A. Introduction

30. The boom in equity and house prices, together with the associated rapid rise in private sector credit, raises the question of whether the Irish economy runs the risk of experiencing an asset price bubble. Speculative bubbles could cause overheating problems and, in the event of a reversal, seriously harm the balance sheets of both borrowers and lenders. From a policy perspective, therefore, it is essential to assess whether inappropriate policies (excessively loose monetary policies, overly generous fiscal incentives, and/or inadequate banking supervision) as opposed to structural determinants of asset prices (income growth, demographics, supply–side, and external variables) have been responsible for the boom.

31. This chapter examines the supervision system in Ireland and the performance of lending institutions, and provides econometric estimates of the contribution of different factors to the asset price boom. It presents evidence suggesting that the financial system is in a reasonably healthy state and systemic risks are low, and that while domestic monetary conditions have played a role in asset price developments, other factors, such as global asset prices, rising incomes, and the supply of housing, have made important contributions. Although the rise in asset prices is not simply caused by inappropriate macroeconomic or prudential policies, and the likelihood of a detrimental asset price reversal may be low, there is clearly a need to continue to monitor lending practices closely in the period ahead, especially as interest rates fall in the run–up to EMU.

B. Financial Supervision and Regulation in Ireland

32. The experience of industrial countries with asset price cycles in the late 1980s and early 1990s highlights the importance of supervision and lending practices, especially in the real estate market. In particular, the financial crises in the Nordic countries, although associated with the removal of financial restrictions and therefore not fully relevant to the case of Ireland, demonstrated how distortions in the financial sector and the absence of adequate supervision could increase the vulnerability of the system to negative shocks.¹⁶

33. In Ireland, the Central Bank is statutorily responsible for the supervision and regulation of banks and other financial institutions. Supervision covers banks, building societies, and other deposit–taking institutions, as well as a range of non–bank institutions and exchanges, including the stock exchange. At present it does not cover insurance companies, life and non–life businesses, insurance intermediaries, credit unions, and credit intermediaries, which are supervised by other government agencies.¹⁷ The principal objectives of supervision include ensuring the stability of the banking and financial system, containing systemic risks, and protecting depositors and investors.¹⁸ The Bank’s view is that regulation, which by nature is an interference in the market, should aim at addressing market failure and help financial institutions in making decisions with imperfect information. Thus, regulation should endeavor to lower the risks of bank failure, but if a firm becomes financially unviable, it should ensure that the exit causes the least disruption to the system rather than impeding its exit. Excessive protection and regulation would likely discourage competition and lower efficiency.

34. Since the supervision act of 1971, the legal framework has evolved considerably in response to implementation of European supervisory legislation and domestic financial innovation, for example to deal with the activities of the International Financial Services Center (IFSC) from the late 1980s. The current legal framework follows European law in the area of supervision. All EU banking and investment services directives have been incorporated into Irish law. The capital adequacy directive, implemented in 1995, requires the maintenance of a minimum level of capital to cover market and other risks. The Bank performs both on–site and off–site inspections of institutions, and also requires a regular flow of detailed financial data. There are areas, however, where this can be improved, in particular in relation to bad loan data. The Bank also requires all supervised entities to put in place and maintain internal control systems to ensure that their businesses are managed in accordance with sound administrative and accounting principles. Since 1989, a deposit protection scheme has been in place funded by all licensed credit institutions.

35. Against the background of the rapid rise in credit and in asset prices, the Bank has endeavored to enhance its supervision capabilities in recent years in order to ensure that lending practices do not fall below reasonable standards. The number of staff engaged in supervision has increased substantially in recent years (at present there are over 100 engaged in this function compared with around 50 in 1994). The Bank follows international norms in regulating aggregate lendings, in particular by requiring that capital adequacy ratios do not fall below a ceiling of 8 percent, but does not indulge in setting rules on individual loans for specific purposes. Lending institutions are expected to follow self–imposed rules of thumb in relation to loan–to–value and income–to–value ratios in the case of house mortgages. In general, borrowers are also required to undertake indemnity insurance arrangements, which in effect transfer risks of default to a third party. The Bank has also continually warned lending institutions of the hazards of excessive lending.

