down from 54 percent in 1993).
  - 2014 quantitative deficit = 9 percent; qualitative deficit = 15 percent of households.
  - 2014 total deficit: rural 48 percent, urban 18 percent.
  - Quantitative deficit in rural areas rose from 7 percent in 1993 to 18 percent in 2014.
- Construction and activity:
  - Construction share in GDP: bottomed around 4 percent in 2000; reached 9.7 percent in 2015.
  - Construction licenses index rose strongly except during 2008 crisis; construction licenses declined in 2016.

### Mortgage Credit and Government Housing Programs
- Mortgage credit:
  - Real mortgage credit growth average = 13 percent per year since 2011.
  - Mortgage credit ≈ 5 percent of GDP.
  - About 30 percent of mortgages originated in 2016 were subsidized.
- Government programs (features and targeting):
  - Mi Casa Ya: targets households with income between 2 and 4 minimum wages, houses up to US$30,000; down payment support and interest subsidy.
  - FRECH: interest-rate subsidies (4−5 percentage points) for households up to 8 minimum wages, houses up to US$50,000.
  - FRECH No-VIS: interest-rate subsidies (2.5 percentage points) for houses up to US$77,000; open to any household regardless of income.
  - Annual ceilings (cupos) determine number of subsidized houses; programs apply only to new houses.
  - Houses sold with mortgage subsidies rose from ≈ 21,000 in 2013 to ≈ 33,000 in 2016.
  - Official estimates: construction sector interconnected with 25 percent of the industrial sector.
  - Fiscal multipliers of housing program vary from 10 for FRECH No-VIS to 1 for the free housing program (official estimates).

### Housing Finance Features and Resilience
- Regulatory safeguards and market features:
  - LTV ceilings: 70 percent (80 percent for social interest housing).
  - Observed LTV in Colombia = 53 percent (below regulatory ceiling).
  - Mortgages usually contracted at fixed interest rates; both variable- and fixed-rate offered.
  - Full recourse on mortgages.
  - Average mortgage maturity in Colombia = 15 years; effective duration ≈ 6−7 years due to prepayments.
  - Funding: banks primarily use retail deposits.
  - No prepayment penalties; borrowers frequently make partial prepayments.
- Cross-country context:
  - Maximum observed LTV median across countries = 83 percent (range 70 percent to 125 percent).

### House Price Misalignment and Modeling
- Model-based misalignment:
  - Estimated misalignment: house prices ≈ 13 percent above equilibrium.
- Determinants and econometric approach:
  - Error-correction model with short-term changes in house prices linked to long-term affordability (price-to-income), per capita real income growth, mortgage credit growth, interest rates, and population growth.
  - Sample for estimation: quarterly data Q1 1994 – Q1 2016.
  - Price measure: existing home price index (BanRep), deflated by CPI; geographic coverage includes Bogota, Medellin, Cali, Barranquilla, Bucaramanga, Cucuta, Manizales, Neiva, Villavicencio and Soacha.
  - VAR for macro effects: quarterly Q1 1994 – Q2 2016; domestic variables: real GDP, real private consumption, real investment, 90-day certificates of deposit nominal rate, real existing house price index; lags = 3.
  - Identification: generalized impulse response approach.
- Econometric results (selected statistics and coefficients):
  - Regression fit: R-square = 0.45; Adjusted R-squared = 0.4147; Prob > F = 0.0000; F(6, 78) = 10.92; Number of obs = 85.
  - Speed of correction: misalignment corrected at about 13 percent every quarter (affordability coefficient).
  - Selected coefficient estimates:
    - lnafford1 (affordability lag): Coef. = -.132468; Std. Err. = .0315063; t = -4.20; P>|t| = 0.000; [95% Conf. Interval] = -.1951922 to -.0697437.
    - hpi1 (lagged hpi): Coef. = -.4009662; Std. Err. = .0926514; t = -4.33; P>|t| = 0.000; [95% Conf. Interval] = -.5854208 to -.2165115.
    - dlnc (credit growth): Coef. = .1200901; Std. Err. = .0214842; t = 5.59; P>|t| = 0.000; [95% Conf. Interval] = .0773184 to .1628619.
    - dlnpop: Coef. = 5.079725; Std. Err. = 2.821283; t = 1.80; P>|t| = 0.076; [95% Conf. Interval] = -.5370173 to 10.69647.
    - rate: Coef. = -.0017835; Std. Err. = .0008625; t = -2.07; P>|t| = 0.042; [95% Conf. Interval] = -.0035006 to -.0000665.
    - _cons: Coef. = .1242847; Std. Err. = .0373041; t = 3.33; P>|t| = 0.001; [95% Conf. Interval] = .0500179 to .1985514.
- Misalignment measures and thresholds:
  - Misalignment procedure uses base years 2005–2008 (index = 100) and averages gaps across base years.
  - Misalignment threshold for 2016 = above 10 percent of equilibrium price.
  - Estimated average 2016 gap = 13.5 percent.
  - Illustrative base-year gaps: 2005 = 17.76; 2006 = 15.4; 2007 = 0; 2008 = 10.96.
- Traditional valuation ratios:
  - Price-to-income and price-to-rent computed versus historical averages.
  - Overvaluation rule: ratio ≥ one standard deviation above historical average.
  - Since 2011 both ratios have risen steadily; only price-to-rent is above the overvaluation threshold.

