## Appendix 1. Details of Regression: Growth in Housing Price (Georgetown)

## Source details

**Canonical URL:** [Appendix 1. Details of Regression: Growth in Housing Price (Georgetown)](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020020-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020020-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020020-print-pdf.pdf.json)

---

### Objective and main conclusion
- Objective:
  - Construct a baseline model for the growth rate in the Urban (Georgetown) Housing Prices and compare it (out-of-sample) with actual changes from 2016 to 2018 to assess whether historical predictors account for the post-2015 rise.
- Main conclusion:
  - Deviations between observed housing price growth and the predictor in 2016–2018 are not statistically significant; differences average around 0.5 standard deviation and hypothesis tests based on t-distribution suggest no significant deviations in 2016, 2017, and 2018.
- Caveats:
  - Absence of a real estate price index; the Urban (Georgetown) Housing Index (a sub-index of the Urban CPI that includes rent, fuel, and power) is used as a proxy.
  - Unavailability of housing market–related indicators: statistics on volume of houses and dwellings, volume of real estate loans, and consumer and business sentiment indices are not produced.

### Data and context (selected statistics and trends)
- Residential real estate prices in Georgetown have risen following the discovery of oil in 2015.
- Commercial banks’ private dwellings mortgage lending:
  - Grew at an average annual rate of 7.7 percent from 2014-2018.
  - Share of private dwellings mortgage loans to total loans increased from 31.5 percent to 34.9 percent during 2014-2018.
- Household exposure to mortgage debt:
  - In Georgetown, exposure accounted for around 350 percent of income in mid-2019 (source: Numbeo).
  - Comparable average in other major cities in neighboring countries: around 190 percent of income (mid-2019, Numbeo).
- Commercial banks’ housing loans growth:
  - From 2014 to 2018, commercial banks’ housing loans grew by an average annual rate of 12.1 percent.
  - Commercial banks account for about 70 percent of financial system total assets and originate two thirds of total real estate mortgage loans.

### Regression specification (baseline predictor)
- Dependent variable:
  - Yi = Urban (Georgetown) housing prices (in percent change, Y/Y).
- Explanatory variables:
  - X1,i = Urban (Georgetown) CPI—All Items (in percent change, Y/Y).
  - X2,i = Oil Price (in percent change, Y/Y) (petroleum spot price (APSP), in US$/barrel).
  - X3,i = Population growth (in percent change, Y/Y).
  - X4,i = Share of urban population (in percent total population).
- Model form:
  - Yi = αi + β1 X1,i + β2 X2,i + β3 X3,i + β4 X4,i + εi.
- Data sources:
  - Urban (Georgetown) housing prices and Urban (Georgetown) CPI—All Items: Bank of Guyana (Statistical Bulletin).
  - Oil Prices: IMF World Economic Outlook (based on Petroleum spot price (APSP), in US$/barrel).
  - Population: United Nations (World Population Prospects 2019).
  - Share of urban population: The World Bank (World Development Indicators).

### Regression results (reported coefficients and diagnostics)
- Two model specifications:
  - Model 1(a): sample period 1995-2015 (used to predict out-of-sample 2016–2018).
  - Model 1(b): sample period 1995-2018 (full sample).
- Coefficient estimates (standard errors in brackets):
  - Urban CPI All Items:
    - Model 1(a): 0.5533** (0.2352)
    - Model 1(b): 0.58802** (0.20928)
  - Change in Oil Price:
    - Model 1(a): 0.04692* (0.02651)
    - Model 1(b): 0.0415* (0.02277)
  - Population Growth:
    - Model 1(a): 6.75491** (2.98852)
    - Model 1(b): 6.51903** (2.71445)
  - Share of Urban Population:
    - Model 1(a): 3.87884*** (0.94874)
    - Model 1(b): 3.87821*** (0.8784)
  - Constant:
    - Model 1(a): -106.4155*** (26.32431)
    - Model 1(b): -106.5723*** (24.36331)
- Goodness of fit and diagnostics:
  - R-squared:
    - Model 1(a): 0.74086
    - Model 1(b): 0.7589
  - Adjusted R-squared:
    - Model 1(a): 0.67608
    - Model 1(b): 0.7082
  - Durbin-Watson statistics:
    - Model 1(a): 2.5705
    - Model 1(b): 2.5199
  - Note: Coefficients are in expected signs; housing prices and the predictor are cointegrated and no significant autocorrelations are detected (Engle-Granger test for cointegration; Breusch-Godfrey serial correlation LM Test referenced).
  - Significance notation: ***p<0.01; **p<0.05; *p<0.10.