C. The Rise in Asset Prices and the Performance of Lending Institutions

36. Asset prices have soared in recent years. House prices were 40 percent higher in 1997 than in 1992, while equity prices rose by 150 percent over the same period. There have been further rapid increases during 1998, although in the case of equity prices the increases have been reversed since July. Credit to the private sector has also risen rapidly in recent years, growing faster than income. The assets of the credit institutions accounted for by credit to the private sector amounted to 94 percent of GDP in 1997, up from 70 percent in 1992 (Table 1). At the same time, the share of residential and other mortgage lending in total lending has remained roughly constant at 35 percent, suggesting that the rise in private credit has been broadly based and not simply reflecting increased borrowing for house purchases.

^{Table 1.}

Credit Institutions: Assets and Liabilities vis–vis Residents, Selected Years

(In billions of pounds; in percent of GDP in parantheses)

Source: Central Bank Annual Report: Various issues

^{Table 1.}

Credit Institutions: Assets and Liabilities vis–vis Residents, Selected Years

(In billions of pounds; in percent of GDP in parantheses)

		1992	1995	1997
Total Liabilities		24.6	33.0	50.2
		(82)	(85)	(107)
Of which:
	Non–government deposits	18.6	27.3	38.0
		(62)	(71)	(82)
Total assets		28.2	37.5	53.4
		(94)	(97)	(114)
Of which:
	Non–government credit	21.0	29.2	44.2
		(70)	(76)	(94)
Of which:
	Residential and other mortg.	7.5	11.1	15.6
		(25)	(29)	(33)

Source: Central Bank Annual Report: Various issues

View Table

^{Table 1.}

Credit Institutions: Assets and Liabilities vis–vis Residents, Selected Years

(In billions of pounds; in percent of GDP in parantheses)

		1992	1995	1997
Total Liabilities		24.6	33.0	50.2
		(82)	(85)	(107)
Of which:
	Non–government deposits	18.6	27.3	38.0
		(62)	(71)	(82)
Total assets		28.2	37.5	53.4
		(94)	(97)	(114)
Of which:
	Non–government credit	21.0	29.2	44.2
		(70)	(76)	(94)
Of which:
	Residential and other mortg.	7.5	11.1	15.6
		(25)	(29)	(33)

Source: Central Bank Annual Report: Various issues

37. Despite anecdotal evidence that mortgage lenders have lowered their standards to attract borrowers, a recent report commissioned by the government, the Bacon Report, does not consider this practice to be pervasive in the housing mortgage market or likely to pose any systemic risks.¹⁹ According to the Report, two factors have put pressure on adhering to lending criteria by lending institutions: rapid house price inflation, which has made it difficult for many borrowers to meet all the criteria, and increased competition. However, there does not seem to be any major systematic deterioration in lending institutions’ aggregate risk or in the quality of their portfolios. At the same time, the Report urges the Central Bank to continue to monitor lending practices in order to ensure that standards are maintained.

38. Despite the rapid growth in lending to the private sector, credit institutions’ capitalization ratios remain at comfortable levels, even though they have tended to fall in recent years, partly in response to increased competition and large dividend distribution (Table 2). Capital adequacy ratios—at around 12 percent for banks and 14 percent for building societies are well above the international norm of 8 percent. Moreover, according to preliminary indications capitalization has improved somewhat in 1998.

^{Table 2.}

Credit Institutions: Measures of Capitalization

(In percent of total risk assets)

Source: Central Bank (Figures submitted to the staff).

^*Defined as ratio of tier 1 capital to total risk assets.

^**Defined as ratio of total own funds to total risk assets.

^{Table 2.}

Credit Institutions: Measures of Capitalization

(In percent of total risk assets)

		1995	1996	1997
Tier 1 ratio *
	Banks	10.73	10.44	9.93
	Building societies	N.A.	13.26	12.36
Solvency ratio **
	Banks	13.88	13.57	13.39
	Building societies	18.60	13.12	13.08
Capital adequacy
	Banks	18.56	15.53	11.86
	Building societies	18.74	14.42	14.12

Source: Central Bank (Figures submitted to the staff).