### Macroeconomic Effects of a House Price Correction (VAR Results)
- House price shock impacts:
  - Housing shock leads to a decline in real house prices of 2.5 percent after one period.
  - Cumulative decline in real house prices ≈ 15 percent after two years (eight quarters).
- Real investment:
  - Initial decline of 2 percent.
  - Cumulative impact on real investment = -21 percent over two years.
  - Real investment is most sensitive to housing shock.
- Real GDP and private consumption:
  - Both decline by more than 3 percent (cumulative) after 8 quarters.
- Confidence intervals: impulse responses shown with ±2 S.E.
- Conclusion and recommendation:
  - Risks appear contained but a negative house price shock could have sizable effects on GDP, consumption, and investment.
  - Authorities should continue close monitoring of credit and house price growth.

### Productivity, Potential Growth, and Policy Scenarios
- Informality and labor quality:
  - Informality fell 7 percentage points since 2001.
- TFP patterns and measurement:
  - TFP grew strongly in the 2000s; weaker recently.
  - Potential TFP obtained by HP filter (λ=6.25) on TFP residual.
  - Recent TFP decline partly reflects measurement issues (labor hoarding) and sectoral shifts (mining expansion into lower productivity areas).
- Productivity dispersion and misallocation:
  - Large productivity differentials across firms and regions; dispersion larger than in OECD.
  - Correcting misallocation in manufacturing could increase growth by 0.3 percentage points (IMF (2017)).
- Productivity projection scenarios:
  - With very strong reform implementation, productivity growth could reach 1 percent by 2022.
  - TFP growth could rise to 1.1 percent in 2022 if the following fully materialize:
    - productivity gains from 4G infrastructure = 0.25 percentage points;
    - peace dividend = 0.5 percentage points (could reach 0.7 percentage points in long run);
    - tax reform productivity gain = 0.1 percentage points;
    - business environment and trade reforms = 0.1 percentage points.
- Potential growth projections:
  - Medium-term potential growth range = 2.8 to 4.1 percent depending on policy strength.
    - Lower bound 2.8 percent: incomplete reforms, no TFP growth, 4G and tax reform effects on investment not realized.
    - Upper bound 4.1 percent: 1 percent TFP growth by 2022 and full policy impacts.
    - Central scenario medium-term potential growth = 3.5 percent.
  - Capital stock projections:
    - Trend net capital stock growth average = 4.3 percent in 2017‒22 (without 4G or tax reform).
    - Construction phase of 4G projects projected to add up to 0.1 percentage points to average capital stock growth in next five years.
    - Structural tax reform projected to add up to 0.1 percentage points to average capital stock growth.
    - Infrastructure and REER depreciation expected to add further 0.1 percentage points.
    - Overall capital stock projected to increase at annual average of 4.3 to 4.7 percent in 2017‒22.
    - Capital/output ratio projected to climb to 2.6 from 2.4 in 2016.
  - Labor inputs:
    - Potential labor projected to grow slightly above 2 percent per year in 2017‒22 (down from 3.8 percent average in last five years).
    - Working-age population projected to grow at 1.1 percent annually in 2017‒22 (down from 1.7 percent in 2000‒16).
    - LFPR climbed to 64‒65 percent in 2015‒6 and projected to remain around current levels.
    - Unemployment projected to fall back to 9 percent as growth picks up after a modest cyclical uptick in 2017.
    - Human capital assumed to increase around 1 percent annually next five years.
- Infrastructure impacts (Box highlights):
  - Road quality improved to sample median → aggregate GDP growth +0.12 percentage points.
  - 10 percent increase in road network size → growth +0.14 percentage points.
  - 4G projects expected to deliver additional 7,000 kilometers of new primary roads.
  - Median-firm investment rate increases by about 0.1 percentage points if road quality improved to sample median.
- Policy priorities:
  - Execute 4G projects without delays; implement structural reforms to boost productivity.
  - Address resource misallocation, R&D spending, and coordination between knowledge centers and private sector.

