Executive Summary
Author: International Monetary Fund. Monetary and Capital Markets Department
Publication Date:
29 Sep 2016
eISBN:
9781475542257
Language:
English
Keywords:
ISCR; CR; interest rate; banking sector; sensitivity analysis; investment funds; negative equity; interest rate shock; long-term debt; household financial assets; primary sector; CRE price; current ratio; Loans; Mortgages; Housing prices; Stocks; Corporate sector; Global; Europe; credit condition; short-term debt
Download PDF (2.7 MB)

This Technical Note discusses the findings and recommendations made in the Financial Sector Assessment Program for Ireland in the areas of nonbank sector stability. Both nonparametric and parametric methods suggest that the residential real estate market in Ireland is close to or moderately below its equilibrium level. Two standard metrics of price-to-income and price-to-rent ratios show that following a protracted period of overvaluation prior to the crisis and a correction afterward, the market has been close to its equilibrium level in recent quarters. Households have deleveraged, but are still highly indebted. The stability analysis results also suggest that vulnerabilities among nonfinancial firms have moderated in recent years.

Abstract

This Technical Note discusses the findings and recommendations made in the Financial Sector Assessment Program for Ireland in the areas of nonbank sector stability. Both nonparametric and parametric methods suggest that the residential real estate market in Ireland is close to or moderately below its equilibrium level. Two standard metrics of price-to-income and price-to-rent ratios show that following a protracted period of overvaluation prior to the crisis and a correction afterward, the market has been close to its equilibrium level in recent quarters. Households have deleveraged, but are still highly indebted. The stability analysis results also suggest that vulnerabilities among nonfinancial firms have moderated in recent years.

Inter-Sectoral Interconnectedness Analysis

Cross-border interlinkages via Irish-domiciled funds industry and multinational companies are a key feature of the financial network. Ireland plays a role as one of a number of significant hubs for the global funds industry. The tight linkages between the rest of the world (ROW) and non-financial corporations reflect the large presence of foreign-controlled multinational companies in Ireland. The direct bilateral connection between the ROW and Irish households is insignificant, but the household sector is indirectly exposed to global shocks through their balance sheets of insurance companies and pension funds.

The banking sector is tightly connected with the real sectors of the Irish economy, but not with many of the other financial sectors. Banks are systemically important sources and destinations of funding in Ireland and have strong bilateral connections with households and non-financial corporates. There are tight bank-sovereign financial linkages through banks’ government bond holdings and the government’s holdings of bank equity. Meanwhile, the linkages between the banking system and the funds industry, insurance companies or pension funds are very limited.

Irish-domiciled other financial intermediaries (OFIs) do not have direct strong financial linkages with the domestic real sectors. The connections between the two sectors mainly reflect intra-company transactions between treasury companies and their parent multinational companies. OFIs do not directly provide loans to households to any significant degree, but are largely connected through purchases of mortgage loan securitizations and loan sales from banks.

Real Estate Market Analysis

Residential real estate (RRE) and commercial real estate (CRE) prices in Ireland have been rising rapidly in recent years, raising concerns about possible overvaluation and a new build-up of imbalances. The rebound in the Irish RRE and CRE markets has been more vigorous than in other countries.

Both non-parametric and parametric methods suggest that the RRE market is close to or moderately below its equilibrium level. Two standard metrics of price-to-income and price-to-rent ratios show that following a protracted period of overvaluation prior to the crisis and a correction afterwards, the market has been close to its equilibrium level in recent quarters.

Similar approaches send mixed signals regarding the valuation of current CRE prices, but early signals of new imbalances in the CRE market should not be overlooked. The price-to-rent ratio suggests that following a sharp correction at the onset of the crisis, CRE prices increased above their historical averages in 2014 and were moderately overvalued. Frequency and HP filters show that CRE prices have recently converged to their long-term trends. Foreign investment inflows or equity funding like REITs can easily reverse if market sentiment changes, which could lead to a sharp drop in CRE values. Lenders with remaining exposures could face another hit from a collapse in collateral values via financial “decelerator” mechanisms, as observed in the post-crisis period. Therefore, the authorities will need to continue to closely monitor CRE lending and evaluate any early signals of a build-up of new market imbalances.

Household Sector Analysis

Households have deleveraged, but are still highly indebted. Some mortgage loans remain in negative equity, and almost all mortgages have variable interest rates (tracker and standard variable interest rates). Negative equity is a well-documented cause of default besides income shocks. Interest rates on existing tracker mortgages are currently very low, reflecting the zero ECB policy rate. If interest rates rise, interest payments will increase and some households may face difficulty servicing their debt.

Existing borrowers with high loan-to-value (LTV) ratios or standard variable rates are more vulnerable than other groups of households to the FSAP adverse scenario. Looking explicitly at the LTV distribution, there is a strictly increasing relationship between LTV levels and the stressed probability of default.

Going forward, it will be crucial to continue to collect loan-level data for systemic risk assessment, and the Central Bank of Ireland should move forward with the establishment of the Central Credit Register (CCR) to ensure that individual households’ credit information can be accurately verifiable.

Corporate Sector Analysis

The stability analysis assesses the financial resilience of the non-financial corporate (NFC) sector. The results suggest that vulnerabilities among NFCs have moderated in recent years. The interest cover ratio (ICR) of the median firm across all categories of firm size and sectors increased steadily in recent years, following a sharp decline in 2008-09, while the share of risky debt (i.e. owned by firms with ICR lower than one) in total debt declined to the pre-crisis level. Nevertheless, the NFC sector—especially smaller firms—remains highly vulnerable. More specifically, the analysis show that small domestic firms account for most of the firms that are under “technical default” (i.e. with ICR less than one), and that the share of risky debt among these firms constituted nearly half of their total debt, well above the shares of medium-sized and large enterprises.

A sensitivity analysis indicates that the sector, and especially smaller firms, remains highly vulnerable even to non-extreme shocks, but, ceteris paribus, banks’ regulatory capital would still be above the minimum requirement. An adverse shock, which comprises a decline in profitability and an increase in interest rates, is likely to push many firms into a vulnerable state. The share of firms with ICR lower than one would triple to nearly fifty percent, largely reflecting the deterioration in the financial health of small firms. In such a scenario, the share of risky debt would increase to the level observed during the financial crisis, resulting in a significant increase in new corporate defaults. Nevertheless, given the moderate share of corporate loans on banks’ loan books, and banks’ current comfortable capital positions, the analysis indicates that banks would still be able to keep their regulatory Tier 1 capital well above the minimum requirement.

Inter-Sectoral Financial Interlinkages1

A. Introduction

1. This note presents a map of inter-sectoral interconnectedness of various financial and real sectors in Ireland and assesses potential channels of risk contagion. As financial interlinkages among the sectors have become complex, a system-wide perspective is required to assess financial stability. The system-wide view can produce relevant insights that are easily overlooked if one analyzes the financial system on a sector-by-sector basis. It also enables to identify which sectors can play a role in transmitting shocks within and across the Irish border.

2. The Central Bank of Ireland has made further enhancements to the Quarterly Financial Accounts (QFA) data.2 It started to publish the “Whom-to-Whom” data for deposits and loans in April 2015, which can provide greater detail on inter-sectoral relationship for financial flows and positions and the transmission of risks among different sectors in Ireland (Cussen, 2015 and forthcoming). Using the QFA data, this note covers nine sectors: (1) monetary financial institutions (MFIs) including deposit-taking institutions and money market funds; (2) non-money market investment funds (NMIFs); (3) other financial intermediaries including financial vehicle corporations and holding companies (OFIs); (4) insurance companies (ICs); (5) pension funds (PFs); (6) governments (GOV); (7) households including non-profit institutions serving households (HHs); (8) non-financial corporations (NFCs); and (9) the rest of the world (ROW). MFIs are divided into three subsectors: domestic-controlled monetary financial institutions, foreign-controlled monetary financial institutions, and money market funds. Bilateral exposures of the Central Bank of Ireland are also examined.3

3. The outline of this section is as follows. Subsection B uses the Quarterly Financial Accounts data to take stock of the relative size and composition trend of balance sheets of individual sectors and describe domestic and cross-border financial linkages among sectors. Subsection C shows results from a domestic financial network analysis to identify systemically important sectors and potential transmission channels that system-wide risk can spread among sectors. Subsection D considers options to improve surveillance.

