Macroprudential policy in a recovering property market: too much too soon?

Abstract

The aftermath of the 2007/2008 financial crisis has resulted in many central banks and regulatory authorities examining the effectiveness of macroprudential policy in preventing the emergence of future credit bubbles. Specific limits on loan-to-value (LTV) and loan-to-income (LTI) ratios have been introduced in a number of economies as a means of ensuring greater financial stability. The Irish property and credit market were particularly affected in the crisis as the domestic housing market had, since 1995, experienced sustained price and housing supply increases. Much of the activity in the Irish market was fuelled by a credit bubble following the emergence of international wholesale funding post 2003. After a period of pronounced declines, Irish house prices in late 2013 started to increase significantly; in early 2015, in response, the Irish Central Bank imposed new LTV and LTI limits to curb house price inflation. However, the introduction of these measures comes at a time when housing supply and mortgage lending are still at historically low levels. Therefore, in this paper we use a newly developed structural model of the Irish property and credit market to examine the implications of these measures for house prices and key activity variables in the mortgage market.

Keywords:

• macroprudential
• house prices
• mortgage credit

Acknowledgements

The authors are economists in the Economic Analysis division of the Economic and Social Research Institute. The views expressed in the paper are not those necessarily held by the ESRI. The authors thank the editor, three anonymous referees and participants at the conference on “Macropudential Regulation: Policy Dynamics and Limiations” held in Dublin in January 2016 for helpful comments. All errors and omissions are the responsibility of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. This is exceptional in euro area terms where the average is about one in twelve.

2. Consequently, one of the major cornerstones of the 2010 program of support agreed between Ireland, the EU and the IMF was a significant commitment to deal with the degree of loan impairment on the mortgage books of Irish financial institutions.

3. In particular, for non-first-time buyers a limit, for primary dwelling purchase, of 80% LTV will now exist, while lending for primary dwelling purchases above 3.5 times LTI is now restricted to no more than 20% of that aggregate value. The regulations are somewhat more lenient for first-time buyers.

4. Fitzpatrick and McQuinn (2007) found evidence of a mutually reinforcing link between mortgage credit and house prices in the Irish market over the period 1980–2004.

5. See Duffy and McQuinn (2014) for example.

6. See Honohan (2010) for more on this.

7. An exact chronology of the control and subsequent liberalisation of the Irish credit market is discussed in detail in Kelly and Everett (2004). See, in particular, Box 1 pp. 96 and 97, which illustrate the building and dismantling of control over the period 1973–1999.

8. Mortgage value as a percentage of property value at the inception of the mortgage contract.

9. Mortgages where the interest charged equals the main ECB refinancing rate plus a premium set at inception.

10. The main model used by McQuinn (2014) is that also applied in Kelly and McQuinn (2014) and is the standard inverted housing demand function commonly applied in such studies as Cameron, Muellbauer, and Murphy (2006); Muellbauer and Murphy (1997); Muellbauer and Murphy (1994); Meen (1996, 2000) and Peek and Wilcox (1991).

11. For a general overview of macroprudential policy, see Galati and Moessner (2013) and Lim et al. (2011).

12. This work is mainly motivated by the call from the G-20 regulatory reform agenda in February 2011 on the IMF, the Financial Stability Board (FSB) and the Bank for International Settlements (BIS) to develop a macroprudential policy framework.

13. In this way, our specification of credit demand is similar to that in Avouyi-Dovi et al. (2014) who estimate a highly parsimonious model of credit and housing markets in France. However, in their study, credit demand is a function of a composite indicator of lending criteria that includes the debt-service-to-income ratio and mortgage duration.

14. See McCarthy and McQuinn (2013), Fernandez-Corugedo and Muellbauer (2006) and Duca et al. (2011) for evidence on how banks affected credit conditions through variation in these ratios in Ireland, the United Kingdom and the United States, respectively.

15. Although, the DF-GLS test rejects stationarity of the LTI variable, other unit root tests such as the KPSS test indicate that the variable is stationary. From a theoretical and intuitive perspective the LTI ratio (as with the LTV ratio) cannot be I(1) as that would imply that it can increase without bounds.

16. This test gives an F-statistic of 15.9 that is significant at the 1% level.

17. See European Commission (2012) for empirical evidence on the link between measures of uncertainty and euro area sovereign bond spreads.

18. See ECB (2009) for evidence that Irish banks used three-month Euribor as the base rate off which the standard variable rate was priced. The impact of monetary policy is captured by Euribor variable and therefore we assume perfect pass-through from changes in the ECB's main refinancing rate to money markets.

19. The concept of ‘undrawn equity’ has been shown to be a significant predictor of mortgage arrears (Whitley, Windram, & Cox, 2004).

20. See the ‘Financial Measures Programme Report’, Central Bank of Ireland, March 2011.

21. See Murphy (2004) for a comprehensive overview of this approach to modelling housing demand.

22. We also considered other normalisations such as by population, the number of households and total deposits but the results do not change significantly.

23. Owing to the small open nature of the economy and the historical linkages between the Irish and UK labour markets.

24. This is consistent with the annual rate depreciation on dwellings assumed by the Central Statistics Office (CSO) in Ireland.

25. See Gerlach-Kristen and McInerney (2014) for details.

26. We discuss the results for the mortgage and housing markets separately for ease of exposition, although the equations are estimated jointly.

27. Additional lags were found to be insignificant.

28. The long-run elasticity is calculated as the short-run coefficient divided by (1(1) --0.67), where 0.67 is the coefficient on the lagged dependent variable.

29. The long-run elasticity in an error-correction model is obtained by dividing the variable's short-run coefficient by the error-correction term.

30. The user cost coefficient does not change significantly when we consider that house price expectations may adapt more slowly than suggested by a four-quarter moving average of lagged house price inflation.

31. We considered three alternative measures of the ‘output gap’: the gap derived from a production function taken from the OECD's Economic Outlook, the difference between actual and potential output where the latter is constructed using a Hodrick-Prescott filtered trend, and a measure of detrended output where GDP is regressed on a linear and quadratic trend. The results reported above refer to the latter but do not change significantly when the alternative measures are used.

32. The restrictions do allow for some relief for first-time buyers who can avail of LTV ratios up to 90% of the purchase price for primary dwellings under 220,000 euro in value. We do not consider this relief in our simulations.

33. It is important to note that these data only pertain to a subset of Irish owned banks. As the newly introduced regulatory limits will affect all retail banks operating in the Irish market, a more relevant data-set for considering the impact of the measures on the housing market is maintained by the CSO. As Duffy and Hanlon (2014) show the share of lending with an LTV of over 80% in that data-set was 60.4% in 2012, which is significantly higher than the share from the CBI data-set for 2013. Unfortunately, as the share of lending at each LTV and LTI is unavailable from the CSO data-set for 2013, we confine our analysis to the CBI data. See Lydon and O'Hanlon (2012) for a discussion of the differences between the two data-sets.

34. CBI (2014) presents the share of lending for LTV and LTI ratios in particular ranges. In calculating the weighted averages, we take the midpoint of the range. For lending at LTV ratios over 100% percent and LTI ratios over 4.5 (as no maximum values are given) we assume that the ranges are 100%–110% and 4.5–5, respectively, and use the midpoints of these ranges.

35. Note that the deviation of the mortgage rate from the baseline level is in percentage points rather than percentage.