^*Defined as ratio of tier 1 capital to total risk assets.

^**Defined as ratio of total own funds to total risk assets.

View Table

^{Table 2.}

Credit Institutions: Measures of Capitalization

(In percent of total risk assets)

		1995	1996	1997
Tier 1 ratio *
	Banks	10.73	10.44	9.93
	Building societies	N.A.	13.26	12.36
Solvency ratio **
	Banks	13.88	13.57	13.39
	Building societies	18.60	13.12	13.08
Capital adequacy
	Banks	18.56	15.53	11.86
	Building societies	18.74	14.42	14.12

Source: Central Bank (Figures submitted to the staff).

^*Defined as ratio of tier 1 capital to total risk assets.

^**Defined as ratio of total own funds to total risk assets.

39. Profitability, at around 1 percent in 1997 (Table 3), also is not low by international standards (see Table 10 in Drees and Pazarbasioglu for an international comparison). For banks, it is in fact higher than in 1992, but lower than the peak of 1994. Loan loss provisions do not indicate a rise in losses as a ratio to total assets. Net interest income for banks fell significantly to 1.7 percent in 1997,²⁰ but the ratio for the two largest banks, the Bank of Ireland and the Allied Irish Bank, was considerably higher at 3.7 percent. (The Moody ratings for these two banks (Al for the former and Aa3 for the latter—upgraded from Al last year) are also favorable. The recent fall in profitability and interest income also reflects in part increased competition and tightening of margins.

^{Table 3.}

Credit Institutions: Indicators of profitability

(In percent of total assets)

Source: Central Bank (Figures submitted to the staff)

^{Table 3.}

Credit Institutions: Indicators of profitability

(In percent of total assets)

		1992	1993	1994	1995	1996	1997
Net interest income
	Banks	3.60	3.61	3.12	2.67	2.31	1.73
	Building societies	2.68	3.14	2.85	2.86	2.57	2.18
Operating expenses
	Banks	3.26	3.29	2.82	2.33	2.03	1.55
	Building societies	1.75	1.84	1.69	1.58	1.44	1.47
Loan loss provisions
	Banks	0.90	0.51	0.23	0.15	0.12	0.10
	Building societies	0.12	0.14	0.07	0.06	0.09	0.06
Profit before tax
	Banks	0.88	1.27	1.44	1.34	1.34	1.05
	Building societies	1.31	1.62	1.35	1.49	1.34	0.96

Source: Central Bank (Figures submitted to the staff)

View Table

^{Table 3.}

Credit Institutions: Indicators of profitability

(In percent of total assets)

		1992	1993	1994	1995	1996	1997
Net interest income
	Banks	3.60	3.61	3.12	2.67	2.31	1.73
	Building societies	2.68	3.14	2.85	2.86	2.57	2.18
Operating expenses
	Banks	3.26	3.29	2.82	2.33	2.03	1.55
	Building societies	1.75	1.84	1.69	1.58	1.44	1.47
Loan loss provisions
	Banks	0.90	0.51	0.23	0.15	0.12	0.10
	Building societies	0.12	0.14	0.07	0.06	0.09	0.06
Profit before tax
	Banks	0.88	1.27	1.44	1.34	1.34	1.05
	Building societies	1.31	1.62	1.35	1.49	1.34	0.96

Source: Central Bank (Figures submitted to the staff)

D. Determinants of House Prices

40. The Bacon Report provides a comprehensive examination of the housing market in Ireland. It finds that the recent trend in house prices is an effect of several factors, including rising living standards, employment growth, social and demographic changes, in particular rapid growth in the rate of household formation, as well as falling interest rates and convergence of interest rates to those in Germany. Using annual data, the Report provides econometric estimates of house price and supply equations that broadly support these conclusions.