B. Sectoral Balance Sheets: Assets and Liabilities

4. Total financial assets in Ireland are large and well-diversified among financial sectors, and have continued to grow in the post-crisis period. As shown in Table 1, the outstanding balance of total financial assets amounted to 2820 percent of GDP (€6053 billion) as of 2015Q4, of which 2,033 percent of GDP was captured by the financial sectors, including monetary financial institutions (533 percent of GDP), non-money market investment funds (811 percent of GDP), other financial intermediaries (515 percent of GDP), and insurance companies and pension funds (174 percent of GDP). During 2011–13, the outstanding balance dropped by 5 percent from the previous peak level in 2010 (2,738 percent of GDP), but exceeded the peak since 2014 (Figure 1). While all the financial sectors had more or less a balanced position, the government and non-financial corporations had a large negative net financial asset position (71 and -119 percent of GDP) and households had a large positive net financial asset position (92 percent of GDP).4

Table 1.

Ireland: Balance Sheet Composition of Financial Assets and Liabilities by Sector

(Assets, percent of total financial assets of each sector, as of 2015Q4)
MFIsNMIFsOFIsICsPFsGOVHHsNFCsShare of each instrument (Percent in total financial assets)ROW
Currency & Deposits25468420378114
Short-Term Debt Security30251220070
Long-Term Debt Securities2133121641200178
Short-Term Loans4090010946
Long-Term Loans152470013011149
Equity1331510474712512521
MMF Shares0022100009
Non-MMF Fund Shares0110403401527
Non-life Insurance Reserves00014002011
Life Insurance and Annuity Entitlement00000012013
Pension entitlements00000033020
Fin Derivatives and Stock Options21011010034
Trade Credits and Advance00040001332
Other Account Receivables1643292745
Total100100100100100100100100100100
Size of Financial Assets (Percent of GDP)533811515122514216558128202213
Share in Total Financial Assets (Percent of Total Financial Assets)192918421621100
(Liabilities, percent of total financial liabilities of each sector, as of 2015Q4)
article image
MFIsNMIFsOFIsICsPFsGOVHHsNFCsShare of each instrument (Percent in total financial assets)ROW
Currency & Deposits42000090089
Short-Term Debt Security10200100110
Long-Term Debt Securities3030016001820
Short-Term Loans04220003464
Long-Term Loans0120102591221212
Equity1001811010602029
MMF Shares41000000070
Non-MMF Fund Shares077000000225
Non-life Insurance Reserves00017000011
Life Insurance and Annuity Entitlement00063000030
Pension entitlements00009900020
Fin Derivatives and Stock Options3900000034
Trade Credits and Advance0014001923
Other Account Receivables1963035454
Total100100100100100100100100100100
Size of Financial Assets (Percent of GDP)535851482124551127469929312103
Share in Total Financial Assets (Percent of Total Financial Assets)182916424324100
Sources: CBI; CSO; and IMF staff calculation.
Figure 1.
Figure 1.

Ireland: Total Financial Assets and Liabilities by Sector

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

5. The recent recovery of total financial assets is led by a rapid growth in the non-money market investment fund industry (Figure 1).5 Financial assets of the investment funds have grown by three folds since the crisis (from about 270 percent of GDP in 2009 to 811 percent of GDP at 2015Q4), while those of monetary financial institutions have decreased by 45 percent, falling from first place (from 963 percent of GDP to 533 percent of GDP).6 Other financial intermediaries also had 17 percent less financial assets in 2015Q4 (515 percent of GDP) compared with the level in 2009 (577 percent of GDP), but still account for 18 percent of total financial assets in Ireland.7 On the other hand, insurance companies and pension funds have played a small role in financial intermediation (6 percent of total financial assets and liabilities) and have increased somewhat since the crisis.

6. For most financial sectors, assets are mainly allocated to equities, long-term debt securities, or long-term loans, but are funded by different instruments (Table 1). As of 2015Q3, monetary financial institutions held more than a half of their financial assets in liquid form, either cash and deposits or short-term debt securities (e.g., certificates of deposit, commercial paper, treasury bills, etc.), and they had two main sources of funding: deposits for retail banks and credit unions, and shares for money market funds. The assets of non-money market funds consisted of holdings of long-term debt securities, equities, and financial derivatives, while the lion’s share of their liabilities was fund shares (77 percent). The bulk of the OFI assets were loans (mostly long-term), while the liabilities were from various sources, such as long-term debt securities, equities, and loans. Insurance companies owed mostly in the form of insurance premia and held non-MMF shares and long-term debt securities, while pension funds’ assets consisted of equities and long-term debt securities and their liabilities were completely funded by pension entitlements.

7. Household financial liabilities have been contracting in a deleveraging process since the financial crisis (Figure 2). The outstanding balance of household debt, almost entirely in the form of long-term mortgage loans, decreased by 27 percent (-€54 billion) between 2008 and 2015Q4. On the other hand, financial assets of households have remained stable, as the increase of pension entitlements offset the decrease of other types of assets, such as deposits, equity holdings, and claims on insurance companies. As of 2015Q4, deposits and claims on pension funds accounted for about 70 percent of household financial assets. Equity holdings and claims on life insurance companies made up the rest. The household sector was a net lender, with net financial assets amounting to €197 billion (92 percent of GDP).

Figure 2.
Figure 2.

Ireland: Changes in Financial Assets and Liabilities of Households

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

8. The volume of corporate financial assets increased 2.4 times, while liabilities doubled during the post-crisis period. As shown in Table 1, equities had the largest share of both assets and liabilities (51 percent and 60 percent), largely due to new entrances of multinational companies. Corporate debt increased from €323 billion to €402 billion (an increase of 15 percentage points of GDP) between 2008 and 2015Q4, but its share in total corporate liabilities has been in a gradual downward trend (Figure 3). Yet, as noted in Cussen (2015), Irish corporate debt would be substantially reduced by the exclusion of foreign-controlled multinational companies.8 External debt, mainly owned by multinational firms, stood at €214 billion in 2015Q4, accounting for more than 50 percent of total corporate debt in Ireland.

Figure 3.
Figure 3.

Ireland: Non-Financial Corporate Debt

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

9. The rest of the world (ROW) has a very large amount of financial assets and liabilities against Irish-domiciled entities and residents, indicating tight cross-border financial linkages (Table 1). The volume of the ROW’s assets and liabilities against Irish-domiciled entities and residents stands out among European countries, next to Luxembourg (Figure 4).9 As of 2015Q4, the ROW’s financial assets represented almost 2,214 percent of GDP. MMF and Non-MMF shares issued by Irish-domiciled funds and held by non-residents amounted to €1,748 billion or 815 percent of GDP, and accounted for 37 percent of total financial asset of the ROW, which have increased rapidly in recent years. The ROW’s holdings of debt and equity reached €1,106 and €1,016 billion (515 and 473 percent of GDP) and accounted for 23 and 21 percent of the ROW’s financial assets, respectively. The ROW’s liabilities were slightly smaller (2,103 percent of GDP), and mainly comprised of equities, fixed-income instruments, and loans (the first two each comprise about 29 percent of total liabilities, and the last 16 percent of the total). The characteristics of cross-border financial linkages are discussed in the following subsection in detail.

Figure 4.
Figure 4.

Selected Countries: Financial Assets and Liabilities of the Rest of the World

(Percent of GDP)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CEIC; WEO; and IMF staff calculation.