### Structural Tax Reform: Components, Revenue, and Simulations
- Reform objective: mobilize about 3 percent of GDP by 2022 to replace lost oil proceeds while fostering employment and growth.
- Key measures and expected revenue yields (in percent of GDP):
  - Total expected revenue: 2017 = 0.7; 2018 = 0.7; 2019 = 1.7; 2020 = 2.5; 2021 = 2.8; 2022 = 3.1.
  - VAT and consumption tax: 2017 = 0.7; 2018 = 0.8; 2019 = 0.8; 2020 = 0.9; 2021 = 0.9; 2022 = 0.9.
    - o/w deduct. VAT on capital goods: 2017 = 0.0; 2018 = -0.2; 2019 = -0.2; 2020 = -0.1; 2021 = -0.1; 2022 = -0.1.
  - Green taxes: 2017 = 0.1; 2018 = 0.1; 2019 = 0.1; 2020 = 0.1; 2021 = 0.1; 2022 = 0.1.
  - Personal income tax: 2017 = 0.0; 2018 = 0.1; 2019 = 0.1; 2020 = 0.1; 2021 = 0.1; 2022 = 0.1.
  - Corporate income tax: 2017 = -0.1; 2018 = -0.3; 2019 = -0.1; 2020 = -0.1; 2021 = -0.1; 2022 = -0.1.
  - Extension of Financial Transaction Tax (FTT): 2017 = 0.0; 2018 = 0.0; 2019 = 0.2; 2020 = 0.4; 2021 = 0.7; 2022 = 0.9.
  - Formalization and Growth 1/: 2017 = 0.0; 2018 = 0.1; 2019 = 0.3; 2020 = 0.7; 2021 = 0.6; 2022 = 0.7.
  - Anti-evasion measures: 2017 = 0.0; 2018 = 0.0; 2019 = 0.3; 2020 = 0.5; 2021 = 0.6; 2022 = 0.6.
- FTT context: current FTT generates about 0.8 percent of GDP; reform counts as net revenue the extension of FTT scheduled to phase out starting 2019.
- Simulation setup (model and assumptions):
  - Model: Flexible System of Global Models (FSGM), semi-structural GE model with OLG and liquidity-constrained consumers, rules-based fiscal and monetary policy.
  - Public capital affects TFP and returns to private investment; fiscal rule ensures stable debt-to-GDP.
  - Simulations devote revenue from PIT, CIT, and VAT to public investment; endogenous monetary policy response included.
  - 4G infrastructure capex (modeled as public investment, 2016–2021):
    - 2016 = 0.23; 2017 = 0.94; 2018 = 1.17; 2019 = 1.05; 2020 = 0.81; 2021 = 0.37 (percent of GDP).
- Key simulation findings:
  - Growth impact: reform could boost growth by about 0.3−0.4 percent of GDP per year, building up as reform yield and additional public investment rise (0.5−0.6 percent of GDP in 2017−18, increasing to ≈0.8 percent thereafter).
  - Reform boosts GDP above trend; contribution to potential GDP (including TFP boost) is milder than headline GDP impact.
  - Private investment and consumption channels:
    - VAT increase → temporarily lower consumption.
    - Lower corporate taxation + additional public capital → higher private investment.
    - After 2018 private consumption recovers and exceeds no-reform scenario.
    - 4G agenda reinforces positive impacts on investment and consumption.
  - Public investment and TFP: reform proceeds enable additional public investment, raising public capital and cumulatively increasing TFP levels.
  - Inflation and monetary policy response:
    - Early GDP above potential → higher inflation and endogenous monetary tightening.
    - Simulated policy rates: about 50bp higher under tax reform scenario; about 125bp higher under tax reform plus 4G scenario.
    - Growth impacts reflect joint fiscal changes and monetary response.
- Concluding remarks and extensions:
  - Main conclusion: the 2016 tax reform advances authorities’ objective to promote growth via private investment boosts and additional public investment.
  - Suggested extensions: explicitly model FTT and green taxes, effects on risk premia and cost of funding, and the quality/effectiveness of public capital.

*Source: IMF staff analysis as presented in the supplied chapter content.*

### 1. Interest Rates _________________________________________________________________________ 7

### 1. Interest Rates

### Introduction
- The monetary policy tightening cycle has come to an end after the authorities raised monetary policy rates by 325 bps to 7.75 percent since September 2015.
- In December 2016 the Monetary Policy Committee (MPC) cut the policy rate by 25 bps. The MPC cited three main reasons:
  - (1) the impact of transitory shocks that pushed inflation away from the target band continued to dissipate rapidly;
  - (2) real GDP growth during 2016Q3 was lower than expected; and
  - (3) ex-ante real interest rates were above their average since 2005.

### Neutral Interest Rate — estimates and DSGE results
- Traditional methodologies (following Magud and Tsounta (2012)) applied yield an estimated real neutral interest rate of 1.7 percent. Methods used include:
  - the uncovered interest parity condition (UIP);
  - an expected-inflation augmented Taylor rule;
  - a consumption-smoothing (consumption-based CAPM) approach;
  - the Hodrick-Prescott (HP) filter.
- A reduced-form small open-economy DSGE model (estimated with Bayesian methods using priors from Gonzalez and Hamann (2011) and Hamann et al (2014)) finds:
  - The real neutral rate has declined since 2007.
  - The DSGE estimate of the current real neutral rate is about 0.7 percent.
  - The level of the estimated neutral rate depends on model specification and shock assumptions.
- Other neutral-rate evidence:
  - The Central Bank provided a range of estimates with the median at 1.4 percent (BanRep 2015).
- Key model facts and parameterization (as presented):
  - IS curve, Phillips curve, and Taylor rule summarized the economy; the neutral rate r* follows a persistent process.
  - The depreciation of the real exchange rate is modeled as an AR(1) with coefficient 0.72.
  - The neutral interest rate process: r*_t = 1.1 + z_t; z_t = 0.91 z_{t-1} + u_t.

### Monetary stance and policy-rate comparisons
- Policy actions and levels:
  - Policy rate increases: raised by 325 bps to 7.75 percent between September 2015 and August 2016.
  - After recent cuts and with inflation expectations at 4 percent, the current (ex-ante) real policy rate is around 3 percent.
- Comparison with neutral rate:
  - Staff’s assessment: neutral range of 1−2 percent.
  - DSGE estimate: real neutral rate about 0.7 percent.
  - Conclusion: The monetary policy tightening has brought the policy rate above the neutral rate; the stance is tight.