41. The Report discusses how the pressure in the housing market may be eased through improving infrastructure and supply conditions, in particular by increasing lands zoned for residential purposes. The Report also recommends fiscal measures to lower incentives for residential investment demand by lowering deductibility on interest payments and imposing a stamp duty on purchases of new houses by nonowner occupiers. At the same time it recommends that the stamp duty for second–hand houses be lowered in order to lower the barrier to entry into the market by first–time buyers. The government implemented most of these recommendations in April 1998.

42. This section provides further analysis of the factors that influence house prices by placing more emphasis on estimation issues (particularly the possibility of the presence of unit roots) and by using quarterly data. The principal objective is to assess the effect of monetary factors on house price inflation.

43. Based on relatively standard arguments, the relative price of new houses, denoted by h_t, is assumed to be influenced by loans advanced to purchase a new house in real terms, l_t, the interest rate, i_t, a measure of activity, y_t, and the addition to the stock of housing, q_t.²¹ The rationale for including loans advanced is obvious. Note that the relationship between this variable and house prices may be bi–directional (see below). The interest rate is the cost of borrowing and, therefore, the opportunity cost of buying a house as opposed to holding assets in an interest–bearing form, while income is a measure of consumers’ purchasing power, which is likely to influence both housing demand and prices. Finally, house completion and house prices are related through both supply and demand channels and the relationship is likely to go both ways: a larger addition to the stock of housing could ease pressure on prices by increasing supply, while higher prices could encourage an increase in house building.

44. The data are quarterly and cover the period 1983Q1–1997Q4.²² All nominal variables are deflated by the CPI. The three–month interest rate is used because a large part of the stock of mortgage loans is still with flexible rates. Quarterly data for GDP or disposable income are not available. Annual GDP series are converted into quarterly series using the cubic spline method. All variables are in natural logarithms.

45. The results of testing for the presence of unit roots (Augmented Dickey–Fuller tests) are reported in Table 4.²³ The tests do not reject the presence of unit roots in any of the series First differences, on the other hand, are all suggested to be free of unit roots.

^{Table 4.}

Augmented Dicky–Fuller Unit Root Tests ^1/

^1/See text for data definitions and sources. All regressions include and intercept and a time trend. Order of lags is chosen to maximize the Schwarz Bayesian Criterion. The crotical value in all cases is –3.48. An asterisk denotes rejection of a unit root at the 5 percent level.

^{Table 4.}

Augmented Dicky–Fuller Unit Root Tests ^1/

Variable	ADF
st	-2.15
Δst	-5.56*
it	-3.42
Δit	-8.32*
mt	-1.09
Δmt	-7.95*
yt	-0.14
Δyt	-5.75*
ust	-1.18
Δust	-6.99*
ht	-0.87
Δht	-8.22*
It	-0.20
Δlt	-4.55*
qt	-1.06
Δqt	-4.07*

^1/See text for data definitions and sources. All regressions include and intercept and a time trend. Order of lags is chosen to maximize the Schwarz Bayesian Criterion. The crotical value in all cases is –3.48. An asterisk denotes rejection of a unit root at the 5 percent level.

View Table

^{Table 4.}

Augmented Dicky–Fuller Unit Root Tests ^1/

Variable	ADF
st	-2.15
Δst	-5.56*
it	-3.42
Δit	-8.32*
mt	-1.09
Δmt	-7.95*
yt	-0.14
Δyt	-5.75*
ust	-1.18
Δust	-6.99*
ht	-0.87
Δht	-8.22*
It	-0.20
Δlt	-4.55*
qt	-1.06
Δqt	-4.07*

^1/See text for data definitions and sources. All regressions include and intercept and a time trend. Order of lags is chosen to maximize the Schwarz Bayesian Criterion. The crotical value in all cases is –3.48. An asterisk denotes rejection of a unit root at the 5 percent level.