C. Inter-sectoral Financial Linkages10

10. Cross-border interlinkages via Irish-domiciled funds industry and multinational companies are a key feature of the financial network (Figure 5).11 Cross border linkages are strongest in two sectors, namely the Irish-domiciled funds industry and Irish-domiciled multinational companies: all bilateral connections over 100 percent of GDP were the ones between the ROW and domestic sectors in 2015Q2 (Table 2). Interesting features of the cross-border interconnections are as follow:

  • Ireland plays a role as one of a number of significant hubs for the global funds industry. As of 2015Q2, non-residents held more than 90 percent of total liabilities of money market funds and non-money market investment funds in gross terms;

  • The tight linkages between the ROW and NFCs reflect the large presence of foreign-controlled multinational companies in Ireland; and

  • The bilateral connection between the ROW and Irish households is insignificant.

Table 2.

Ireland: Financial Network Matrix (Gross Bilateral Position)

article image
Sources: CBI; and IMF staff calculation.
Figure 5.
Figure 5.

Ireland: Financial Network Map1/,2/,3/

(As of 2015Q2)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Note:1/ DMFIs = Domestic Monetary Financial Institutions, FMFIs = Foreign Monetary Financial Institutions, MMFs = Money Market Funds.2/ The size of each vertex represents the size of total financial assets.3/ The thickness of arrows from a sector to another depicts the volume of bilateral exposures from a creditor to a debtor.Sources: CBI calculation; NodeXL; and IMF staff calculation.

Properties of Financial Network in Ireland

A financial network is a set of elements called nodes (sectors) connected by links (exposures). The network is “directed,” as the relationship goes from a sector to another. As of 2015Q2, it is an incomplete network because some sectors do not have bilateral exposures bigger than 1 percent of GDP.

Degree refers to the sum of incoming links (In-Degree) and outgoing links (Out-Degree) of a sector. A sector with higher degree is more central, meaning that the sectoral is highly connected. In-and Out-Degree make reference to the direction of these links.

The following figures indicate the dominance of the ROW, NFCs, OFIs, and domestic banks in the financial network. They show In-Degree, Out-Degree, and the size of total assets of each sector in Ireland. The ROW has connections with all the other sectors, NFCs are exposed to all the counterparties except MMFs and PFs, and domestic banks have very small exposures against non-money market investment funds, PFs, and foreign-controlled banks.

uA01fig01

Degree Distribution and Total Assets in the Irish Financial Network

(Number of In-and Out-Degree, Percent of GDP)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; and IMF staff calculation.

Focusing on the Irish domestic financial network, the following figure crystallizes the potential role of four domestic sectors (i.e., domestic banks, OFIs, NFCs, HHs) in distributing and absorbing funding.

uA01fig02

Degree Distribution and Total Assets excluding the ROW

(Number of In-and Out-Degree, Percent of GDP)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; and IMF staff calculation.

11. Several connections can act as key transmission channels of risks across borders. First, domestic banks and foreign bank’s subsidiaries hold both financial assets and liabilities against the ROW and thus are exposed to negative spillovers from credit and funding shocks that non-residents face. However, the spillovers appear to be less of a concern compared with the pre-crisis period, given the fact that both foreign banks’ exposures to Ireland and Irish domestic banks’ claims to non-residents have reduced significantly in recent years.12 Second, Irish insurance companies and pension funds are net creditors to the ROW, and thus any negative shock originated outside of Ireland could be channeled through their balance sheets (Table 3). Third, the ROW possesses a large amount of government bonds. A sharp withdrawal of non-residents from the bond market could generate an interest rate shock and cause distress in the Irish financial system. Fourth, the multinational companies play an important role in connecting the Irish economy with the ROW through their external financing (e.g., inter-company loans).

Table 3.

Ireland: Financial Network Matrix (Net Bilateral Position)

(Percent of GDP, as of 2015Q2)

article image
Sources: CBI calculations; and IMF staff calculation.

12. Exclusion of the ROW shows that the banking sector, NFCs, and OFIs are the three main sectors in the domestic financial network (Figure 6 and Table 4).13 The banking sector was the largest financial sector (financial assets at 169 percent of GDP), followed by other financial institutions (117 percent of GDP), and non-financial corporations (107 percent of GDP) in June 2015. As shown in Box 1, they are ones with more connections (high in-and out-degree) than any other sectors, excluding the ROW. Table 5 and 6 show that many domestic sectors had large financial claims on and obligations to these three sectors relative to their balance sheets. As such, shocks to these sectors can affect other sectors via these direct and indirect financial interlinkages.14

Figure 6.
Figure 6.

Ireland: Domestic Financial Network Map 1/

(As of 2015Q2)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI calculation; NodeXL; and IMF staff calculation.1/ The size of each vertex shows the size of total financial assets. The thickness of arrows depicts the volume of bilateral gross exposures from a creditor to a debtor; loops represent gross intra-sectoral claims.
Table 4.

Ireland: Domestic Financial Network (Gross Bilateral Position)

(Percent of GDP, as of 2015Q2)

article image
Sources: CBI calculations; and IMF staff calculation.
Table 5.

Ireland: Domestic Financial Network (Relative Claim Matrix)

(Percent of total assets of each sector, as of 2015Q2)

article image
Sources: CBI calculation; and IMF staff calculation.

13. The Irish banking sector is tightly connected with the real sectors of the Irish economy, but not with many of the other financial sectors. Sources of banks’ obligations were well diversified, with households playing the largest role (over 30 percent of bank liabilities and household assets), and destinations of banks’ claims were mainly the real sectors, including households and NFCs (Table 5 and 6). There is a tight bank-sovereign financial linkage in Ireland through banks’ government bond holdings and the government’s holdings of bank equities. Interestingly, the Irish funds industry and the banking system did not depend on each other for funding and asset allocation. Insurance companies and pension funds had a small exposure to banks (4 percent and 1 percent of GDP, respectively), and banks had no meaningful exposure to insurance companies and pension funds, as shown in Table 4. Meanwhile, inter-bank linkages amounted to 41 percent of GDP (red box in Table 4), and the bilateral exposures between the banking sector and OFIs amounted to be 34 percent and 14 percent of GDP (blue boxes in Table 4).

Table 6.

Ireland: Domestic Financial Network (Relative Obligation Matrix)

(Percent of total liabilities of each sector, as of 2015Q2)

article image
Sources: CBI calculation; and IMF staff calculation.

14. The link between other financial intermediaries and the domestic real sector reflects group financing transactions and as such is not a potential source of contagion. In Table 5, OFIs held 36 percent of their total assets in the form of NFC equity or debt and NFCs invested 14 percent of assets into OFIs. The connection between the two sectors, however, mainly reflects intra-company transactions between treasury companies and their parent multinational companies. OFIs had some exposures to households (19 percent of their total assets). They do not provide loans directly to households to any significant degree, but are largely connected through purchases of mortgage loan securitizations and loan sales from banks.

15. Non-financial corporate sector is a net borrower and the most indebted sector in Ireland. In 2015Q2, NFCs’ liabilities to domestic sectors were 153 percent of GDP (Table 4). Most of the claims were held by the non-financial corporate sector itself, OFIs, and banks (60 percent, 42 percent, and 25 percent of GDP, respectively).15 As mentioned in the previous paragraphs, the role of treasury companies within the group of multinational companies should be noted.

16. The household sector is the largest net lender in Ireland, holding a large amount of claims on insurance companies and pension funds. Households held sizable retirement savings in pension funds (33 percent of household financial assets) and banks’ deposits or debentures (32 percent of household financial assets). The claim on the insurance companies constituted 23 percent of household financial assets. Conversely, the largest claim on households was held by banks in the form of mortgages (47 percent of GDP and 62 percent of household liabilities), amounting to 31 percent of bank assets.

17. Irish domestic inter-sectoral financial linkages are less dense than those in Denmark at end-2013.16 The 2014 Denmark FSAP found that, with the data at end-2013, the interbank claims were the largest connection among domestic financial institutions (110 percent of GDP), and those between institutional investors (insurance companies and pension funds) and OFIs (including non-money market investment funds) were the second largest linkage (55 percent of GDP).17 Relative to these numbers, the interbank claims in Ireland amounted to 47 percent of GDP, and the bilateral exposures between institutional investors and OFIs and NMIFs in Ireland were only 14 percent of GDP. The bilateral linkages between domestic financial sectors and real sectors were also tighter in Denmark with banks’ loans to households (129 percent of GDP) and to NFCs (64 percent of GDP) at end-2013, while those in Ireland were 49 percent and 26 percent of GDP, respectively, in 2015Q2.