### Financial Conditions Index (FCI) — methodology and findings
- Purpose: capture overall financial conditions through credit, leverage, and risk channels while purging direct effects of monetary policy and business cycle.
- Methodology:
  - A factor-augmented vector-autoregression (FAVAR) framework is used to estimate the FCI.
  - Time-varying parameter FAVARs (TVP-FAVARs) following Koop and Korobilis (2014) are implemented.
  - Two additional specifications estimated: PCA-FAVAR (Stock and Watson 2002; Bernanke et al 2005) and KF-FAVAR (Doz, Giannone, and Reichlin 2011).
  - Empirical model: X_t (20 financial variables) and Y_t (inflation, real GDP growth, REER, policy rate) with an unobservable factor f_t interpreted as the FCI.
- Composition:
  - Variables cover three main dimensions: risk (interest rate spreads and asset price volatilities), collateral values and leverage (bank leverage, housing and stock prices), and credit (commercial, consumption, and mortgage).
- Normalization and controls:
  - FCI normalized around 0 over 2001–16: values above 0 = “loose”, below 0 = “tight.”
  - The constructed FCI is purged of cyclical influence by controlling for GDP growth and controls for the policy rate so it does not capture direct effects of monetary policy decisions.
- Key empirical findings:
  - Financial conditions deteriorated and remained tight during 2016.
  - Impulse response estimates: a 1 standard deviation tightening of financial conditions lowers GDP growth by about 0.25 percentage points within 12 months.
  - The estimated FCI suggests financial conditions tightened by one standard deviation in 2015, and by one fifth of one standard deviation in 2016.
  - The cumulative tightening implies a drag on growth in 2017 subtracting around 0.2 ppts.

### Conclusions and policy implications
- Assessment:
  - The monetary policy tightening brought the policy rate above the neutral rate; with inflation on a decreasing path and expectations inside the target band, monetary policy stance is tight.
  - Broad financial conditions remained tight during 2016 and are expected to continue weighing on growth.
- Policy guidance:
  - The central bank has room to continue with the easing cycle given inflation trends and expectations; however, the precise timing of cuts should depend on a consolidation of these trends.
  - Monitoring of financial conditions is important because tightening in the FCI has a measurable negative impact on GDP growth (about 0.25 percentage points per 1 standard deviation tightening within 12 months).

*Prepared by Francisco Roch; IMF staff analysis as presented in the Monetary Policy and Financial Conditions chapter.*

### 1.      Colombian house prices have increased significantly between 2005 and 2016. In the main

### cr17139 - 1.      Colombian house prices have increased significantly between 2005 and 2016. In the main

### Recent developments in house prices and housing market
- House prices in the main three cities (Bogota, Cali, and Medellin) rose by around 200 percent in nominal terms (110 percent in real terms) from 2005 to mid-2016.
- Average home-price index grew by an annual real rate of 5.18 percent between 2005 and 2016.
  - Bucaramanga: highest average annual real growth rate, 8 percent.
  - Armenia: lowest average annual real growth rate, 3.4 percent.
- Price growth was stronger for middle and upper income levels across cities.
- Home ownership and housing deficits:
  - Out of 14 million households in 2015: 46 percent live in their own homes, 38 percent rent, and 16 percent enjoy free usufruct.
  - Total housing deficit estimated at 25 percent of households in 2014, down from 54 percent in 1993.
  - In 2014, the quantitative deficit was 9 percent and the qualitative deficit was 15 percent of households.
  - Total housing deficit in 2014: rural areas 48 percent, urban areas 18 percent.
  - Quantitative deficit in rural areas increased from 7 percent in 1993 to 18 percent in 2014.
- Construction activity:
  - Construction share in GDP bottomed at around 4 percent in 2000 and reached 9.7 percent in 2015.
  - Construction licenses index rose strongly except during the 2008 financial crisis; construction licenses declined in 2016 suggesting potential slowing.

### Mortgage credit and government programs
- Mortgage credit developments:
  - Real mortgage credit has grown by an annual average of 13 percent since 2011.
  - Mortgage credit stands at about 5 percent of GDP.
  - Around 30 percent of the mortgages originated in 2016 were subsidized.
- Government housing programs (summary of features and scale):
  - Programs combine support for down payments and subsidized mortgage interest rates.
  - Mi Casa Ya: targets households with income between 2 and 4 minimum wages and for houses worth up to US$30,000; offers down payment support and an interest subsidy.
  - FRECH: offers interest rate subsidies (4−5 percentage points) targeting households earning up to 8 minimum wages and houses worth up to US$50,000.
  - FRECH No-VIS: provides interest rate subsidies (2.5 percentage points) for houses worth up to US$77,000 and is open to any household regardless of income.
  - The government sets annual ceilings (cupos) for the number of houses to be subsidized under each program; the programs are only for new houses.
  - Number of houses sold with mortgage subsidies increased from about 21,000 in 2013 to about 33,000 in 2016.
  - Official estimates suggest the construction sector is interconnected with 25 percent of the industrial sector.
  - Official estimates suggest the fiscal multipliers of the housing program vary from 10 for FRECH No-VIS to 1 for the free housing program.

### Housing finance characteristics and resilience factors
- Regulatory and market features that mitigate fragility:
  - Ceiling on loan-to-value (LTV) ratios of 70 percent (80 percent for social interest housing).
  - Mortgages are usually contracted at a fixed interest rate (though both variable- and fixed-rate mortgages are offered).
  - There is full recourse on mortgages.
- Cross-country context and Colombian specifics:
  - Maximum observed LTV ratios across countries have a median of 83 percent; they range from 70 percent (Colombia, Hong Kong SAR, Hungary) to 125 percent (the Netherlands).
  - In Colombia, the LTV ratio is at 53 percent, below the regulatory ceiling.
  - Term to maturity:
    - Mortgages across countries range from 7 years (Turkey) to 45 years (Sweden), median 25 years.
    - Colombia: average maturity is 15 years; effective duration is around 6−7 years due to early prepayments.
  - Funding model: banks primarily use retail deposits as the primary source of financing in Colombia.
  - Degree of lender recourse: Colombia has full recourse on mortgages.
  - Interest type: both variable- and fixed-rate mortgages are offered; majority of households tend to choose fixed-rate mortgages.
  - Prepayment penalties: there are no prepayment penalties in Colombia and borrowers frequently make partial prepayments.