### Interpretation of results and hypothesis tests
- The predictor explains a large share of variation in housing price growth (R-squared 0.74086 in Model 1(a); 0.7589 in Model 1(b)).
- Out-of-sample predictions for 2016–2018 closely track actual housing price growth; average deviation ~0.5 standard deviation.
- From 2017 to 2018 the pace of growth diverged: predictor showed deceleration while actual growth accelerated slightly—this could signal either a potential structural break or omitted variables.
- Hypothesis testing (H0: (A)-(B) = 0; N = 24; Mean = -0.10; Standard deviation = 2.3308):
  - 2016: Test statistic Zi = 0.13; tcrit,i = 1.711; Result: Accept H0.
  - 2017: Test statistic Zj = -0.52; tcrit,j = -1.711; Result: Accept H0.
  - 2018: Test statistic Zk = -0.49; tcrit,k = -1.711; Result: Accept H0.
- Annual Actual (A), Predictor (B), and Difference (A)-(B) for 1995–2018 are reported; selected entries:
  - 2016: Actual (A) -0.19; Predictor (B) 0.00; Difference (A)-(B) 0.19
  - 2017: 1.82; 0.51; -1.31
  - 2018: 2.34; 1.11; -1.24

### Missing variables and suggested extensions
- Potential omitted variables to include in extended models:
  - Changes in property taxes (2018 National Budget reduced property tax rates from 0.75 percent to 0.5 percent for the first GY$ 20 million in taxable net property, with the remainder taxed at 0.75 percent).
  - Demand for real estate from foreigners and Guyanese residing abroad.
  - Changes in housing supply (construction and residential permits).
  - Distribution of income and household debt across different population groups.
  - Business and household sentiment indicators.
- Data collection improvements:
  - Expand data collection to include commercial real estate indicators, volume of houses and dwellings, volume of real estate loans, and consumer and business sentiment indices.
  - Increase frequency of data to improve forecasting accuracy and surveillance.

### Policy implications and recommendations
- Surveillance and supervision:
  - Step up monitoring of housing market developments, household debt, and bank lending practices given the potential for a structural break or missing variables.
  - Enhance supervisory surveillance on banks’ mortgage loans and underwriting standards to mitigate risks.
  - Intrusive and tighter supervision, including enforcement actions, would be needed if lax underwriting standards are detected.
- Macroprudential toolkit:
  - Fortify the macroprudential policy toolkit to be ready to tighten measures when necessary to safeguard financial and macroeconomic stability.
  - Potential longer-term tools: countercyclical capital buffers or systemic risk surcharges.
  - Targeted sectoral macroprudential tools: sectoral capital requirements (risk weights or LGD floors); limits on loan-to-value (LTV) ratios; caps on debt-service-to-income (DSTI) or loan-to-income (LTI) ratios.
  - Activation guidance: intensify supervisory scrutiny and communication first; tighten less distortionary sectoral capital requirements next; consider tighter LTV and/or DSTI limits if needed. LTV and DSTI caps should be imposed on the flow of new household loans (not retroactively).
  - Combine tools to reinforce effectiveness and mitigate single-tool shortcomings.
  - Expand the regulatory perimeter to include non-banks and foreign branches; inter-agency cooperation may be required.
  - Loosening guidance: sectoral tools can be loosened during housing busts to contain feedback loops, but relaxation should respect prudential minima and not go beyond a “permanent floor” considered safe.