18. Calculating steady-state Markov chain probabilities with the relative claim matrix (Table 6) as a transition matrix, NFCs, domestic banks, and OFIs are identified as important destinations of funding in Ireland (Figure 7).18 The steady-state probabilities indicate that, if one euro circulates through the domestic financial network, 29 cents will flow through NFCs no matter which sector the circulation starts from, making it the primary systemic funding destination. The analysis also shows that 18 cents will flow through domestic banks and other financial institutions, making these sectors the second largest funding destinations. Thus, income and profit shocks to NFCs and domestic banks can affect upstream funding sources.

Figure 7.
Figure 7.

Ireland: Steady-State Destination and Source of Domestic Financial Flows

(Percent of funding, 2015Q2)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; and IMF staff calculation.

19. The same analysis with the relative obligation matrix (Table 7) as a transition matrix finds that domestic banks are the most important source of funding in the Irish economy, followed by OFIs and households (Figure 7). The steady-state probability vector suggests that if 1 euro of funding is traced back through the relative obligation matrix, 24 cents flows from domestic banks, 16 cents from OFIs, and 15 cents from households. Non-money market investment funds are an important source of funding internationally, but mainly affect the domestic economy through economic channels, such as employment and investments, rather than through financial links. As a result, liquidity shocks to domestic banks can significantly affect the funding positions of downstream destinations.

Table 7.

Denmark: Domestic Financial Network Matrix

(Gross Bilateral Position, as of 2013Q4)

article image
Source: IMF (2014) “Technical Note on Macroprudential Policy Framework in Denmark” prepared by Prachi Mishra.

20. Recognizing the importance of monitoring interconnections among and with sectors in the Irish financial system, the Central Bank of Ireland has been making efforts to fill data gaps in the banking and non-banking sector. The large exposures data, which is mainly on the asset side of balance sheets, is used to examine the bilateral interbank exposures of Irish banks (Hallissey, 2016). The interlinkages on the liability side of balance sheets will be monitored with a new data template that is being finalized by the EU Commission. Kenny and others (2015) study the interconnectedness between non-domestic banks and their Irish-domiciled counterparties in the credit default swap market using data on derivative markets under the European Market Infrastructure Regulation. The Central Bank collects security level granular data quarterly from banks and non-money market investment funds and somewhat less granular data from FVCs in line with ECB regulations. The Central Bank has, on its own initiative, extended security level reporting to MMFs on a monthly basis from December 2014 and has, in an internationally unique measure, extended the FVC reporting form in full to other Special Purpose Vehicles from 2015Q3. In addition, new detailed exposure data of insurance companies will be available in 2016 under the Solvency II reporting regime.

D. Conclusions and Policy Implications

21. The inter-sectoral network analysis with QFA data shows that the Irish economy is highly connected across its border. Despite the recent deleveraging, banks are still interconnected with the ROW. If banks do not have sufficient capital and liquidity buffers, they can act as shock amplifiers. Insurance companies and pension funds can propagate negative foreign shocks into Ireland because domestic households are exposed to non-residents through their savings in these financial sectors, which invest in foreign assets. In addition, the government and NFCs rely on funding from nonresidents, and remain vulnerable to a reversal of sentiment in the global financial market.

22. The banking sector, NFCs, and OFIs (largely connected to MNC activities) are important sources and destinations of funding within the domestic-domiciled financial network in Ireland. Income/profit or liquidity shocks in these sectors can originate negative spillovers to other sectors in the economy, unless they maintain sufficient capital or liquidity buffers.

23. Given the large volume of intra-sectoral exposures, it is important to continue to enhance the collection of granular bilateral exposure data within both the banking sector and non-bank financial institutions as well as across these sectors. Quarterly financial accounts data show bilateral exposures at the sectoral level only. Securities data is collected for banks, investment funds, MMFs, and FVCs, with the addition of SPVs from Q3 2015 and insurers from Q1 2016. However, granular bilateral exposure data for banks are currently available for asset exposures for the most part. There are a number of data initiatives underway within the Central Bank of Ireland to fill data gaps and enhance understanding of financial interconnectedness in Ireland. These are welcome but will require sufficient resources and strong inter-departmental collaboration.

Real Estate Market Analysis19

A. Introduction

25. Ireland has experienced a historic financial crisis with the bursting of a real estate boom during the global financial crisis (text figure). The real estate boom in the first decade of the 21st century was fueled by fast credit growth, funded by domestic bank loans and cross-border capital flows. Prices in both the residential real estate (RRE) and commercial real estate (CRE) sectors doubled and total banking assets tripled from 2000 to 2007. Neglecting the RRE or CRE booms and the associated rapid credit growth had disastrous consequences: debt overhang and deleveraging spirals threatened financial and macroeconomic stability. GDP growth was negative for two years in a row in 2008–09, and unemployment increased to 15.1 percent by 2012. House and CRE prices fell by about 50 and 70 percent from their peak in 2007, respectively, with banks facing significant losses on their real estate exposures. The authorities had to recapitalize the banking system in the amount of €64 billion (about 40 percent of GDP).

uA01fig03

Property Prices

(Index, 1995Q1=100)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; IPD; and OECD.

26. Ireland still faces the challenge of dealing with the stock of mortgage and CRE NPLs. The authorities have deployed various measures to accelerate the resolution of problem loans and especially mortgages, CRE loans, and loans to small and medium-sized enterprises.20 Nonetheless, the system still holds a large stock of NPLs, composed to a significant degree of long-overdue mortgages and CRE loans (text figure).

uA01fig04

Non-Performing Loans

(Percent of gross loans)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: Central Bank of Ireland.

27. In addition, coming out of the crisis, real estate markets bounced back from their lows, prompting concerns for possible overvaluation and a build-up of new imbalances. RRE and CRE prices in 2015Q4 are now about 35 and 60 percent higher than the trough in 2013Q1, but still remain 33 and 48 percent below the peak levels, respectively.

28. Thus, developments in the real estate sector remain of central importance to macroeconomic and financial stability. Both CRE and REE have shown a proclivity towards price booms and busts. Construction activity is likewise highly cyclical. The funding of real estate purchases and construction has been a major part of banks’ business but also a source of major vulnerabilities.

29. Therefore, this note reviews recent developments in the two property markets, assesses certain systemic risk, and considers preemptive policy options to contain a potential build-up of imbalances. First, it summarizes price dynamics and transaction activities in each market, and discusses demand and supply factors driving the markets. It also looks at sources of funding and the extent to which banks are exposed to the two sectors. Second, it estimates the level of misalignment between actual and fundamental prices in both RRE and CRE markets, using a variety of non-parametric and parametric approaches. Third, it examines existing and potential policy options to address detected risk.

B. Recent Developments in Real Estate Markets

Residential Real Estate Market

30. National house prices have rebounded strongly since 2013. Based on the CSO data, nominal prices recorded the first increase in June 2013 (1.2 percent y-o-y) since February 2008, and growth rates have remained positive since then. In 2014, nominal house prices rose as fast as in the pre-crisis boom period (16 percent y-o-y) and continued to increase by 7 percent in 2015 (text figure). House price growth reached 16.3 percent at end-2014 (y-o-y, 3 month moving average), accompanied by a growth in housing transactions.

uA01fig05

Property Price Growth Rate

(Percent, year-on-year change)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; IPD; and OECD.

31. The rebound in the Irish RRE market is more vigorous than in other OECD countries (Figure 8). An international comparison reveals that Ireland has the most volatile RRE market: it had the largest boom and was also hit hardest during the crisis. Irish house prices are now returning from the post-crisis trough faster than in other OECD countries.21

Figure 8.
Figure 8.

