Selected Issues and Analytical Notes

Abstract

Selected Issues and Analytical Notes

Macroprudential Tools at Work in Canada1

1. The aim of this paper is to assess which macroprudential policy measures have been effective in containing house price and mortgage credit growth in Canada and other economies. Our analysis indicates that macroprudential policy measures have had a moderating effect on house prices and mortgage credit in Canada since 2010. International experience suggests that lower caps on debt-service-to-income (DSTI) ratios and loan-to-value ratios could be effective in containing both mortgage credit and house price growth.

2. The discussion below is organized as follows. First, we analyze the effectiveness of macroprudential measures in Canada, then we discuss lessons from the international experience with macroprudential measures. Finally, we offer concluding remarks.

A. Assessing the Effectiveness of Macroprudential Measures in Canada

3. A distinctive feature of Canada’s prudential toolkit stems from a large government role in mortgage insurance. More than a half of outstanding mortgage debt is covered by mortgage insurance due to a requirement that borrowers must obtain insurance for high loan-to-value (LTV) mortgages (with LTV over 80 percent). 2 Government-backed mortgage insurance is provided by the Canada Mortgage and Housing Corporation (CMHC) - a crown corporation with half the market share of new insured mortgages3 - and two private companies. The government guarantees 100 percent of CMHC’s obligations and backs private mortgage insurers’ obligations subject to a deductible equal to 10 percent of the original principal amount of the mortgage loan. To address risks associated with the provision of these guarantees, the government sets eligibility requirements for insured mortgages. This serves as an important macroprudential tool in addition to traditional macroprudential tools available to the regulators responsible for financial stability.

4. Since 2000s Canada’s policies managing housing market risks have gone through two contrasting stages:

  • Easing up to 2007. Mortgage insurance rules were relaxed in the mid-2000s. Together with lower interest rates, this boosted mortgage credit and housing prices. The higher house prices led to a sharp expansion of home equity credit lines.

  • Tightening since 2008. As house prices and mortgage credit surged, the government’s focus changed to containing the growth of imbalances in the housing market as well as to reduce its exposure to mortgage insurance. Since 2008, the federal government has undertaken multiple rounds of measures to tighten mortgage insurance rules, going beyond a reversal of the loosening in the mid-2000s (Table A1). In addition, OSFI introduced numerous microprudential measures to strengthen bank balance sheets (Table A2).

5. Our analysis focuses on macroprudential policy measures related to changes in mortgage insurance rules during the tightening stage. Here we distinguish six rounds of macroprudential measures introduced in 2008, 2010, 2011, 2012, early-2016 and late-2016 (Table 1). We did not include the increases in mortgage insurance premia in 2014 and 2015 into our empirical exercise as the changes were too small to have an impact on credit demand.4

Table 1.

Canada: Macroprudential Measures Related to Changes in Mortgage Insurance Rules Since 2008

article image
article image
Sources: Department of Finance Canada, CMHC, OSFI

Methodology

6. We estimate two separate equations, for mortgage credit and house prices, assuming that macroprudential measures affect house prices indirectly through the mortgage credit. Each equation is estimated separately using OLS.5 All variables are monthly in a sample from August 1998 to February 2017.

7. For mortgage credit growth, we estimate the following specification:

Mt=αm+βmXt+γmDti+εmt

where Mt is mortgage credit year-on-year growth rate;

Xt is a matrix of control variables including the unemployment rate and hourly wage growth, five-year mortgage interest rate,6 and house price growth. The control variables enter the equations with lags to account for delays in the response of mortgage credit to changes in the explanatory variables.

Dti stands for a dummy variable equal to 1 in the months following the implementation of a set of macroprudential measures i (2008, 2010, 2011, 2012, and two 2016 rounds) and zero otherwise. To isolate the effects of individual rounds of measures, each dummy variable takes a value of 1 until the end of the sample. We assess the impact of the rounds of measures on mortgage credit using the entire sample, but also test the impact over 1, 3, 6, and 9 months after they were introduced, and for the whole period between rounds.7

Separately, we test the effectiveness of individual measures and replace the dummy variable in specification (1) with changes in a specific instrument, such as the maximum LTV ratio for new and refinancing loans and amortization period.