### Extensions and linkage to housing loans analysis
- Recommended extension: include commercial real estate in the analysis and establish a commercial property price index in addition to a residential property price index to capture potential spillovers from anticipated oil production.
- Appendix 2 (Growth in Housing Loans) provides complementary regressions and diagnostics that indicate:
  - From 2014 to 2018, commercial banks’ housing loans grew by an average annual rate of 12.1 percent.
  - Share of commercial banks’ mortgage loans to total loans increased from 31.5 percent in 2014 to 34.9 percent in 2018.
  - Regression models for housing loan growth (including prime lending rate, construction growth, population growth, nominal GDP growth) show statistically significant coefficients for prime lending rate, construction growth, and population growth in several specifications, but nominal GDP growth is not statistically significant.
  - R-squared for housing-loan models: ranged from 0.48036 to 0.52889 across reported specifications.
  - Deviations between actual and predicted growth in housing loans in 2018: divergence of 7 percentage points, accounting for 0.38 standard deviations; hypothesis tests indicate deviations in 2016–2018 are not statistically significant; evidence is inconclusive on structural break.
  - Conclusion: analyses are early signals; authorities should strengthen monitoring and data collection to identify “missing variables” explaining divergences.

*Source: wpiea2020020-print-pdf - Appendix 1. Details of Regression: Growth in Housing Price (Georgetown) — wpiea2020020-print-pdf*

### Appendix 1. Details of Regression: Growth in Housing Price (Georgetown) ____________21

### Appendix 1. Details of Regression: Growth in Housing Price (Georgetown) ____________21

### III. Are Urban Real Estate Prices Showing Signs of "Overheating"?
- Objective:
  - Construct a baseline model for the growth rate in the Urban (Georgetown) Housing Prices and compare it (out-of-sample) with actual changes from 2016 to 2018 to assess whether historical predictors account for the post-2015 rise.
- Main conclusion:
  - Deviations between observed housing price growth and the predictor in 2016–2018 are not statistically significant; differences average around 0.5 standard deviation and hypothesis tests based on t-distribution suggest no significant deviations in 2016, 2017, and 2018 (Appendix 1).
- Caveats emphasized:
  - Absence of a real estate price index; the Urban (Georgetown) Housing Index (a sub-index of the Urban CPI that includes rent, fuel, and power) is used as a proxy.
  - Unavailability of housing market-related indicators: statistics on volume of houses and dwellings, volume of real estate loans, and consumer and business sentiment indices are not produced.

### Data and Context (selected statistics and trends)
- Residential real estate prices in Georgetown have risen following the discovery of oil in 2015.
- Commercial banks’ private dwellings mortgage lending:
  - Grew at an average annual rate of 7.7 percent from 2014-2018.
  - Share of private dwellings mortgage loans to total loans increased from 31.5 percent to 34.9 percent during 2014-2018.
- Household exposure to mortgage debt:
  - In Georgetown, exposure accounted for around 350 percent of income in mid-2019 (source: Numbeo).
  - Comparable average in other major cities in neighboring countries: around 190 percent of income (mid-2019, Numbeo).
- Commercial banks’ housing loans growth:
  - From 2014 to 2018, commercial banks’ housing loans grew by an average annual rate of 12.1 percent.
  - Commercial banks account for about 70 percent of financial system total assets and originate two thirds of total real estate mortgage loans (note in footnote).