OECD Countries: International Comparison of Real House Prices

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

32. This price growth appears to be driven by factors such as supply shortages. Housing construction is currently subdued, resulting in supply shortages, while it increased during the boom (Figure 9). The number of houses completed in 2015 was 12,666, well below the estimated 25,000 units required to meet household formation (CBI, 2015a; Duffy and others, 2014). Correlation between the annual house price growth and private credit growth was 0.67 before the crisis (1990–2007Q3), while the correlation after the crisis (2007Q4–2015Q3) dropped and reversed to -0.04. Mortgage loan approval has started to increase again since 2012H2, but its nominal amount is still below the pre-2000 level (Figure 9). Moreover, non-mortgage buyers play a significant role in the recent recovery, with about half of the transactions in cash (CBI, 2015b).

Figure 9.
Figure 9.

Ireland: House Completion and Permission, and Mortgage Loan Approval

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

33. In the post-crisis period, there was a significant divergence in prices and market conditions across different regions in Ireland. House prices in Dublin recovered earlier and about three times faster than in the rest of the country (Figure 10). The annual house price growth rate in Dublin reached 25 percent in August 2014, supported by strong rental growth in the face of limited new supply and weak construction activity. Expectations of further appreciation might also defer sales, contributing to housing shortages in Dublin.

Figure 10.
Figure 10.

Ireland: House Price Growth Rate by Regions

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CSO; and CEIC.

34. House price growth slowed to 6.6 percent (y-o-y) at end-2015, from 16 percent at end-2014. The slowdown in the Dublin market has been more noticeable, with the growth rate falling back to 2.6 percent at end-2015. Market expectations for further increases have also moderated. In the 2015Q3 Central Bank of Ireland survey of residential property price expectations, the percentage of respondents expecting prices to rise across 1 quarter, 1 year, and 3 year-time horizons were 46, 82, 93 percent, respectively, down from 90, 97, and 98 percent, respectively, in 2014Q3 (CBI, 2015b).

Commercial Real Estate Market

35. The CRE market has also been picking up strongly after plummeting from the pre-crisis boom. Based on the MSCI Investment Property Databank (IPD), capital values grew by 30 and 19 percent (y-o-y) at end-2014 and 2015Q4, respectively (Figure 11). Rental values have also been increasing rapidly at above 14 percent (y-o-y) for the last four quarters, and are 41 percent higher than the trough in 2013. Performance in the Dublin office sector has been the most robust, and is now spreading beyond the capital. Capital values in the CRE sector have increased by 25 percent y-o-y since 2014Q2, the fastest growth since 1999.

Figure 11.
Figure 11.

Ireland: Developments in the CRE Market1

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Note: 1/ Note that the data are a mix of annual and quarterly based on data availability. The last observation for Ireland, the U.K. and the U.S. are quarterly.

36. Strong returns have attracted investors in the search for yield as interest rates remain low. The Irish CRE market has been one of the best performing asset classes in Europe since 2014. Total returns for Irish CRE have outperformed other countries (Figure 11). Returns on 10-year Irish sovereign debt have trended to a twenty-year low and the spread between total returns for CRE and the long-term sovereign bonds reached approximately 8.5 percentage points.

37. The favorable economic outlook has provided a boost to the CRE market recovery. Investors have been encouraged by the broad-based improvement in domestic economic performance, which drove the rental growth in the office sector (text figure).

uA01fig06

CRE Capital Growth vs Real GDP Growth

(Percentage change y/y)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: MSCI/IPD; and Central Statistics Office.

38. The lack of new construction activity has contributed to the supply shortages (text figure). While activities are slowly picking up, the level of CRE stock under construction fell dramatically between 2008 and 2014, reflecting the low profitability margins in an environment of depressed prices (early in the period); construction firms’ stretched balance sheets and tighter lending standards by banks; and possibly other factors, such as the limited availability of suitable sites.

uA01fig07

Dublin Office Stock under Construction

(Millions of square meters)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: JLL.Note: It is originally sourced from Figure 4 of Duffy, David and H. Dwyer (2015). In the paper, data are presented as square feet but converted to square meters here for consistency.

39. The investor base has changed considerably since the crisis. Unlike the pre-crisis period, the majority of investment activity is now being funded through foreign investors and equity funds, such as real estate investment trusts (REITs) (text figure).22 Large foreign pension and insurance funds from the U.S., the U.K., and Germany have invested in the Irish CRE market, in part for balance sheet management, matching long-term liabilities with long-term assets. The attractiveness of Ireland as a leading destination for FDI has also increased demand for the CRE properties from multinational companies. CBI (2015a) notes that while only 2 percent of the value of CRE transactions was from foreign investors in 2006, this figure was over 40 percent in 2014.

uA01fig08

CRE Investment Turnover by Type of Investors

(Millions of euros)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: Central Bank of Ireland.

40. Total outstanding balance of CRE loans has been decreasing, while the flow of new lending to CRE has slowly picked up. Banks have reduced their exposure in the CRE market through the deleveraging process that occurred from 2009 to 2014. New bank financing to the CRE sector is slowly increasing, but new lending to CRE has been a minor component of total new lending volumes (Figure 12).

Figure 12.
Figure 12.

Ireland: Outstanding Balance of Private Credit and New Loan Composition

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

C. Empirical Analysis of Price Misalignment in Real Estate Markets

41. RRE and CRE prices in Ireland have been rising rapidly as shown in the previous sections due to various demand and supply factors, causing concerns about possible overvaluation and a build-up of new imbalances. We employ several non-parametric and parametric methods in order to estimate a long-term trend or equilibrium price level and compare it with actual prices to test whether the two sectors are overvalued.

Residential Real Estate Market

Non-parametric approach: price-to-income and price-to-rent ratios

42. In view of the recent recovery, common measures of RRE valuation based on deviation from long-run historical trends, such as price-to-income and price-to-rent ratios, suggest that the market is neither undervalued nor overvalued in 2015Q2. The two metrics show that there was an apparent overshooting in house prices before the crisis and a correction afterwards (text figure).23 However, caution should be taken before confirming this result with more sophisticated methods, because absolute level of valuation depends on the length of the period for which the historical mean is calculated.

uA01fig09

Price-to-Income and Price-to-Rent Ratio

(Percent, deviation from historical mean, 1990Q1-2015Q2)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: OECD; and IMF staff calculation.

43. An international comparison of the two metrics demonstrates that the deviation of Irish house prices from a long-term average is in the middle of the sample. Both ratios in Ireland have been historically more volatile than ones in comparator countries. However, when compared with 25 OECD countries, using the price-to-rent ratio, Ireland does not stand out as an over-or underpriced market as of 2015Q2, and, in fact, is close to the middle of the distribution (Figure 13). Similarly, based on the price-to-income ratio, the Irish RRE market does not appear to be overvalued.

Figure 13.
Figure 13.

OECD Countries: Price-to-Income and Price-to-Rent Ratios

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

44. It is worth focusing on RRE price developments since the introduction of macroprudential measures. The Central Bank of Ireland imposed the proportionate limits on loan-to-value (LTV) and loan-to-income (LTI) ratios in February 2015. The aims were: (i) to strengthen the resilience of lenders and borrowers to financial shocks; and (ii) to reduce the risk of future credit and RRE price spirals.24 It is still too early to properly evaluate the effectiveness of the measures. There is some evidence that these tools have had an effect in reducing price pressures, following very strong growth in 2013 and 2014. Moreover, market expectations for future house price increases have also moderated (expectation channel).25 The 2015Q3 survey of RRE price expectations shows that the percentage of respondents expecting prices to rise across 1 quarter, 1 year, and 3 year-time horizons were 46, 82, 93 percent, respectively, down from 90, 97, and 98 percent, respectively, in 2014Q3 (CBI, 2015a). House price growth rate had moderated to 6.6 percent (y-o-y) at end-2015, from 16 percent at end-2014.

Parametric methods: statistical filters, error correction models, and Markov regime switching model26

45. A more analytical way of looking at the RRE valuation is to calculate the gap between the actual house prices and their predicted fundamental (equilibrium) level based on econometric models. For this purpose, a broad range of models are used: (i) statistical filters; (ii) time-series approaches using economic fundamentals as explanatory variables (e.g., a vector error correction model (VECM) and an OLS regression with a pseudo error correction term); and (iii) a Markov regime switching model. It should be emphasized, however, that estimating equilibrium levels of house prices is still challenging and can be imprecise. Therefore, results should be interpreted with caution.