46. In the presence of series with unit roots an appropriate methodology to use is the Johansen maximum–likelihood procedure to test for and estimate both long–run cointegrating and short–run error–correction relationships. The standard Johansen procedure, however, assumes that all the included variables are endogenous to the system, which may not always be a reasonable assumption.²⁴ A modified version of this procedure is used that allows for the presence of exogenous non–stationary variables.²⁵ It is not unreasonable to assume that short–term interest rates and GDP are exogenous variables in this model, while house prices and house loans are endogenous. It is also possible, as suggested above, that housing completion is influenced by other variables in the model. This was tested and not supported by the data (the error–correction equation for this variable did not include any significant variables; the results are not reported here). Therefore, the estimations reported below assume that house prices and housing loans are endogenous I(1) variables, whereas the other three variables are exogenous I(1) variables.

47. Table 5 reports cointegration likelihood ratio tests for a VAR(l) relationship (the order of the VAR was selected by the SB Criterion; see also footnote 25). The results support the existence of one cointegrating vector. Table 6 presents the maximum–likelihood estimates of the cointegrating vector, including exclusions tests (chi–squared) for the long–run coefficients. The results suggest that house prices and housing loans are positively correlated. They also indicate that housing completion has a negative effect on prices, consistent with the hypothesis that increased supply eases pressure on prices. Income and the interest rate do not seem to have significant coefficients in the long–run relationship.

^{Table 5.}

Cointegration Likelihood Ratio Tests for the House Price Equation ^1/

^1/Based on a cointegrating VAR of order 1, Selected using the Schwarz Bayesian Criterion, with unrestricted intercepts and restricted trend, which includes h_t and l_t as endogenous I(1) variables, i_t, y_t, and q_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1983q1–1997q4.

^{Table 5.}

Cointegration Likelihood Ratio Tests for the House Price Equation ^1/

		Maximal Eigenvalue		Trace
Null	Alternative	Statistic	95% CV	Statistic	95% CV
r = 0	r = 1	44.85	27.80	63.21	40.37
r ≤ 1	r = 2	18.37	20.47	18.37	20.47

^1/Based on a cointegrating VAR of order 1, Selected using the Schwarz Bayesian Criterion, with unrestricted intercepts and restricted trend, which includes h_t and l_t as endogenous I(1) variables, i_t, y_t, and q_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1983q1–1997q4.

View Table

^{Table 5.}

Cointegration Likelihood Ratio Tests for the House Price Equation ^1/

		Maximal Eigenvalue		Trace
Null	Alternative	Statistic	95% CV	Statistic	95% CV
r = 0	r = 1	44.85	27.80	63.21	40.37
r ≤ 1	r = 2	18.37	20.47	18.37	20.47

^1/Based on a cointegrating VAR of order 1, Selected using the Schwarz Bayesian Criterion, with unrestricted intercepts and restricted trend, which includes h_t and l_t as endogenous I(1) variables, i_t, y_t, and q_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1983q1–1997q4.

^{Table 6.}

Maximum–Likelihood Estimates of the Coefficients in the Cointegrating VAR(1) for Relative House Prices and Likelihood Ratio Tests of Exclusions

1/ The sample period is 1983ql–1997q4. x² statistic in each case tests the restriction that the coefficient is equal to zero in the cointegrating relationship; and an asterisk denotes significance at the 5 percent level.

^{Table 6.}

Maximum–Likelihood Estimates of the Coefficients in the Cointegrating VAR(1) for Relative House Prices and Likelihood Ratio Tests of Exclusions

Variable	Cointegrating Coefficients		X2(1) [Rejection Probability]
ht	-1.00	*	16.56	[0.00]
lt	0.52	*	22.84	[0.00]
it	0.03		0.48	[0.49]
yt	0.46		0.82	[0.37]
qt	-0.33	*	22.08	[0.00]
t	-0.02		1.84	[0.17]

1/ The sample period is 1983ql–1997q4. x² statistic in each case tests the restriction that the coefficient is equal to zero in the cointegrating relationship; and an asterisk denotes significance at the 5 percent level.