### Regression specification (baseline predictor)
- Dependent variable:
  - Yi = Urban (Georgetown) housing prices (in percent change, Y/Y).
- Explanatory variables:
  - X1,i = Urban (Georgetown) CPI—All Items (in percent change, Y/Y).
  - X2,i = Oil Price (in percent change, Y/Y) (petroleum spot price (APSP), in US$/barrel).
  - X3,i = Population growth (in percent change, Y/Y).
  - X4,i = Share of urban population (in percent total population).
- Model form:
  - Yi = αi + β1 X1,i + β2 X2,i + β3 X3,i + β4 X4,i + εi.
- Data sources:
  - Urban (Georgetown) housing prices and Urban (Georgetown) CPI—All Items: Bank of Guyana (Statistical Bulletin).
  - Oil Prices: IMF World Economic Outlook (based on Petroleum spot price (APSP), in US$/barrel).
  - Population: United Nations (World Population Prospects 2019).
  - Share of urban population: The World Bank (World Development Indicators).

### Regression results (reported coefficients and diagnostics)
- Two model specifications:
  - Model 1(a): sample period 1995-2015 (used to predict out-of-sample 2016–2018).
  - Model 1(b): sample period 1995-2018 (full sample).
- Coefficient estimates (standard errors in brackets):
  - Urban CPI All Items:
    - Model 1(a): 0.5533** (0.2352)
    - Model 1(b): 0.58802** (0.20928)
  - Change in Oil Price:
    - Model 1(a): 0.04692* (0.02651)
    - Model 1(b): 0.0415* (0.02277)
  - Population Growth:
    - Model 1(a): 6.75491** (2.98852)
    - Model 1(b): 6.51903** (2.71445)
  - Share of Urban Population:
    - Model 1(a): 3.87884*** (0.94874)
    - Model 1(b): 3.87821*** (0.8784)
  - Constant:
    - Model 1(a): -106.4155*** (26.32431)
    - Model 1(b): -106.5723*** (24.36331)
- Goodness of fit and diagnostics:
  - R-squared:
    - Model 1(a): 0.74086
    - Model 1(b): 0.7589
  - Adjusted R-squared:
    - Model 1(a): 0.67608
    - Model 1(b): 0.7082
  - Durbin-Watson statistics:
    - Model 1(a): 2.5705
    - Model 1(b): 2.5199
  - Note: Coefficients are in expected signs; housing prices and the predictor are cointegrated and no significant autocorrelations are detected (Engle-Granger test for cointegration; Breusch-Godfrey serial correlation LM Test referenced in Appendix 1).
  - Significance notation: ***p<0.01; **p<0.05; *p<0.10.

### Interpretation of results
- The predictor explains a large share of variation in housing price growth (R-squared 0.74086 in Model 1(a); 0.7589 in Model 1(b)).
- Out-of-sample predictions for 2016–2018 closely track actual housing price growth; average deviation ~0.5 standard deviation.
- From 2017 to 2018 the pace of growth diverged: predictor showed deceleration while actual growth accelerated slightly—this could signal either a potential structural break or omitted variables.

### Policy implications and recommendations
- Continuous surveillance is important:
  - Step up monitoring of housing market developments, household debt, and bank lending practices given the potential for a structural break or missing variables.
  - Fortify the macroprudential policy toolkit to be ready to tighten measures when necessary to safeguard financial and macroeconomic stability.
- Data collection improvements recommended:
  - Expand data collection to include commercial real estate indicators, volume of houses and dwellings, volume of real estate loans, and consumer and business sentiment indices.
  - Increase frequency of data to improve forecasting accuracy and surveillance.

### Missing variables and suggested extensions to the study
- Variables that could explain the divergence and should be included in extended models:
  - Changes in property taxes (noting the 2018 National Budget reduced property tax rates from 0.75 percent to 0.5 percent for the first GY$ 20 million in taxable net property, with the remainder taxed at 0.75 percent).
  - Demand for real estate from foreigners and Guyanese residing abroad.
  - Changes in housing supply (construction and residential permits).
  - Distribution of income and household debt across different population groups.
  - Business and household sentiment indicators.