46. Results with statistical filters suggest that house prices are somewhat undervalued by 9–10 percent (text figure). One can assess where house prices are in a cycle without taking a view on whether the trend is driven by macro-financial factors. Using either one-sided or two-sided HP filter, Irish house prices are estimated to be about 9 percent below the trend.27 Isolating a component of house prices that lies within an 8–30 year interval, longer than a business cycle, a band-pass filer suggests an undervaluation of about 10 percent.28

uA01fig10

House Price Valuation with Statistical Filters

(Percent, deviation of actual prices from a trend)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: OECD; and IMF staff calculation.

47. Econometric models indicate that, as of 2015Q2, house prices were around the equilibrium level (Box 2). Specifically, using quarterly data between 1990Q1 and 2015Q2:

Summary of Econometric Model Specification

Several versions of vector or single error-correction models are used to estimate equilibrium levels of house prices on the period 1970Q1–2015Q2 (RRE market evaluation) and 1990Q1–2015Q3 (CRE market evaluation). Markov regime switching models with two or three states are estimated to identify boom-bust cycles in the two markets and find out where the current period stands in a new cycle. The data are compiled from various sources, such as the CBI, CSO, OECD, BIS, CEIC, and other sources, which are described in greater details in Appendix I. The data are deflated with GDP deflator or CPI index, and transformed into a long term.

• Vector error correction model:

Yt=∑pA*Yt−p+EtandΔYt=Π*Et−1+∑ip−1Γi*Yt−i+ωt,

where Yt is a k-vector of endogenous non-stationary I(1) variables and Et and ωt are vectors of white-noise innovations. Yt includes the log of the following variables: real house prices, real GDP per capita, real private credit, long-term interest rate (government bond yield), and real construction cost.

• Regression model with a pseudo error correcting term:

Following the approach in Igan and Loungani (2012), we set up

Δ⁢ln⁡hpt=α+γ*ln⁡PIRt−1+β1*Δ⁢ln⁡gdpt+β2*Δ⁢ln⁡credt+β3*ln⁡tt+β4*ln⁡cos⁡tt+ωt,

Where hp, PIR, gdp, cred, Int, cost denote real house prices, price-to-income ratio, real GDP per capita, real private credit, long-term interest rate (government bond yield), and real construction cost, respectively. ωt are white-noises. Besides real GDP, other real-term variables are deflated with CPI index.

• Markov regime switching model (MRSM):

Δlnhpt=αSt+γSt*lnPIRt−1+β1*Δlngdpt+β2*Δlncredt+β3*Intt+β4*Δlncostt+ωSt,εSt∼N(0,σSt2),St=1,2,…,kP=[p11⋯p11⋮⋱⋮pk1⋯pkk]

where k = 2 or 3 in the model specification and pij controls the probability of a switch from state j to state i. For example, for a model with two regimes, a parameter vector θ=(α1,.α2,.γ1,γ2,σ12,σ22,p11,p22) is estimated by maximum likelihood methods with Hamilton’s filter. Two parameters, a constant term αSt and the coefficient of price-to-rent ratio γSt are allowed to change across two regimes (S1 = high or S2 = low) or three regimes (S1 = high, S2 = normal, or S3 = low). Variance of the white noise term σt2 is allowed to change over states only for residential real estate prices to let the maximum likelihood estimation converge. Perlin (2015) provides further information of the MRSM algorithm used in this note.

  • Results from a VECM suggest that a negative house price gap opened up after the crisis but reduced to around -5 percent in 2015 (Figure 14). In the VAR and corresponding VECM, we use five endogenous variables: real house prices, real GDP per capita, real private credit, long-term interest rates (10-year government bond yield), and real construction costs. Because lag criteria and Johansen cointegration test support different model specifications, two specifications of the model are estimated and their results are averaged: one with three lags and two cointegration vectors and the other with two lags and three cointegration vectors. The estimated results show that Irish house prices have been volatile, moving away from the equilibrium level during the recent boom-bust cycle. Focusing on the recent period, the large negative house price gap that opened up right after the crisis has been closing quite rapidly to -5 percent; and

  • Second, an OLS regression with the price-to-income ratio as a pseudo error correction term shows no over-or undervaluation in house prices (Figure 14). In the model, similar to the one in Igan and Loungani (2012), house prices depend on household affordability (price-to-income ratio), GDP per capita, private credit, long-term interest rates, and construction costs. The choice of these variables stems from the desire to capture both demand and supply factors.29 The price-to-income ratio acts as a pseudo error correction term for an unsustainable deviation from the equilibrium level of house prices to normally stabilize by itself over the long run. This model specification yields R-squared of 36 percent and shows that the deviation from the equilibrium level is closed by 2015Q2. We also ran the model replacing the price-to-income ratio with an estimated error correction term from the long-run equation, yielding a very similar result and confirming the findings with the affordability indicator.

Figure 14.
Figure 14.

Ireland: Evaluation of RRE Prices with Econometric Models

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

48. Markov regime switching model (MRSM) analysis suggests that the probability of being in a high growth regime (boom) increased in 2013–14, but dropped at the beginning of 2015 (Figure 15). The Markov regime switching model allows three parameters of the OLS regression with the price-to-income ratio as a pseudo error correction term to change between two regimes (high or low), considering that economic agents would behave differently along the boom-bust cycle. In this MRSM, three parameters (the constant term, the coefficient of the error correction term, and the variance of the white noise) can switch between two regimes.

Figure 15.
Figure 15.

Ireland: Evaluation of House Prices with Markov Regime Switching Models

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

49. Results suggest that the MRSM specification appears to capture the dynamics of house prices well. Most parameters are statistically significant and with the expected sign and size. Especially, the error correction speed is slower in the “high” regime than in the “low” regime, as expected. The high regime has occurred for half of the sample period, and its expected duration is estimated to be four years (16 quarters). Two booms in 1998–2001 and 2002–07 stand out markedly, and one reappeared in 2014. The MRSM with three regimes (high, normal, or low) confirms the findings with two regimes as shown in the following figure. The probability of a high regime in both specifications started to decrease in 2015 (Figure 15).

Summary

50. Non-parametric and parametric analyses suggest that Irish REE prices are currently around equilibrium levels. Prices rebounded during 2013–14, but the pace started to slow at the beginning of 2015. Recent data show that the newly introduced macroprudential measures on mortgage loans may have started to have some impact via the expectation channel.30

Commercial Real Estate Market

Non-parametric method: price-to-rent ratio

51. The deviation from a long-term trend of the price-to-rent ratio suggests that the CRE sector was moderately overvalued as of 2015Q3 (text figure).31 The metric show that the CRE prices were also exuberant before the crisis, growing significantly above the rental yield. The adjustment after the crisis was higher than that in the RRE market. From 2014, the ratio breached the historical average again. As mentioned above, the absolute level of overvaluation depends on the choice of the period over which the historical average is calculated.

uA01fig11

CRE Price-to-Rent Ratio

(Percent, deviation from historical mean)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; and IMF staff calculation.
Parametric methods: statistical filters, error correction models, and Markov regime switching model 32

52. Results from HP and Band-pass filtering show that CRE prices are near the long-run statistical trends (text figure). Using either one-sided or two-sided HP filter, Irish CRE prices are estimated to be close to the trend. Isolating a component of house prices that lies within an 8–20 year interval, longer than a business cycle, a band-pass filer show that, as of 2015Q3, CRE prices were close to the equilibrium level in the range of +2 percent, while a frequency filter, which extracts components within an 8–25 years interval, indicates 8 percent of undervaluation.33

uA01fig12

CRE Price Valuation with Statistical Filters

(Percent, deviation of actual prices from a trend)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; and IMF staff calculation.