### Main findings on misalignment and risk assessment
- Estimated misalignment:
  - Model results suggest that house prices are around 13 percent above equilibrium.
- Risk mitigation:
  - After the financial crisis of 1999, authorities implemented measures addressing mortgage financing vulnerabilities.
  - The soundness in housing finance contracts and regulations (LTV ceiling, predominance of fixed-rate mortgages, full recourse) would mitigate negative spillovers from a potential correction in house prices.

### Modeling approach and determinants of house prices
- Determinants included in standard models and the paper’s approach:
  - Past growth rates of real house prices (persistence).
  - Past housing affordability ratio (ratio of house prices to income) as the long-term equilibrium relationship.
  - Economic fundamentals: per capita real income growth (positive effect), mortgage credit growth (positive effect), interest rates (negative effect), and population growth (positive effect).
- Empirical method:
  - Misalignments estimated using an error-correction model relating short-term changes in house prices to a long-term equilibrium relationship, interest rates, changes in income per capita, and credit growth.
  - Regression specification (changes in house prices as dependent variable) uses demand-side explanatory variables including:
    - Change in real GDP per capita over the last quarter.
    - Affordability level of housing in the previous period (ratio of house prices to income).
    - Change in real mortgage over the past year.
    - An average lending interest rate.
    - Change in population over the past year.

*Prepared by Francisco Roch. Staff from the Banco de la República provided useful comments.*

### 18.      The estimation uses quarterly data from the first quarter of 1994 until the first quarter

### 18.      The estimation uses quarterly data from the first quarter of 1994 until the first quarter

### Data and methodology
- Sample period for econometric estimation of house prices: first quarter of 1994 until the first quarter of 2016.
- Price measure: existing home price index from the BanRep database, deflated by the consumer price index (CPI).
- Geographic coverage of the price index: Bogota, Medellin, Cali, Barranquilla, Bucaramanga, Cucuta, Manizales, Neiva, Villavicencio and Soacha.
- Housing affordability metric: ratio of house prices to real GDP per capita.
- VAR estimation for macroeconomic effects: quarterly data from the first quarter of 1994 until the second quarter of 2016.
  - Domestic variables in the VAR: real GDP, real private consumption, real investment, 90-day certificates of deposit interest rate (nominal), and the real existing house price index.
  - Variable transformations: natural logarithms for all variables except the nominal interest rate (introduced in levels).
  - VAR lag length: 3 lags.
- Identification of house price shocks: generalized impulse response approach (invariant to variable ordering).
- Robustness note: results could change if the new home price index is used due to the historical gap between the existing and new home price indices. This robustness check is left for future work.

### Econometric results on house prices
- Long-run behavior:
  - Real house prices in Colombia show long-run reversion to fundamentals.
  - R-square of the regression: 0.45.
  - Adjusted R-squared: 0.4147.
  - Prob > F = 0.0000; F(6, 78) = 10.92; Number of obs = 85.
- Speed of correction:
  - Affordability is negatively related to changes in house prices and is statistically significant: misalignment is corrected at about 13 percent every quarter.
- Coefficient signs and significance:
  - Better mortgage credit availability and higher income per capita promote higher house prices.
  - Demographic factors have a positive effect on house prices.
  - The coefficient for the past growth rate of real house prices is negative (current growth rate is negatively correlated with the past growth rate), contrary to expectation.
- Selected regression output (excerpt):
  - lnafford1 (affordability lag): Coef. = -.132468; Std. Err. = .0315063; t = -4.20; P>|t| = 0.000; [95% Conf. Interval] = -.1951922 to -.0697437.
  - hpi1 (lagged hpi): Coef. = -.4009662; Std. Err. = .0926514; t = -4.33; P>|t| = 0.000; [95% Conf. Interval] = -.5854208 to -.2165115.
  - dlnc (change in credit?): Coef. = .1200901; Std. Err. = .0214842; t = 5.59; P>|t| = 0.000; [95% Conf. Interval] = .0773184 to .1628619.
  - dlnpop: Coef. = 5.079725; Std. Err. = 2.821283; t = 1.80; P>|t| = 0.076; [95% Conf. Interval] = -.5370173 to 10.69647.
  - rate: Coef. = -.0017835; Std. Err. = .0008625; t = -2.07; P>|t| = 0.042; [95% Conf. Interval] = -.0035006 to -.0000665.
  - _cons: Coef. = .1242847; Std. Err. = .0373041; t = 3.33; P>|t| = 0.001; [95% Conf. Interval] = .0500179 to .1985514.

### Price misalignment and traditional valuation measures
- Misalignment estimation method:
  - Use levels of house prices as of the first quarter in years from 2005 to 2008 as alternative base levels (index = 100 at chosen base).
  - Compute predicted index values from regression-implied house price changes; estimated price gap = actual index value − predicted index value.
  - Misalignment measure: average estimated gap across the alternative base years to ensure robustness.
- Misalignment threshold for 2016: prices considered misaligned if the average estimated gap is above 10 percent of the equilibrium price.
- Estimated average 2016 gap: 13.5 percent (average estimated price gap).
- Table 2 (illustrative gaps by base year, excerpt):
  - Base year 2005 gap: 17.76
  - Base year 2006 gap: 15.4
  - Base year 2007 gap: 0
  - Base year 2008 gap: 10.96
  - (Presentation in source shows a set of gaps; the average is reported as 13.5 percent.)
- Traditional valuation ratios:
  - Price-to-income and price-to-rent ratios computed using historical averages over the entire data period.
  - Overvaluation rule: ratio at least one standard deviation above its historical average.
  - Both valuation measures have been increasing steadily since 2011, but only the price-to-rent ratio is above the overvaluation threshold (i.e., suggests overvaluation).