### IV. Have There Been Changes in Bank Mortgage Lending Policies? (introductory findings)
- Rationale:
  - A structural break between actual housing loan growth and a predictor could indicate shifts in banks’ credit policies (e.g., relaxation of underwriting standards) that could fuel higher real estate demand and prices.
- Observed lending trends:
  - From 2014 to 2018, commercial banks’ housing loans grew by an average annual rate of 12.1 percent, close to three times the average annual growth rate of total loans.
  - Share of commercial banks’ mortgage loans to total loans increased from 31.5 percent in 2014 to 34.9 percent in 2018.
- Model to explain housing loan growth (specified for subsequent analysis):
  - Dependent variable: Zj = Growth in commercial banks' housing loans (in percent, Y/Y).
  - Explanatory variables:
    - V1,j = Prime Lending Rate (in percent).
    - V2,j = Construction growth (in percent, Y/Y).
    - V3,j = Population growth (in percent, Y/Y).
    - V4,j = Nominal GDP growth (in percent change, Y/Y).
  - Data sources:
    - Commercial banks' housing loans and Prime Lending Rate: Bank of Guyana (Statistical Bulletin).
    - Construction sector: Guyana Bureau of Statistics.
    - Population: The World Bank (World Development Indicators).
    - Nominal GDP: IMF.

*Source: wpiea2020020-print-pdf - Appendix 1. Details of Regression: Growth in Housing Price (Georgetown) ____________21*

### 17. The predictor derived from the regression provides a reasonable in-sample fit.

### 17. The predictor derived from the regression provides a reasonable in-sample fit.

### Regression results and model specification
- Models used three explanatory variables: prime lending rate (proxy for the price of credit); growth in the construction sector (proxy for volume); and population growth (proxy for demand for housing).
- Model 3(a) and Model 3(b) include an additional variable—nominal GDP growth—as a proxy for income levels.
- The coefficients of the prime lending rate, construction growth, and population growth are reported as statistically significant in the models without and with nominal GDP growth.
- Nominal GDP growth is not statistically significant but its inclusion increases the in-sample fit (R-squared).

### Estimated coefficients (Figure 2 table summary)
- Sample period: 1995-2015 and 1995-2018 variants reported.
- Prime Lending Rate:
  - Model 2(a): 13.38224*** (3.96019)
  - Model 2(b): 13.43991*** (3.62313)
  - Model 3(a): 11.46512** (4.18723)
  - Model 3(b): 11.40515*** (3.81101)
- Construction growth:
  - Model 2(a): 1.01015* (0.55929)
  - Model 2(b): 1.00453* (0.51442)
  - Model 3(a): 1.07465* (0.55267)
  - Model 3(b): 1.0864** (0.50494)
- Population growth:
  - Model 2(a): 48.43238** (20.55105)
  - Model 2(b): 47.99398** (18.82824)
  - Model 3(a): 39.74414* (21.38052)
  - Model 3(b): 39.33508* (19.34107)
- Nominal GDP growth (Models 3):
  - Model 3(a): 1.22742 (0.98192)
  - Model 3(b): 1.25768 (0.88274)
- Constant terms:
  - Model 2(a): -196.6612*** (61.25398)
  - Model 2(b): -197.6339*** (55.89663)
  - Model 3(a): -177.8352** (62.11833)
  - Model 3(b): -177.2343*** (56.35975)
- Goodness of fit:
  - R-squared: 0.48036; 0.47856; 0.52659; 0.52889
  - Adjusted R-squared: 0.38866; 0.40034; 0.40824; 0.42971
  - Durbin-Watson statistics: 2.30322; 2.43972; 2.49322; 2.7175
- Significance notation preserved: ***p<0.01; **p<0.05; *p<0.10
- Note: Standard errors are in brackets.

### Statistical diagnostics
- No significant autocorrelation detected from the Breusch-Godfrey serial correlation LM Test.
- The Engle-Granger test for cointegration strongly shows that housing loan growth and the predictor are cointegrated (Appendix II referenced for details).