53. Econometric models indicate that CRE prices are marginally undervalued in the range of 5–8 percent (text figure).34 In the error correction models, similar to the one used in the previous section and O’Brien and Woods (2015), CRE prices depend on economic activity (GDP), credit availability, and long-term interest rates in the long-run equation, and the CRE price growth rate is explained by an estimated residual from the long-run equation (or the price-to-rent ratio), a change in interest rates, and the growth rate of GDP and credit.35 Using quarterly data between 1990Q1 and 2015Q3, the long-run equation yields a good R-squared of 74 percent, while R-squared for the short-run equation is 30 percent. A few other variables were tested as part of independent variables. For example, FDI inflows were included among the explanatory variables, but turned out to be unimportant (statistically insignificant). It is not surprising if one considers that foreign investors were mostly not major direct players in Irish CRE markets before the crisis, and FDI includes mainly investment outside the CRE sector.

uA01fig13

CRE Price Valuation with Error Correction Model

(Percent deviation of actual prices from the equilibrium level)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: CBI; CEIC; and IMF staff calculation.

54. Similar results were obtained using an alternative specification. In addition to an estimated residual from the long-run equation, the price-to-rent ratio is also used as a pseudo error-correction term in a model specification, because any exuberant growth of CRE prices beyond the rental value cannot be sustainable and would stabilize by itself over the long run. Results from the two specifications show that there was a large boom-bust cycle before and after the crisis, but the post-crisis “negative” deviation has been rapidly closing from 40 to 5–8 percent in a two-year horizon by 2015Q3.

55. Analyses with Markov regime switching models (MRSM) suggest that CRE markets entered into a high regime probability in the second half of 2013, which can be an early warning signal of another prolonged boom as shown in the last cycle.

  • Two regimes (high or low): The latest boom-bust cycle in the CRE market lasted for 20 years, which is longer than a normal business cycle (Figure 16).36 The cycle started around 1993 and ended at 2013. The boom period almost coincided with one in the RRE market. The estimated transition matrix shows that there is a long swing in the CRE market. That is, once the CRE market enters into a high regime, it tends to stay in the regime for a while: the expected duration of the high regime is estimated to be over nine years (33 quarters). The boom regime has occurred about 65 percent of the sample period 1990–2015, longer than the bust regime.

  • Three regimes (high, normal, or low): The MRSM with three regimes appears to capture dynamics of the CRE market better than one with two regimes. It detects a temporary slow-down period between two high growth periods during 1993–2007. It also hints a recent “pick-up” of CRE prices in recent years, which was an early warning signal of a prolonged boom in the last cycle (Figure 16).

Figure 16.
Figure 16.

Ireland: Evaluation of CRE Prices with Markov Regime Switching Models

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Summary

56. The mixed signals on the valuation of current CRE prices obtained from the non-parametric and parametric analyses point to the need for vigilance. While error correction models suggest a marginal undervaluation, the price-to-rent ratio and the MRSMs indicate an early warning of an incipient new boom period. Such mixed results are common but do point to the possibility of an emerging risk. Therefore, it is necessary to monitor the market and assess the need for action, close data gaps, review the available toolkit, and be ready for immediate policy actions.

D. Conclusion and Policy Implications

57. It would be important to rigorously evaluate the effectiveness and examine policy leakages of the two instruments with the new wave of loan-level data. We welcome the Central Bank of Ireland’s analytical framework to assess the effectiveness of the instruments in achieving their stated objectives. The new mortgage loan data will allow the Central Bank to test the framework and produce a comprehensive evaluation report. The authorities will also need to investigate if there has been any policy leakage or violation, for example where the provision of credit migrates from mortgage loans to unsecured consumer loans.

58. Close surveillance is needed to evaluate the early signals of a build-up of new imbalances in the commercial real estate markets. Even if different approaches send mixed signals regarding the overvaluation of CRE prices, it is undeniable that the recent CRE price growth rate has been high. Higher CRE asset valuations may help banks improve their balance sheets with higher recovery rates on the previously defaulted CRE loans. Yet, FDI inflows or equity funding like REITs can easily reverse if market sentiment changes, which could lead to a sharp drop in CRE values. Lenders with remaining exposures could face another hit from a collapse of collateral values via financial “decelerator” mechanisms, as observed in the post-crisis period. In addition, even if the flow of new CRE loans is a minor component of total new lending, it can continue to pick up. The ongoing intensive monitoring of CRE lending by the Central Bank of Ireland is hence welcome.

59. The authorities should enhance data collection and continue to allocate sufficient resources for CRE market analyses. The Central Bank of Ireland staff has made efforts to improve analyses on CRE market developments, which will need to continue with a support of sufficient resource allocation. In this regard, the mission welcomes the latest announcement that the Central Bank of Ireland and NAMA will co-fund the development of a CRE statistical system by 2018, which will be maintained by the CSO and give detailed information on sales and lease transactions, and construction activities, such as permissions, commencements, and completions.

Household Sector Analysis37

A. Introduction

60. The legacy of the financial crisis left behind household debt overhang in Ireland, but the process of balance sheet repair has begun. As the real economy experiences recovery and employment improves, the level of household debt has decreased in recent years. Due to the improving economy and ongoing resolution efforts, the overall number and volume of mortgage arrears have maintained a downward trend since 2013Q4.

61. Despite progress on arrears resolution, some segments are still vulnerable to domestic and external shocks. As of December 2015, fifteen percent of Primary Dwelling House (PDH) and 23 percent of Buy to Let (BTL) mortgages remained in negative equity, notwithstanding a recent recovery of house prices (Central Bank of Ireland, 2015a). A majority of mortgage loans is on either tracker rates or standard variable rates.38 External macro-financial risks are broadly on the downside with weak euro area growth and a possible reversal of the global search for yield. Stresses—triggered by a rapid increase in interest rates, a halt of recovery in the real economy and labor market, a sharp decline in real estate prices, or a combination thereof—can have a significant impact on the financial sector via intricate inter-sectoral financial interlinkages. Understanding how fragile the household sector would be against shocks is key to securing a robust recovery and financial stability in Ireland.

62. The rest of the paper is outlined as follows. Section II reviews the current state of household balance sheets using aggregate data. Section III assesses the vulnerability of the Irish household sector and its financial resilience to the FSAP stress test scenario, using loan-level data and the loan-loss forecasting model developed by the Central Bank of Ireland. Section IV concludes and considers policy options to address current and potential vulnerabilities.

B. Recent Developments in Household Indebtedness and Vulnerabilities

63. Households have deleveraged and debt sustainability improved (text figure). During 2008Q3–2015Q4, household financial assets increased by 20 percent from €294 billion to €355 billion, and household debt fell by 27 percent from €204 billion (the previous peak) to €150 billion. Unlike the pre-crisis period, the household sector became a net lender in 2009 and has kept this status since then (text figure).39

uA01fig14

Ireland: Household Debt

(Billions of euro)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: Central Bank of Ireland.

Households’ debt-to-disposable income ratio decreased from 215 percent in 2011Q1 to 155 percent in 2015Q4, and debt as a proportion of total assets decreased from 28 percent in 2012Q1 to 19 percent in 2015Q4 due to the deleveraging, as well as a rise in disposable income and total assets, supported by economic recovery and employment growth (Box 3).

uA01fig15

Household Financial Assets and Liabilities Transactions

(Billions of euros)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: CBI.

64. Two main components of household debt, mortgage loans and consumer credits, have declined from their peak in 2009Q1 (text figure). The growth rate (q-o-q) of mortgage loans, which account for almost 90 percent of total household loans, was minus 1.1 percent on average during 2009Q2–15Q3, and that of consumer loans was -3.3 percent in the same period.40 The deleveraging happened in both of the two subcomponents of mortgage loans, PDH and BTL loans, while the pace of negative growth in the PDH market has declined in recent quarters. New mortgage loan approval has started to increase again since 2012H2, but its nominal amount is still below the pre-2000 level.

uA01fig16

Composition of Household Loan

(Percent of total household loan, as of 2015Q3)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: CBI; and IMF staff calculation.