View Table

^{Table 6.}

Maximum–Likelihood Estimates of the Coefficients in the Cointegrating VAR(1) for Relative House Prices and Likelihood Ratio Tests of Exclusions

Variable	Cointegrating Coefficients		X2(1) [Rejection Probability]
ht	-1.00	*	16.56	[0.00]
lt	0.52	*	22.84	[0.00]
it	0.03		0.48	[0.49]
yt	0.46		0.82	[0.37]
qt	-0.33	*	22.08	[0.00]
t	-0.02		1.84	[0.17]

1/ The sample period is 1983ql–1997q4. x² statistic in each case tests the restriction that the coefficient is equal to zero in the cointegrating relationship; and an asterisk denotes significance at the 5 percent level.

48. Since there are two endogenous variables in the system, the maximum–likelihood estimation gives rise to error–correction equations for both variables, which include the same error–correction term derived from the long–run relationship. Table 7 contains the results for house prices. GDP now has a significant coefficient (at the 10 percent level) but the interest rate remains insignificant. Importantly (see below), the error–correction term also has a significant coefficient. Table 8 presents the error–correction equation for housing loans. Here, both GDP and housing completion have positive significant coefficients, as does the error–correction term, although the equation seems to suffer from serial correlation.²⁶ The significance of both error–correction terms indicates that house prices and housing loans are simultaneously determined. Housing loans, moreover, are also affected by the increase in the stock of housing and factors that influence this variable. This is an important result because it implies that the rise in house prices is not simply driven by increased credit.

^{Table 7.}

Estimated Error–Correction Model for New House Prices Based on the Cointegrating VAR(1) in Table 6 ^1/

^1/The dependent variable, Δh_t, is the change in the logarithm of new house prices deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

^{Table 7.}

Estimated Error–Correction Model for New House Prices Based on the Cointegrating VAR(1) in Table 6 ^1/

Regressor				Coefficient		Standard Error
Δit				-0.02		0.02
Δyt				0.75	**	0.39
Δqt				-0.02		0.02
ecmt-1				-0.24	*	0.02
R2	=	0.38
s.e.	=	0.022
X2sc (4)	=	6.57	[0.16]	Test for serial correlation
X2ff (1)	=	0.07	[0.79]	Test for functional form
X2n (2)	=	1.07	[0.59]	Test for normality
X2he (1)	=	1.21	[0.27]	Test for heteroscedasticity

^1/The dependent variable, Δh_t, is the change in the logarithm of new house prices deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

View Table

^{Table 7.}

Estimated Error–Correction Model for New House Prices Based on the Cointegrating VAR(1) in Table 6 ^1/

Regressor				Coefficient		Standard Error
Δit				-0.02		0.02
Δyt				0.75	**	0.39
Δqt				-0.02		0.02
ecmt-1				-0.24	*	0.02
R2	=	0.38
s.e.	=	0.022
X2sc (4)	=	6.57	[0.16]	Test for serial correlation
X2ff (1)	=	0.07	[0.79]	Test for functional form
X2n (2)	=	1.07	[0.59]	Test for normality
X2he (1)	=	1.21	[0.27]	Test for heteroscedasticity

^1/The dependent variable, Δh_t, is the change in the logarithm of new house prices deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

^{Table 8.}

Estimated Error–Correction Model for Loans Paid for New Houses Based on the Cointegrating VAR(2) in Table 6 ^1/

^1/The dependent variable, Δl_t, is the change in the logarithm of loans deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

^{Table 8.}

Estimated Error–Correction Model for Loans Paid for New Houses Based on the Cointegrating VAR(2) in Table 6 ^1/

Regressor				Coefficient		Standard Error
Δit				-0.04		0.09
Δyt				4.56	*	1.98
Δqt				0.71	*	0.10
ecmt-1				1.15	*	0.11
R2	=	0.59
s.e.	=	0.11
X2sc (4)	=	12.58	[0.01]	Test for serial correlation
X2ff (1)	=	0.07	[0.78]	Test for functional form
X2n (2)	=	0.83	[0.66]	Test for normality
X2he (1)	=	0.08	[0.78]	Test for heteroscedasticity

^1/The dependent variable, Δl_t, is the change in the logarithm of loans deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

View Table

^{Table 8.}

Estimated Error–Correction Model for Loans Paid for New Houses Based on the Cointegrating VAR(2) in Table 6 ^1/