### Deviations between actual and predicted growth (2016–2018)
- Growth in housing loans was slightly faster than the predictor in 2018.
- The divergence between actual and predicted growth rates in 2018 is 7 percentage points.
- The 7 percentage point divergence accounts for 0.38 standard deviations.
- Hypothesis tests based on t-distribution suggest that deviations between observed growth in housing loans and the predictor are not statistically significant in 2016, 2017, and 2018, indicating inconclusive evidence of any structural break in the relationship between housing loans and the predictor (Appendix 2 referenced).

### Interpretation, data limitations, and recommended surveillance
- The regression R-squared and adjusted R-squared are not high, indicating the possibility of “missing variables” or that it is too early to tell if the sharp increase in housing loan in 2018 was due to changes in bank mortgage lending practices.
- Further analysis is needed by expanding the explanatory variables to improve the fit when more comprehensive data become available.
- Policy recommendation: It would be prudent to enhance supervisory surveillance on banks’ mortgage loans and underwriting standards to mitigate risks.
- While it is too early to conclude that banks’ mortgage lending practices have changed, stepping up surveillance is advised.

### Macroprudential context and guidance (selected points)
- The 2016 FSAP recommends that the Bank of Guyana (BoG) consider developing a formal macroprudential framework in the medium term; under existing legislation, the BoG has powers to use countercyclical tools, including loan-to-value (LTV) and debt-to-income (DTI) requirements.
- Potential longer-term tools: countercyclical capital buffers or systemic risk surcharges.
- Targeted sectoral macroprudential tools include:
  - Sectoral capital requirements (risk weights or loss given default (LGD) floors).
  - Limits on loan-to-value (LTV) ratios.
  - Caps on debt-service-to-income (DSTI) or loan-to-income (LTI) ratios.
- Core indicators to monitor: growth of mortgage loans and house prices; house price-to-rent and house price-to-income ratios; averages and distributions of LTV, DSTI, and LTI; share of foreign currency–denominated mortgage loans; housing price growth by region and property type.
- Activation guidance:
  - Sectoral tools should be activated or tightened when multiple indicators point to rising systemic risk.
  - Policymakers should take a gradual approach: intensify supervisory scrutiny and communication first; tighten less distortionary sectoral capital requirements next; consider tighter LTV and/or DSTI limits if needed.
  - LTV and DSTI caps should be imposed on the flow of new household loans (not retroactively on outstanding loans).
- Combining tools reinforces effectiveness and mitigates single-tool shortcomings (higher risk weights, LTV limits, DSTI/LTI caps work through complementary channels).
- Expand the regulatory perimeter to include non-banks and foreign branches to contain leakages; inter-agency cooperation may be required.
- When to loosen:
  - Sectoral tools can be loosened during housing busts to contain feedback loops between falls in credit and house prices.
  - Fast-moving indicators (house transaction volumes, spreads on household loans) can inform relaxation decisions, but evidence of systemic stress (simultaneous decline in prices and credit, increase in non-performing loans) is vital.
  - Relaxation should respect prudential minima and not go beyond a “permanent floor” level considered safe in downturns.
  - Relaxation may have limited effect if banks are risk-averse or capital-constrained (“pushing on a string”), but can still help contain spillbacks from falling prices and credit.

### Conclusions and policy implications
- Housing prices in Georgetown and commercial banks’ lending to the housing sector in Guyana appear to be in their early stages of growth.
- No solid evidence yet supports structural changes in dynamics determining housing price growth or commercial banks’ housing loan practices during 2016 to 2018.
- The analyses should be treated as early signals; authorities should strengthen monitoring of the housing market, household debt, and bank lending practices.
- Fortify the macroprudential policy framework with effective toolkits the central bank could deploy early when necessary.
- Intrusive and tighter supervision, including enforcement actions, would be needed if lax underwriting standards are detected.
- Identify “missing variables” explaining divergences between actual and model-predicted growth rates by collecting new data and increasing data frequency to enhance surveillance and forecast accuracy.
- Extension recommended: include commercial real estate in the analysis and establish a commercial property price index in addition to a residential property price index to capture potential spillovers from anticipated oil production.