8. For house price growth, we estimate the following specification:

Ht=αh+βhYt+εht

where Ht is house price year-on-year growth rate;

Yt is a matrix of lagged control variables including the growth rate of the number of completed but unsold houses, mortgage credit growth, nominal GDP growth, and growth of sales of existing houses.8 The control variables enter the equations with lags to account for the delay in the house price response to changes in their determinants.

Results

9. The estimated specification (1) suggests the following effectiveness of the individual rounds of macroprudential measures (Table A3):

  • 2008 round does not appear to have had an impact on mortgage credit growth. The estimated coefficient for the 2008 measures is not statistically significant across the different specifications. The lack of effects could be related to the limited scope of the measures: the maximum amortization period for new government backed insured mortgages was lowered from 40 to 35 years but remained high; and even though the maximum LTV for new mortgages was reduced from 100 to 95 percent, the effective LTV ratio was still at 100 percent.9 Credit growth did decelerate in the 12 months following the measures, but this was likely due to the cyclical downturn in the economy accompanied by higher unemployment and a decline in household income.

  • 2010 round had a statistically significant impact on mortgage credit growth, dampening it by about 2.3 percentage points on average. The package of measures included a lower LTV ratio on refinancing loans, a significant increase in the down payment on properties not occupied by owners (from 5 to 20 percent), and tighter eligibility criteria that prevented some potential borrowers from qualifying for mortgages.

  • 2011 round also appears to have reduced mortgage credit growth, by about 0.9 percentage points on average. The implemented measures further lowered the LTV ratio on refinancing loans and brought the maximum amortization period closer to the average. This likely prevented more borrowers from taking new loans and reduced the size of the loans.10

  • 2012 round seems to have been the most effective one as it trimmed down mortgage credit growth by an estimated 3 percentage points on average. The effect appears to have gotten stronger over time. The measures focused on the amortization period, the LTV ratio for mortgage refinancing and DTIs. The new LTV ratio on refinance loans (down to 80 percent) was likely quite effective, as more than half of the new insured refinance loans in the period before the 2012 measures had a LTV ratio higher than 85 percent. These measures reduced the effective LTV for first-time home buyers from 100 to 95 percent.11 The effects of the 2012 measures might have also captured new mortgage underwriting standards implemented by OSFI at the end of fiscal year 2012.

  • Early 2016 round did not seem to have an impact on mortgage credit growth, possibly because it affected a relatively small proportion of the mortgage market. The 2016 tightening of mortgage insurance rules included increasing the minimum down payment for new insured mortgages from 5 to 10 percent but only for the portion of the house price above $500,000.12

  • Late 2016 round seems to have been effective in reducing mortgage credit growth. Estimates suggest that more stringent qualifying rules had an immediate effect on mortgage credit growth, as they tightened underwriting standards and affected all new borrowers with a down payment below 20 percent. These results, however, are based on few observations and should be interpreted with caution.

10. The results for individual measures suggest similar effectiveness of tighter LTVs for new mortgages and refinancing loans. The estimates indicate that a one percentage point reduction in both the maximum LTV for new mortgages and for refinancing loans tends to reduce y/y credit growth by about 0.5 percentage points (Table A4). Reducing the amortization period by one year appears to have lowered credit growth by 0.2 percentage points.13

11. Estimates suggest that there is a strong link between mortgage credit and house price growth. The estimated equation for house price growth indicates that other things being equal mortgage credit growth has a one-for-one effect on house price growth (Table A5). This means that tighter mortgage insurance rules indirectly dampened house price growth by reducing mortgage credit growth.

12. The household debt to income ratio would have likely been even higher had the authorities not acted. We calculate the fitted regression values of mortgage growth rates both with the measures and without them. Other things being equal, without these measures the average mortgage credit growth since April 2010 would have been higher by more than 5 percentage points. The household debt-to-income ratio would have been closer to 200 percent as of the third quarter of 2016, instead of the actual 167 percent.