### Macroeconomic effects of a house price correction (VAR results)
- Impact on house prices:
  - Housing shock leads to a decline in real house prices of 2.5 percent after one period.
  - Cumulative decline in real house prices of about 15 percent after two years (eight quarters).
- Impact on real investment:
  - Initial decline of 2 percent following the shock.
  - Cumulative impact on real investment of -21 percent over two years.
  - Real investment is the macroeconomic variable most sensitive to the housing shock.
- Impact on real GDP and private consumption:
  - Both real GDP and private consumption decline by more than 3 percent (cumulative) after 8 quarters.
- Confidence intervals: responses shown with ±2 S.E. in Figure 8 (impulse responses).

### Conclusions and policy implications
- Overall assessment:
  - Risks from the housing market seem to be contained.
  - VAR results indicate a negative house price shock could have sizable implications for GDP, consumption, and investment.
  - House prices do not seem to be largely above levels justified by economic fundamentals (average estimated gap is at 13 percent).
- Mitigating factors:
  - Post-1999 financial crisis macroprudential measures such as the use of LTV limits, together with other housing financing characteristics, limit vulnerabilities from the housing market.
  - Current slowdown in economic activity should decelerate mortgage growth and impact house price growth, which has started to show signs of weakening.
- Policy recommendation:
  - Authorities should continue to monitor closely developments in credit and house price growth.

### Brief links to potential growth analysis (production-function insights)
- Production function used to estimate and project potential output for 1990‒2022: Y = A * K^{α1} * L^{1−α1} (formula symbolically presented in source).
- Capital stock:
  - Machinery and equipment increased from 3 percent to 9 percent of the total capital stock over the last three decades (due to oil/mining).
  - Oil/mining investment increased from 2½ percent of GDP in 2005 to 4½ percent of GDP in 2009‒13.
  - Trend scenario projection for net capital stock growth: average of 4.3 percent in 2017‒22 (without 4G projects or tax reform).
  - Construction phase of 4G projects projected to add up to 0.1 percentage points to the average growth rate of the capital stock in the next five years.
  - Structural tax reform projected to boost private investment, adding up to 0.1 percentage points to the average growth rate of the capital stock.
  - Infrastructure and real exchange rate depreciation expected to add a further 0.1 percentage points to the average growth rate of the capital stock.
  - Overall capital stock projected to increase at an annual average of 4.3 to 4.7 percent in 2017‒22.
  - Capital/output ratio projected to climb to 2.6 from 2.4 in 2016.
- Labor input:
  - Potential labor projected to grow slightly above 2 percent per year in 2017‒22, down from an average of 3.8 percent in the last five years.
  - Working age population projected to grow at an annual average of 1.1 percent in 2017‒22, down from 1.7 percent in 2000‒16.
  - Labor force participation rate (LFPR) climbed to an all-time high of 64‒65 percent in 2015‒6 and is projected to remain around current levels.
  - Unemployment projected to fall back to 9 percent as growth picks up after a modest cyclical uptick in 2017.
  - Human capital assumed to continue increasing at around one percent annually in the next five years.
- Impact of infrastructure (Box 1 highlights):
  - If road quality improved to the sample median, aggregate GDP growth would increase 0.12 percentage points.
  - A 10 percent increase in the size of the road network would increase growth by 0.14 percentage points.
  - 4G infrastructure projects expected to deliver an additional 7,000 kilometers of new primary roads.
  - Corporate investment effects: median-firm investment rate would increase by about 0.1 percentage points if road quality improved to the sample median.

*Source: IMF staff analysis as presented in the supplied chapter content.*

### 11.      Falling labor market informality helped improve

### 11.      Falling labor market informality helped improve

### Labor market informality and labor quality
- Informality fell 7 percentage points since 2001, indicating that the average quality of the Colombian labor force improved.
- Informal labor is less productive than formal labor (Perry and others 2007, Dabla-Norris and others 2005).
- Some provisions in the structural tax reform might extend this trend.

### Productivity: recent patterns and measurement
- Total factor productivity (TFP) grew significantly in the 2000s but was weaker in recent years.
- TFP is obtained as a residual from the log version of equation (1), using actual GDP, the capital stock adjusted by utilization, and actual employment adjusted by human capital. Potential TFP is the result of applying an HP filter (λ=6.25) to the TFP residual.
- Factors explaining the significant productivity improvements starting at the turn of the century include:
  - a sharp reduction in crime,
  - progress in financial deepening and financial inclusion (Karpowicz 2014),
  - a better business environment (e.g., cutting almost by half the days it takes to get a construction permit),
  - favorable terms of trade dynamics (IMF 2017).
- The recent decline in trend TFP is influenced by the sharp drop in measured TFP at the time of the financial crisis, which may largely reflect factors such as labor hoarding rather than fundamental productivity changes.
- Sosa and others (2013) hypothesize part of the recent decline in TFP could be related to the expansion of mining into areas of lower marginal productivity where production became profitable due to the commodity price boom.
- The decline in TFP growth is also a global phenomenon (Jones 2015, Haldane 2017).