Regressor				Coefficient		Standard Error
Δit				-0.04		0.09
Δyt				4.56	*	1.98
Δqt				0.71	*	0.10
ecmt-1				1.15	*	0.11
R2	=	0.59
s.e.	=	0.11
X2sc (4)	=	12.58	[0.01]	Test for serial correlation
X2ff (1)	=	0.07	[0.78]	Test for functional form
X2n (2)	=	0.83	[0.66]	Test for normality
X2he (1)	=	0.08	[0.78]	Test for heteroscedasticity

^1/The dependent variable, Δl_t, is the change in the logarithm of loans deflated by CPI; ecm_t-1 is lagged residuals from the conintegrating relationship in Table 6; sample period is 1983q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

E. Determinants of Equity Prices

49. In this section we examine the determinants of equity prices. In particular, we test whether equity prices are influenced by the nominal interest rate, credit to the private sector, the level of activity, and global equity prices. As in the case of house prices, both the interest rate—representing the opportunity cost of holding equity—and credit to the private sector are included in order to allow for the possibility that credit rationing could affect equity holding. The level of activity, on the other hand, is an indicator of the stage of the business cycle and the economic agents’ purchasing power, while the price of global equities, which for investors are an alternative to domestic equities, is also likely to influence domestic equity prices.²⁷

50. Irish equity prices and broad money are deflated by the consumer price index, and are denoted by s_t and m_t, respectively. The U.S. equity prices are converted into domestic currency and deflated by domestic CPI, and are denoted by us_t. The interest rate and GDP, as before, are denoted by i_t and y_t, respectively. All variables are in natural logarithms.²⁸

51. The Augmented Dickey–Fuller unit root tests, reported in Table 4, do not reject the presence of unit roots in any of the levels but suggest that first differences are all I(0). Table 9 presents the cointegration test (under the reasonable assumption that only equity prices are endogenous in the system). The results, perhaps not surprisingly, do not provide support for the existence of a cointegrating relationship, reflecting the difficulty of explaining long–run equity price movements.

^{Table 9.}

Cointegration Likelihood Ratio Tests for the Equity Price Equation ^1/

^1/Based on a cointegrating VAR of order 2, with unrestricted intercepts and restricted trend, which includes s_t as the endogenous I(1) variable, m_t, i_t, y_t, and us_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1980q1–1997q4.

^{Table 9.}

Cointegration Likelihood Ratio Tests for the Equity Price Equation ^1/

		Maximal Eigenvalue		Trace
Null	Alternative	Statistic	95% CV	Statistic	95% CV
r = 0	r = 1	15.84	23.17	15.84	23.17

^1/Based on a cointegrating VAR of order 2, with unrestricted intercepts and restricted trend, which includes s_t as the endogenous I(1) variable, m_t, i_t, y_t, and us_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1980q1–1997q4.

View Table

^{Table 9.}

Cointegration Likelihood Ratio Tests for the Equity Price Equation ^1/

		Maximal Eigenvalue		Trace
Null	Alternative	Statistic	95% CV	Statistic	95% CV
r = 0	r = 1	15.84	23.17	15.84	23.17

^1/Based on a cointegrating VAR of order 2, with unrestricted intercepts and restricted trend, which includes s_t as the endogenous I(1) variable, m_t, i_t, y_t, and us_t as exogenous I(1) variables, and first differences of the latter variables as I(0) exogenous variables. Sample period is 1980q1–1997q4.

52. Since no long–run relationship between the levels of the variables appears to exist, we test for the presence of a relationship between first differences, This can be done using OLS since all first differences are I(0). The results are reported in Table 10 (longer lags were not significant). They strongly indicate that the main determinants of growth in equity prices are growth in U.S equity prices and in activity. The interest rate is significant at the 10 percent level only and, moreover, its lagged value is significant with the opposite sign and a similar magnitude. The stock of money has an insignificant coefficient.