B. International Experience

13. Economies around the world have used a variety of policy tools to deal with house price and mortgage credit booms. These include traditional monetary and fiscal policies, including transaction tax, property tax, sellers’ and buyers’ duty (Crowe et al., 2011 and Dell’Ariccia et al., 2012) and other macroprudential measures (see Appendix in Lim and others (2013)). Caps on LTV, and the DSTI ratio, provisioning requirements and risk weights are among the most frequently used tools.14

14. We update the macroprudential dataset in Lim and al. (2013) and examine the international experience over 2000–16 with four most frequently used measures: limits on LTV ratios; caps on DSTI ratios; greater risk weights for banks’ credit assets; and higher provisioning requirements for banks. We estimate their impact on mortgage credit and house price growth with panel data regressions across a sample of 35 economies controlling for the business cycle and the cost of borrowing.15

Economies in the Dataset

article image

For mortgage credit growth, we estimate the following specification:

Mit=α1i+α2Xit+α3Dit+εit

where M is real mortgage credit year-on-year growth;

X is a matrix of variables including lagged real mortgage credit growth, real GDP growth, and real lending rates;

D stands for an MPP measure. Following Krznar and Morsink (2014), we use a “step function variable” for each macroprudential instrument, that is, a variable that increases (decreases) by one every time the instrument is tightened (loosened) and stays there until the instrument is changed. For house price growth, we estimate the following specification:

Hit=β1i+β2Yit+β3Dit+εit

where H is real house price year-on-year growth;

Y is a matrix of variables including lagged real house price growth, real GDP growth, and real lending rates;

D stands for an MPP measure, constructed as a “step function variable”.

We use GMM Arrellano-Bond estimator, which is designed for the situations with fixed individual effects and explanatory variables that are not strictly exogenous.16

15. The results suggest that all four measures can be effective in containing mortgage credit growth (Table A6). The economies that lowered the caps on the LTV and DSTI ratios, tightened provisioning requirements or raised risk weights for banks’ credit assets moderated the mortgage credit growth rate on average by 0.9-2 percentage points.

16. Lower caps on LTV and DSTI ratios appear to have a statistically significant but economically small effect on house price growth (Table A7). Both measures contained the house price growth by about 0.4 percentage points.

C. Conclusion

17. Canada’s macroprudential policy seems to have lowered mortgage credit growth thus moderating the surge in house prices on a national basis. The 2010, 2011, 2012 and late 2016 rounds of measures are estimated to have been effective, possibly due to their larger scope and size.

18. International experience over 2000-16 suggests that tighter LTV limits and lower caps on DSTI ratios could be effective in moderating both mortgage credit and house price growth. Canada’s DSTI limits are relatively stringent compared to the other economies, but the limits on the LTV ratio are above the average in our sample of economies. Further tightening of the LTV ratios could help contain growth in mortgage credit and house prices.

A04ufig1

Debt Service-To-Income Limits, End-2016

(Percent)

Citation: IMF Staff Country Reports 2017, 211; 10.5089/9781484309650.002.A004

Sources: Authorities and the IMF Database of Macroprudentail Measures.
A04ufig2

Loan-To-Value Limits, End-2016

(Percent)

Citation: IMF Staff Country Reports 2017, 211; 10.5089/9781484309650.002.A004

Sources: Authorities and the IMF Database of Macroprudentail Measures.

Annex I. Annex Tables

Table A1.

Canada: Mortgage Insurance Products Until 2008

article image
Source: CMHC, Genworth.
Table A2.

Canada: Microprudential Measures

article image
article image
Table A3.

Canada: Mortgage Credit Equation

article image

*,**,*** indicate respectively statistical significance at the 10, 5, and 1 percent level. Standard deviations in italic.

The estimation period is 1999:2-2017:2, using montly, seasonally adjusted data. Newey-West consistent variance estimator is used to calculate coefficients’ standard deviation.

Regressions I to IV estimate macroprudential measures effects after 1, 3, 6 and 9 months respectively after their implementation. Regression V estimates effects of each macroprudential measure between rounds of measures.

Table A4.

Canada: Effects of Specific Macroprudential Measures on Mortgage Growth, OLS Estimation

article image

*,**,*** indicate respectively statistical significance at the 10, 5, and 1 percent level.