### Measurement limitations and alternative metrics
- TFP, as a residual, can be contaminated by measurement errors in labor and capital inputs (e.g., changes in the quality of the capital stock or fluctuations in hours worked).
- Labor productivity in the non-primary sector is an alternative measure that avoids some pitfalls of TFP; Colombia’s labor productivity grew strongly in 2003-08 by this metric, though part of the improvement is due to capital accumulation rather than pure productivity.
- In international perspective, Colombia’s productivity is relatively low compared to other EMs.

### Productivity dispersion and resource misallocation
- Aggregate TFP figures mask large productivity differentials across firms and regions.
- The dispersion in firm productivity in Colombia is larger than in OECD countries (Brown and others 2013).
- IMF (2017) finds evidence of resource misallocation in the manufacturing sector in Colombia. Correcting it could increase growth by 0.3 percentage points.
- Low spending in R&D and weak coordination between centers producing knowledge and the private sector may hinder productivity growth; CONPES (P.40) notes this issue and points to policy measures to address it.

### Productivity projections and reform scenarios
- Going forward, productivity growth could reach one percent in a scenario with very strong reform implementation but could be much weaker in the absence of reforms.
- TFP growth would gradually increase to 1.1 percent in 2022 if the positive effects of the following factors materialize in full:
  - Productivity gains from better infrastructure: 0.25 percentage points (improvements from the 4G projects).
  - The peace dividend: 0.5 percentage points (further improvements in security from the peace agreement with the FARC). In the long run, the peace dividend could grow further, reaching up to 0.7 percentage points in terms of productivity growth.
  - Productivity gains from the tax reform: 0.1 percentage points (simplification of the tax system reduces filing costs).
  - Ongoing efforts to improve the business environment and reduce barriers to trade: 0.1 percentage points (streamlining regulations, customs code, simplification of import tariffs, removing nontariff barriers, promoting new export markets and products).

### Potential growth: drivers and projections
- Factor accumulation, especially labor, drove potential growth up to an average of 4 percent in the last 15 years.
  - High population growth, a large drop in the unemployment rate, and improving schooling rates increased labor inputs significantly.
  - Together with increasing capital accumulation and some TFP gains, these developments lifted potential growth to above 4 percent on the eve of the global financial crisis.
- Medium term potential growth would be in the 2.8 to 4.1 percent range depending on the strength of policies.
  - The lower bound (2.8 percent) corresponds to an unlikely scenario with incomplete reform implementation resulting in no TFP growth, and the effects of 4G infrastructure projects and the tax reform on investment not materializing.
  - The upper bound (4.1 percent) corresponds to one percent TFP growth by 2022 due to very strong structural reforms and full impact of policies on capital accumulation.
  - A central scenario with medium term potential growth of 3.5 percent is presented; this central scenario still involves a significant turnaround in productivity that requires strong reform implementation.
- Further reductions in the natural unemployment rate and informality could increase potential growth but may be offset by growing skill mismatches and concerns about the quality of education.
- Declining potential growth is a common theme among commodity exporters and highlights the importance of structural reforms to grow in a world of low oil prices.

### Conclusion: policy implications and priorities
- Potential growth is likely to moderate to a range of 2.8 to 4.1 percent in a world of lower oil prices.
- The production function results indicate potential growth rose to 4−5 percent in the era of high oil prices and fell considerably with the sharp drop in oil prices.
- The outlook is positive under successful implementation of structural reforms:
  - The 4G infrastructure projects and the tax reform will increase investment, offsetting the sharp decline in capital accumulation in the oil sector. Capital accumulation is projected to contribute 1.4 to 1.6 percentage points to average annual potential growth in 2017‒22, highlighting the importance of executing 4G projects without delays.
  - Improvements in productivity are essential to lift potential growth. The large increases in the labor force observed in the last 15 years are unlikely to continue, making productivity-enhancing reforms central for growth. Better infrastructure, the peace agreement with the FARC, and ongoing efforts to improve the business environment and reduce trade barriers are identified as key steps to boost productivity and growth.

### Tax reform: revenue and key measures (summary)
- The structural tax reform approved in December 2016 aims to mobilize revenue to replace lost oil proceeds while fostering employment and growth.
- The reform aims to collect about 3 percent of GDP by 2022 to keep a broadly constant level of primary expenditure as a ratio of GDP.
- Key measures and expected revenue yields:
  - An increase in the VAT general rate by 3 percentage points (from 16 to 19 percent) will generate about 1 percent of GDP.
  - Anti-evasion measures and gains from formalization and growth are each expected to generate about 0 (text truncated in source at that point).