*Source: wpiea2020020-print-pdf - 17. The predictor derived from the regression provides a reasonable in-sample fit.*

### APPENDIX 1. DETAILS OF REGRESSION: GROWTH IN HOUSING PRICE (GEORGETOWN)

### APPENDIX 2. DETAILS OF REGRESSION: GROWTH IN HOUSING LOANS

### (a) Partial Sample (1995-2015) — Regression Results and Diagnostic Tests
- Conclusion statements:
  - Based on the Breusch-Godfrey Serial Correlation LM Test, the null hypothesis (no serial correlation) is accepted at 5 percent significance level.
  - Based on the Engel-Granger test of cointegration, the null hypothesis (actual housing loans growth and the predictor are not cointegrated) is rejected, at 5 percent significance level.

- Dependent Variable: Growth in Housing Loans
- Method: Least Squares
- Sample : 1995 2015

- Coefficients and inference:
  - Constant: -177.835262.11833; t-Statistic: -2.862846; Prob.: 0.0113
  - Prime Lending Rate: 11.465124.1872332.738113; Prob.: 0.0146
  - Construction Growth: 1.0746460.552671; t-Statistic: 1.944460; Prob.: 0.0696
  - Population Growth: 39.7441421.380521; t-Statistic: 1.858895; Prob.: 0.0815
  - Nominal GDP Growth: 1.2274170.981919; t-Statistic: 1.250019; Prob.: 0.2293

- Model fit and statistics:
  - R-squared: 0.526591    Mean dependent var: 15.4978
  - Adjusted R-squared: 0.408239    S.D. dependent var: 28.1575
  - S.E. of regression: 21.66043    Akaike info criterion: 9.193108
  - Sum squared resid: 7506.789    Schwarz criterion: 9.441804
  - Log likelihood: -91.52764    Hannan-Quinn criter.: 9.247082
  - F-statistic: 4.449361    Durbin-Watson stat: 2.493197
  - Prob(F-statistic): 0.013151

- Breusch-Godfrey Serial Correlation LM Test (Null: No serial correlation at up to 2 lags):
  - F-statistic: 1.741037    Prob. F( 2,14): 0.2112
  - Obs*R-squared: 4.182774    Prob. Chi-Square(2): 0.1235

- Engel-Granger Cointegration Test (Series: ACTUAL FITTED; Sample : 1995 2015; Included observations: 21; Cointegrating equation deterministics: C; Automatic lags specification based on Schwarz criterion (maxlag=4)):
  - ACTUAL: tau-statistic: -6.076332; Prob.*: 0.0004; z-statistic: -28.553250; Prob.*: 0.0001
  - FITTED: tau-statistic: -4.735105; Prob.*: 0.0063; z-statistic: -22.730230; Prob.*: 0.0023

### (b) Full Sample (1995-2018) — Regression Results and Diagnostic Tests
- Conclusion statements:
  - Based on the Breusch-Godfrey Serial Correlation LM Test, the null hypothesis (no serial correlation) is accepted at 5 percent significance level.
  - Based on the Engel-Granger test of cointegration, the null hypothesis (actual housing loans growth and the predictor are not cointegrated) is rejected, at 5 percent significance level.

- Dependent Variable: Growth in Housing Loans
- Method: Least Squares
- Sample : 1995 2018

- Coefficients and inference:
  - Constant: -177.234356.35975; t-Statistic: -3.144697; Prob.: 0.0053
  - Prime Lending Rate: 11.405153.8110132.992681; Prob.: 0.0075
  - Construction Growth: 1.0864040.504944; t-Statistic: 2.151534; Prob.: 0.0445
  - Population Growth: 39.3350819.341072; t-Statistic: 2.033759; Prob.: 0.0562
  - Nominal GDP Growth: 1.2576810.882736; t-Statistic: 1.424753; Prob.: 0.1704