The estimation period is 1999:2-2017:2, using monthly seasonally adjusted data. Newey-West consistent variance estimator is used to calculate coefficients’ standard deviation. Standard deviations in italic.

All regressions include control variables as in Table A3 but are not shown here.

Table A5.

Canada: House Price Equation

article image

*,**,*** indicate respectively statistical significance at the 10, 5, and 1 percent level. Standard deviations in italic.

OLS estimation, period of 1999:2-2017:2. Monthly, seasonally adjusted data are used.

Newey-West consistent variance estimator is used to calculate coefficients’ standard deviation.

The dependent variable is the y-o-y change in house price index(source: CREA).

Table A6.

Effects of Macroprudential Measures on Mortgage Credit Growth-Panel GMM Estimation (2000-16)

Dependent variable: Mortgage credit growth (real)

article image

*,**,*** indicate respectively statistical significance at the 10, 5, and 1 percent level. Standard deviations are in italic.

The estimation period is 2000:1-2016:4; quarterly, seasonally adjusted data.

Data for China, India, Colombia, Romania, Lithuania, Luxembourg and Uruguay pertain to claims to private sector since mortgage credit data is unavailable.

Real mortgage credit growth is defined as the y-o-y growth rate of nominal mortgage credit deflated by CPI.

Table A7.

Effects of Macroprudential Measures on House Price Growth-Panel GMM Estimation (2000-16)

Dependent variable: House price growth (real)

article image

*,**,*** indicate respectively statistical significance at the 10, 5, and 1 percent level. Standard deviations in italic.

The estimation period is 2000:1-2016:4; quarterly, seasonally adjusted data.

Real house prices are defined as house price indices deflated by CPI (Source: OECD, GlobalProperty Guide, IMF dataset).

References

  • Allan C. and U. Faruqui, 2011, “What Explains Trends in Household Debt in Canada?Bank of Canada Review, Winter 2011–12.

  • Allen, J., 2011, “Competition in the Canadian Mortgage Market,Bank of Canada Review, Winter 2010–11.

  • Allen, J., Grieder T., Peterson B., and T. Roberts, 2017, “The Impact of Macroprudential Housing Finance Tools in Canada,Working Papers No 632 (Basel: Bank for International Settlements.

    • Search Google Scholar
    • Export Citation
  • Ahuja, A. and M. Nabar, 2011, “Safeguarding Banks and Containing Property Booms: Cross-Country Evidence on Macroprudential Policies and Lesson from Hong Kong SAR,IMF Working Paper, 11/284 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Akinci, O. and J. Olmstead-Rumsey, 2015, “How Effective are Macroprudential Policies? An Empirical Investigation,International Finance Discussion Papers 1136, Board of Governors of the Federal Reserve System.

    • Search Google Scholar
    • Export Citation
  • Arregui, N., Benes, J., Krznar, I., Mitra, S. and A. Oliveira Santos, 2013, “Evaluating the Net Benefits of Macroprudential Policy: A Cookbook,IMF Working Paper 13/167 (Washington: International Monetary Fund).

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Almeida, H., Campello, M., and C. Liu, 2006, “The Financial Accelerator: Evidence from International Housing Markets,Review of Finance, (10) 132 2006.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Canada Mortgage and Housing Corporation, 2011, “Annual Report: A Solid Foundation for Generations”.

  • Canada Mortgage and Housing Corporation, 2012, “Canadian Housing Observer”.

  • Canada Mortgage and Housing Corporation, 2016, “Forests and Trees: Housing Finance and Macroprudential Policy in Canada”, at https://www.cmhc-schl.gc.ca/en/corp/nero/sp/2016/2016-11-18-0800.cfm

    • Search Google Scholar
    • Export Citation
  • Crawford, A. and U. Faruqui, 2012, “What Explains Trends in Household Debt in Canada?

  • Bank of Canada Review, Winter 2011–12.

  • Crowe, C., G. Dell’Ariccia, D. Igan, and P. Rabanal, 2011, “Policies for Macrofinancial Stability: Options to Deal with Real Estate Booms,IMF Staff Discussion Note, SDN/11/02.