*Source: IMF staff report (chapter content provided).*

### 0.6 percent of GDP. The reform also includes green taxes and some reductions in the exemptions for

### Colombia: Key Elements of Structural Tax Reform

### Reform components and revenue impact
- Reform includes green taxes and some reductions in the exemptions for Personal Income Tax (PIT).
- The reform counts as net revenue the extension of the existing Financial Transaction Tax (FTT) that was scheduled to gradually disappear starting in 2019 (currently, the FTT generates about 0.8 percent of GDP).
- Reform expected revenue by component (In Percent)
  - Total: 2017 0.7; 2018 0.7; 2019 1.7; 2020 2.5; 2021 2.8; 2022 3.1
  - VAT and consumption tax: 2017 0.7; 2018 0.8; 2019 0.8; 2020 0.9; 2021 0.9; 2022 0.9
    - o/w deduct. VAT on capital goods: 2017 0.0; 2018 -0.2; 2019 -0.2; 2020 -0.1; 2021 -0.1; 2022 -0.1
  - Green taxes: 2017 0.1; 2018 0.1; 2019 0.1; 2020 0.1; 2021 0.1; 2022 0.1
  - Personal income tax: 2017 0.0; 2018 0.1; 2019 0.1; 2020 0.1; 2021 0.1; 2022 0.1
  - Corporate income tax: 2017 -0.1; 2018 -0.3; 2019 -0.1; 2020 -0.1; 2021 -0.1; 2022 -0.1
  - Extension of Financial Transaction Tax: 2017 0.0; 2018 0.0; 2019 0.2; 2020 0.4; 2021 0.7; 2022 0.9
  - Formalization and Growth 1/: 2017 0.0; 2018 0.1; 2019 0.3; 2020 0.7; 2021 0.6; 2022 0.7
  - Anti-evasion measures: 2017 0.0; 2018 0.0; 2019 0.3; 2020 0.5; 2021 0.6; 2022 0.6
- For simulation:
  - VAT excluding capital goods: 2017 0.7; 2018 1.0; 2019 1.0; 2020 1.0; 2021 1.0; 2022 1.0
  - CIT + VAT on capital goods: 2017 -0.1; 2018 -0.5; 2019 -0.2; 2020 -0.2; 2021 -0.2; 2022 -0.2
- Expected revenue from selected measures (in percent of GDP) as presented:
  - 2017 2018 2019 2020 2021 2022
  - Combined No Reform424334343434
  - CIT343434343434
  - CREE890000
  - Combined Reform403733333333
  - CIT343333333333
  - CREE640000
  - Memo: Reform-No Reform-2-6-1-1-1-1

### Simulations and modeling approach
- Literature benchmarks:
  - Fiscal multipliers in emerging markets are generally smaller; tax multipliers are similar to expenditure multipliers (in the range of 0.1−0.4).
  - Rincon et al (2014) find a public investment multiplier between 0.4 and 0.5; public consumption multiplier on average about -0.02.
  - Vargas et al (2012) estimate expenditure multipliers of about 0.4.
- Model used: Flexible System of Global Models (FSGM)
  - Semi-structural general equilibrium model calibrated to specific countries.
  - Incorporates rules-based fiscal and monetary policy, OLG (Non-Ricardian) and liquidity constrained consumers (LIQ).
  - Private consumption and investment are micro-founded; other parts use reduced-form.
  - Public capital (deviations from trend) positively affect total factor productivity and returns to private investment (Tobin’s Q model of quadratic adjustment costs).
  - Fiscal policy modeled as combination of indirect and direct taxes, oil-related royalties, and lump-sum transfers; fiscal rule ensures stable debt-to-GDP ratio; monetary policy guided by a Taylor rule.

### Simulation results and channels
- Scope of simulations:
  - Joint impact of revenue and expenditure measures: expected revenue yield for PIT, CIT and VAT is devoted to public investment.
  - Endogenous monetary policy response is included: if GDP rises above potential, inflationary pressures and rate hikes will follow.
  - Additional growth reported is relative to a no-reform steady state.
- 4G infrastructure agenda modeling:
  - Simulation includes expected 4G capex between 2016-2021 (modeled as public investment although in practice PPP-based).
  - 4G Infrastructure Agenda Capex (In Percent of GDP): 2016 0.23; 2017 0.94; 2018 1.17; 2019 1.05; 2020 0.81; 2021 0.37
- Key findings:
  - Growth impact: the reform could boost growth by about 0.3−0.4 percent of GDP per year which builds up gradually as reform yield and additional public investment start small (0.5-0.6 percent of GDP in 2017−18) and then increase slightly after (to 0.8 percent of GDP).
  - The reform would boost GDP above its trend level; contribution to potential GDP (including TFP boost from public investment) will be milder than the impact on headline GDP.
  - Private investment and consumption:
    - Increase in VAT leads to temporarily lower consumption.
    - Combination of lower corporate taxation and additional public capital boosts private investment.
    - After 2018, private consumption recovers and exceeds the no-reform scenario, aided by additional labor income from expanded production and employment.
    - The 4G agenda reinforces positive impacts on investment and consumption.
  - Public investment and total factor productivity:
    - Reform proceeds allow additional public investment, increasing public capital and cumulatively increasing TFP levels.
  - Inflation and monetary policy response:
    - Headline GDP exceeding potential in early years leads to higher inflation and endogenous monetary policy tightening.
    - Simulations show monetary policy rates would be about 50bp higher under a tax reform scenario and about 125bp higher under a tax reform plus 4G scenario.
    - The reported growth impacts are the joint result of fiscal changes and the monetary policy response.

### Concluding remarks and potential extensions
- Main conclusion:
  - The simulations suggest the tax reform approved in 2016 stands to advance the authorities’ objective to promote growth, driven by boosts to private investment and additional public investment enabled by the reform.
- Potential extensions and refinements:
  - Model the impact of the extension of the FTT and green taxes explicitly; FTT may impact growth via the credit channel; green taxes may affect sectors differently based on fossil fuel reliance.
  - Incorporate potential effects on risk premia and the cost of funding for firms and consumers; inclusion of a reduction in risk premia could generate higher private investment.
  - Account for the quality/effectiveness of public capital by parameterizing how resources devoted to public investment become productive public capital.

*Source: National authorities and Fund staff estimates*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2017/cr17139.pdf_