^{Table 10.}

Estimated Model for the Rate of Change in Equity Prices using OLS ^1/

^1/The dependent variable, Δs_t, is the change in the logarithm of equity prices deflated by CPI; sample period is 1980q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

^{Table 10.}

Estimated Model for the Rate of Change in Equity Prices using OLS ^1/

Regressor				Coefficient		Standard Error
Δst-1				0.40	*	0.12
Δmt				-0.01		0.19
Δit				-0.09	**	0.05
Δyt				2.81	*	1.25
Δust				0.54	*	0.10
Δmt-1				0.04		0.19
Δit-1				0.09	**	0.05
Δyt-1				-2.17		1.33
Δust-1				-0.10		0.12
R2	=	0.53
s.e.	=	0.067
X2sc (4)	=	3.97	[0.41]	Test for serial correlation
X2ff (1)	=	1.79	[0.18]	Test for functional form
X2n (2)	=	3.90	[0.14]	Test for normality
X2he (1)	=	1.47	[0.23]	Test for heteroscedasticity

^1/The dependent variable, Δs_t, is the change in the logarithm of equity prices deflated by CPI; sample period is 1980q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

View Table

^{Table 10.}

Estimated Model for the Rate of Change in Equity Prices using OLS ^1/

Regressor				Coefficient		Standard Error
Δst-1				0.40	*	0.12
Δmt				-0.01		0.19
Δit				-0.09	**	0.05
Δyt				2.81	*	1.25
Δust				0.54	*	0.10
Δmt-1				0.04		0.19
Δit-1				0.09	**	0.05
Δyt-1				-2.17		1.33
Δust-1				-0.10		0.12
R2	=	0.53
s.e.	=	0.067
X2sc (4)	=	3.97	[0.41]	Test for serial correlation
X2ff (1)	=	1.79	[0.18]	Test for functional form
X2n (2)	=	3.90	[0.14]	Test for normality
X2he (1)	=	1.47	[0.23]	Test for heteroscedasticity

^1/The dependent variable, Δs_t, is the change in the logarithm of equity prices deflated by CPI; sample period is 1980q1–1997q4; and single and double asterisks, respectively, denote significance at the 5 and 10 percent levels.

F. Concluding Remarks

53. Evidence presented in this chapter does not seem to suggest that recent developments in asset prices and private sector credit pose significant systemic risks to the financial system in Ireland. Capital adequacy and profitability, although lower than in recent years, are both at comfortable levels. Econometric analysis, moreover, does not implicate loose monetary conditions as the principal exogenous cause of asset price inflation. In the case of house prices, the results suggest that while the availability of credit at low interest rates has contributed to higher house price inflation, so have rising incomes and supply pressures. Moreover, increased mortgage lending has been largely a response to higher prices and increased demand for housing and the factors that have contributed to this. The results for equity prices suggest that their growth has been influenced by movements in global equity prices and income growth, rather than domestic monetary conditions per se.

54. The significance of structural factors in the evolution of house prices, the likelihood of low interest rates in the period ahead, and the apparent strength of the banking sector in Ireland imply that the risk of a major house price reversal with detrimental ramifications may be small. On the other hand, the risk of a reversal in equity prices is higher since they seem to respond heavily to movements in global equity prices. Indeed the Irish stock index has fallen by more than 25 percent from its highs earlier in the year. However, while fluctuations in equity prices affect the ability of businesses to raise capital, they are unlikely to have a large impact on consumers’ and banks’ balance sheets because of the relatively small share of equities in total wealth.

55. While asset price inflation in Ireland may largely reflect fundamental and exogenous factors, rather than inappropriately loose monetary and financial policies, tighter macroeconomic policy will be necessary in the period ahead as interest rates fall further. Recently implemented policies to restrict fiscal subsidies to mortgage borrowing and to improve supply conditions in the housing market are also likely to moderate the boom in house prices. Moreover, even though capital adequacy ratios remain well above international norms, and imposing unnecessarily restrictive regulations would likely impede economic growth and competition in financial markets, the Central Bank will need to continue to monitor lending practices closely and to warn lending institutions of the hazards of imprudence.