    • Search Google Scholar
    • Export Citation
  • Dell’Ariccia, G., Igan, D., Laeven, L., and H. Tong, 2012, “Policies for Macrofinancial Stability: Options to Deal with Credit Booms,IMF Staff Discussion Note SDN/12/06.

    • Search Google Scholar
    • Export Citation
  • Department of Finance Canada (2016), “Balancing the Distribution of Risk in Canada’s Housing Finance System: A Consultation Document on Lender Risk Sharing for Government-Backed Insured Mortgages”, available at https://www.fin.gc.ca/activty/consult/lrs-prp-eng.asp

    • Search Google Scholar
    • Export Citation
  • Dunning, W., 2009, “The Canadian Residential Mortgage Market during Challenging Times”, Canadian Association of Accredited Mortgage Professionals.

    • Search Google Scholar
    • Export Citation
  • Dunning, W., 2011, “Revisiting the Canadian Mortgage Market - The Risk is Minimal”, Canadian Association of Accredited Mortgage Professionals.

    • Search Google Scholar
    • Export Citation
  • Dunning, W., 2012, “Annual State of the Residential Mortgage Market in Canada”, Canadian Association of Accredited Mortgage Professionals.

    • Search Google Scholar
    • Export Citation
  • Dunning, W., 2015, “Annual State of the Residential Mortgage Market in Canada,Canadian Association of Accredited Mortgage Professionals.

    • Search Google Scholar
    • Export Citation
  • Gravelle, T., Grieder, T., and S. Lavoie, 2013, “Monitoring and Assessing Risks in Canada’s Shadow Banking Sector,Bank of Canada Financial System Review, June 2013.

    • Search Google Scholar
    • Export Citation
  • Hallissey, N., Kelly, R. and T. O’Malley, 2014, “Macro-prudential Tools and Credit Risk of Property Lending at Irish banks,Central Bank of Ireland, Economic Letter Series, Vol. 2014, No. 10.

    • Search Google Scholar
    • Export Citation
  • Igan, D. and H. Kang, 2011, “Do Loan-to-Value and Debt-to-Income Limits Work? Evidence from Korea,IMF Working Paper No. 11/297.

  • IMF, 2011, “Housing Finance and Financial Stability—Back to Basics?Chapter 3 in Global Financial Stability Report, September (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2013, “The Interaction of Monetary and Macroprudential Policies—Background Paper,” (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • IMF, 2014a, Canada—2013 Article IV Consultation, February (Washington: International Monetary Fund).

  • IMF, 2014b, Canada—2014 Financial Sector Stability Assessment, February (Washington: International Monetary Fund).

  • IMF, 2014c, Canada—Financial Sector Assessment Program-Stress Testing-Technical Note, March (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Joint Forum, 2013, “Mortgage Insurance: Market Structure, Underwriting Cycle, and Policy Implications—Consultative Document,Basel Committee on Banking Supervision, Joint Forum, February.

    • Search Google Scholar
    • Export Citation
  • Kiff, J., 2009, “Canadian Residential Mortgage Markets: Boring but Effective?IMF Working Paper No. 09/130.

  • Krznar, I. and J. Morsink 2014, “With Great Power Comes Great Responsibility: Macroprudential Tools at Work in Canada,IMF Working Paper No. 14/83.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kuttner, K.N. and I. Shim, 2014, “Can non-interest rate policies stabilise housing markets? Evidence from a panel of 57 economies,Working Papers No 433, Bank for International Settlements.

    • Search Google Scholar
    • Export Citation
  • Lim, C.H., Krznar, I., Lipinsky, F., Otani, A., and X. Wu, 2013, “The Macroprudential Framework: Policy Responsiveness and Institutional Arrangements,IMF Working Paper No. 13/166 (Washington: International Monetary Fund).