- Model fit and statistics:
  - R-squared: 0.528888    Mean dependent var: 14.77082
  - Adjusted R-squared: 0.429707    S.D. dependent var: 26.45349
  - S.E. of regression: 19.97708    Akaike info criterion: 9.0101
  - Sum squared resid: 7582.59    Schwarz criterion: 9.255528
  - Log likelihood: -103.1212    Hannan-Quinn criter.: 9.075212
  - F-statistic: 5.332538    Durbin-Watson stat: 2.717526
  - Prob(F-statistic): 0.004727

- Breusch-Godfrey Serial Correlation LM Test (Null: No serial correlation at up to 2 lags):
  - F-statistic: 1.565836    Prob. F( 2,17): 0.2376
  - Obs*R-squared: 3.733427    Prob. Chi-Square(2): 0.1546

- Engel-Granger Cointegration Test (Series: ACTUAL FITTED; Sample : 1995 2018; Included observations: 24; Cointegrating equation deterministics: C; Automatic lags specification based on Schwarz criterion (maxlag=4)):
  - ACTUAL: tau-statistic: -6.8732030; z-statistic: -31.404060
  - FITTED: tau-statistic: -5.8372170.0005; z-statistic: -25.079940.0013

### (c) Hypothesis Test: Are Observed Growths in Commercial Banks’ Housing loans Significantly Different from the Predictor in 2016, 2017, and 2018?
- Hypothesis:
  - H0: (A)-(B) = 0
  - H1: (A)-(B) ≠ 0
- Mean = -0.02; Standard deviation = 18.1578; N = 24

- 2016:
  - Test statistic, Zi = -0.209
  - Critical value, tcrit,i = -1.711 (t-distribution, 5 percent significance level, n=24)
  - Result: Accept H0 since Zi > tcrit,i. Therefore, the observed growth in housing loans is not significantly different from the predictor in 2016.

- 2017:
  - Test statistic, Zj = -0.199
  - Critical value, tcrit,j = -1.711 (t-distribution, 5 percent significance level, n=24)
  - Result: Accept H0 since Zi > tcrit,j. Therefore, the observed growth in housing loans is not significantly different from the predictor in 2017.

- 2018:
  - Test statistic, Zk = -0.385
  - Critical value, tcrit,k = -1.711 (t-distribution, 5 percent significance level, n=24)
  - Result: Accept H0 since Zk > tcrit,k. Therefore, the observed growth in housing loan is not significantly different from the predictor in 2018.

- Annual Actual, Predictor, and Difference (A)-(B) for Growth in Housing Loans:
  - 1995: Actual (A) 85.53; Predictor (B) 91.86; Difference (A)-(B) 6.25
  - 1996: 46.52; 54.47; 7.84
  - 1997: 25.09; 16.2; -8.90
  - 1998: -3.24; 19.22; 22.43
  - 1999: 6.30; 13.36; 6.97
  - 2000: 16.25; 24.48; 8.16
  - 2001: 12.26; -23.4; -35.64
  - 2002: 7.95; 21.21; 13.23
  - 2003: 6.32; -5.3; -11.58
  - 2004: 5.40; -28.1; -33.53
  - 2005: 8.22; 14.76; 6.52
  - 2006: 16.62; 24.98; 8.32
  - 2007: 24.76; 12.9; -11.90
  - 2008: 6.80; 29.72; 2.89
  - 2009: 0.98; -29.9; -30.87
  - 2010: 23.38; 33.4; 10.06
  - 2011: 16.93; 22.55; 5.55
  - 2012: 0.42; -24.5; -24.90
  - 2013: 20.05; 26.86; 6.77
  - 2014: 14.41; 18.7; -5.67
  - 2015: -15.50; 22.53; 38.00
  - 2016: 10.17; 6.3; -3.82
  - 2017: 6.84; 3.2; -3.63
  - 2018: 12.52; 19.56; 6.97

*Source: APPENDIX 2. DETAILS OF REGRESSION: GROWTH IN HOUSING LOANS — wpiea2020020-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020020-print-pdf.pdf_