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lim, C., Columba, F., Costa, A., Kongsamut, P., Otani, A., Saiyid, M., Wezel, T., and X. Wu, 2011, “Macroprudential Policy: What Instruments and How to Use Them? Lessons from Country Experiences,” IMF Working Paper No. 11/238 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Office of the Superintendent of Financial Institutions, 2016, “OSFI Tightens Supervisory Expectations for Mortgage Underwriting,” available at http://www.osfi-bsif.gc.ca/Eng/osfi-bsif/med/Pages/nr20160707.aspx

    • Search Google Scholar
    • Export Citation
  • Peterson, B. and Y. Zheng, 2011, “Medium-Term Fluctuations in Canadian House Prices,Bank of Canada Review, Winter 2011-2012

  • Vandenbussche, J., Vogel, U., and E. Detragiache, 2012, “Macroprudential Policies and Housing Prices—A New Database and Empirical Evidence for Central, Eastern, and Southeastern Europe,IMF Working Paper 12/303 (Washington: International Monetary Fund).

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wong, E., Fong, T., Li, K., and H. Choi, 2011, “Loan-to-Value Ratio as a Macroprudential Tool: Hong Kong’s Experience and Cross-Country Evidence,Hong Kong Monetary Authority Working Paper No. 01/2011.

    • Search Google Scholar
    • Export Citation
1

Prepared by Ivo Krznar, Zsofia Arvai (both MCM), and Yulia Ustyugova (WHD).

2

In Ontario, provincially-regulated credit unions are also required to have mortgage insurance in cases where the LTV exceeds 80 percent. Mortgages with LTV ratios of 80 percent or below (“low LTV” or low-ratio mortgages) may also be insured on a portfolio basis.

3

CMHC captures close to 65 percent of outstanding insured mortgage debt.

4

In 2014, CMHC increased its mortgage loan insurance premiums by approximately 15 percent for homeowners and 1-4 unit rental properties which raised monthly mortgage payment by approximately $5. In 2015, mortgage loan insurance premiums for homebuyers with less than a 10% down payment increased by approximately 15%, resulting in additional increase of approximately $5 to their monthly mortgage payment.

5

The results of the model estimated using SUR or 3SLS estimators are consistent with the results of the model estimated using OLS estimator.

6

Source: Canada Mortgage and Housing Corporation, conventional mortgage lending rate, five year term.

7

Given that the annual mortgage credit growth rate is regressed on monthly data, the estimated effect of macroprudential measures could be muted on impact and increase overtime,

8

This follows Peterson and Zheng (2011). Igan and Kang (2011) also use similar specifications for Korea.

9

The government set a minimum down payment of 5 percent for insured loans, but “cash backs,” unsecured borrowing and gifts could have been considered part of the down payment.

10

CMHC (2011) suggested that the volume of refinance loans dropped by 22 percent following the 2011 measures. Dunning (2012) estimated that the 2011 measures would push debt-service ratios above the maximum limit for about 6 percent of the high LTV mortgages taken out during 2010. He also suggested that about 11 percent of the borrowers in 2011 would have not been able to access credit following the latest reduction of the maximum amortization period.

11

OSFI’s B-20 guideline stipulates that banks should make reasonable efforts to determine if down payment is sourced from the borrower’s own resources or savings. CMHC (2012a) indicates that 35 percent of households who purchased a house in 2011 were first-time borrowers and about 15 percent of them borrowed at least part of the down payment.

12

Dunning (2015) suggests that only a small minority of home buyers (2 percent) and mortgage borrowers were affected and for most of them the additional required down payments were relatively insubstantial.

13

Using loan-level administrative data and household survey data for 2005-10, Allen et al. (2017) find that policies targeting the loan-to-value ratio have a larger impact on demand than policies targeting the debt-service ratio, such as amortization.

14

See Lim et all (2011), Crowe et al. (2011), Akinci and Olmstead-Rumsey (2015), Almeida, Campello and Liu (2005), Wong et al. (2011), Ahuja and Nabar (2011), Igan and Kang (2011), IMF (2011), Arregui and others (2013), Kuttner and Shim (2014), Hallissey and others (2014), Vandenbussche, Vogel, and Detragiache (2012) for more detailed discussion on implementation and evidence of instruments’ effectiveness.

15

The regressions broadly follow the approach in Arregui and others (2013).

16

Given the challenges associated with finding valid instruments, we also estimated the same specifications using fixed effects models that produced generally consistent results (except for the coefficient for provisioning that lost significance in the mortgage credit growth equation).

Canada: Selected Issues and Analytical Notes
Author: International Monetary Fund. Western Hemisphere Dept.