65. However, households are still highly indebted. The outstanding balance, however stood at €150 billion in 2015Q4 (155 percent of household disposable income and 70 percent of nominal GDP). As shown in a text figure, the Irish household debt-to-disposable income ratio is one of the highest in Europe (text figure). The NPL ratio has declined from the post-crisis peak (25.6 percent in 2014Q1), but remains at a high level (20.2 percent in 2015Q4).

uA01fig17

Household Debt-to-Disposable Income

(Percent of gross disposable income)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Sources: ECB; and Haver Analytics.

Ireland: Household Net Worth

After a large swing, household net worth started to increase again in recent years (text figure). It is still 13 percent lower than the previous peak at 2007Q2, but amounted to about three times nominal GDP as of 2015Q4. It declined by 60 percent during 2008–12Q2, but has increased by about 41 percent from the lowest level (€444 billion) in 2012Q2. The recent increase reflects not only the deleveraging process in the sector, but also a strong increase of housing asset values. The housing asset values have increased by more than 36 percent since 2013, as house prices strongly rebound.

uA01fig18

Dynamics of Household Balance Sheet Components

(Billions of euro)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: CBI.

66. A vast majority of mortgage loans (about 91 percent) have variable interest rates, and thus most households with mortgages are vulnerable to an increase in interest rates. As of 2015Q3, tracker loans and standard variable rate (SVR) loans accounted for 50.4 percent and 40.1 percent of total mortgage loans, respectively, while the share of loans with fixed rates over 1 year was only 7.5 percent (Table 8). Interest rates on existing tracker mortgages are current very low (about 1 percent), reflecting the ECB’s accommodative monetary stance. The average interest rate on SVR loans, where banks have more flexibility to set rates in line with market conditions, was 3.96 percent at 2015Q4 (CBI, 2016). If rates rise, interest payments will increase and thus some households will have difficulty servicing their debt.

Table 8.

Ireland: Mortgage Loans by Types of Interest Rates and Associated Interest Rates

Rate Types Market Share, Percent, as of 2015Q3

article image
Source: CBI Household Credit Market Report 2016H1.

67. A sizable share of mortgage loans remains in negative equity.41 10 percent of PDH loans were in negative equity and performing and 5 percent of PDH loans were in negative equity and non-performing at end-2015. On the other hand, for BTL loans, 12 percent were in negative equity and performing and 11 percent were in negative equity and non-performing in the period. Negative equity is a well-documented cause of default (Lydon and McCarthy, 2011; Kelly and O’Malley, 2016). Even if the share of new loans with high loan-to-value (LTV) ratios decreased significantly during the post-crisis period, the large share of mortgage loans originated with high LTV ratios before the crisis continue to be a cause of concern in Ireland.42

uA01fig19

Distribution of Originating Loan-to-Value Ratio

(Percent of new lending in each year)

Citation: IMF Staff Country Reports 2016, 317; 10.5089/9781475542257.002.A001

Source: CBI.

68. Arrears remain high in mortgage loans, and are increasingly prolonged despite a progress in restructuring distressed mortgages over recent years. The number and volume of mortgage arrears has fallen by about 36 percent and 35 percent since 2013Q3, respectively. Yet, the share of arrears in the greater-than 720 days-past-due category continues to rise, accounting for about 85 percent of total arrears as of 2015Q4.

69. The aggregate picture can mask large variations in financial soundness across households. The CBI (2015c) shows that 43 percent of Irish households do not hold any debt, and the debt burden faced by indebted households varies substantially across different segments of households. In particular, indebtedness is a particularly heavy burden for younger cohorts (34–45 age group). One needs to know which segments of households (e.g., the young, BTL borrowers, etc.) are vulnerable to shocks. The burden of the debt overhang is unevenly distributed across different groups of households, and thus detailed loan-level information of vulnerable segments would be useful to design and select targeted policy measures.

C. Simulation Analyses of Probabilities of Default in Household Sector

Methodology and Data

70. In this section, the Central Bank of Ireland’s internal Loan Loss Forecasting (LLF) models with loan-level data are used to assess vulnerabilities across the Irish household sector. These models have been in usage since 2011 both for stress tests and research purposes. The models combine an approach—jointly computing probability of default (PD), exposure at default (EAD), and loss given default (LGD)—where any set of house price, interest rate and unemployment rate scenarios can be applied to the population of mortgages at the Bank of Ireland (BOI), Allied Irish Bank (AIB), Educational Building Society (EBS, now merged with AIB but a distinct entity at the date of origination of many mortgages in the sample), and Permanent TSB (PTSB).

71. The sample amounts to 515,210 mortgage loans in Ireland, totaling €78 billion and 68.4 percent of total mortgage loans at end-2014.43 Table 9 reports the composition of the sample under study. 72 percent of the loans in the sample are outside Dublin; the vast majority are PDH loans (85 percent); Previous Owner mortgages represent 60 percent of the sample, as opposed to 40 percent for First-Time Buyer (FTB) mortgages; the most common age categories are between 35–44 and 45–55; 45 percent of loans have LTV of below 60, with over twenty percent being in negative equity; most loans have variable interest rates, with 51 percent on tracker contracts and 41 percent on SVR loans.

Table 9.

Ireland: Sample Composition of Residential Mortgages in December 2014

article image
Sources: IMF; and CBI in collaboration with the ECB.

72. The framework and mechanics behind the LLF are outlined in detail in Gaffney and others (2013) and Kelly and O’Malley (2016). For the purposes of the current analysis, only the PD module of the model is utilized. This module applies the coefficients of a Multi-State Model (MSM, see Jackson (2011) for methodological detail) to the following set of explanatory variables: bank, interest rate, current LTV ratio, regional unemployment, interest rate type, BTL status, and loan age (and its natural logarithm). The model then calculates a loan-level PD and Probability of Cure (PC) for each loan depending on their values as of December 2014 and the stress test scenario values for interest rates, unemployment and house prices, which affect the loan-level interest rate, regional unemployment and loan-level LTV ratio, respectively. The model calculates PD and PC over a three-year horizon from December 2014, as reported in all charts in this section.

Opening position: default stocks at December 2014.

73. There are two distinctive aspects to household financial vulnerability that can be measured with the data. Firstly, one can observe “backward looking” vulnerability by calculating the “opening stock” of non-paying loans across segments of the market, and secondly, one can observe “forward looking” vulnerability by predicting the probability of default over a three-year horizon for those loans still fully paying at end-2014. In all cases default rates are reported on a count basis rather than the balance-weighted default rates more familiar to banking and stress testing practitioners. This is because the current analysis is focused on vulnerability across households, rather than across monetary volumes of mortgages. Table 4 focuses on the “backward looking” measure of vulnerability by showing how defaults are distributed across the Irish mortgage market at end-2014.

74. The value in exploring the heterogeneity underlying the headline default stock in the Irish mortgage market is clear from Table 11. Many of these patterns have previously been highlighted in regression analyses by McCarthy (2014) and Kelly and O’Malley (2016).

  • Small differences in default propensity are uncovered across geography (where Dublin mortgages are less likely to default), FTB status (where previous-owners are more likely to default, in line with the findings of Kelly and others (2014)), and across the originating income distribution (where higher-income households are less likely to default, particularly when excluding BTL loans);

  • A much starker default differential is shown when comparing PDH to BTL loans, with the latter loans for investment purposes having a default stock that is double that in the owner-occupier segment (26.4 percent vs. 12.4 percent);

  • The age distribution also exhibits important differences, with default rates on existing loans increasing with age (excluding the 75+ category, which only accounts for 1 percent of mortgages, and where it is possible that borrowers have acted as guarantors for younger borrowers). Mortgages where the primary borrower is between 65 and 75 years of age have a default stock of 16.6 percent, while those under 35 years of age have default stock of 9.9 percent when BTL mortgages are excluded. This pattern partially reflects the relationship between borrower age and the housing and credit cycles: for those under 35, fifty percent have had their mortgage originated since 2010, by which point banks’ credit standards had tightened considerably, and housing values had already begun to collapse. Of those in the 65–75 category, on the other hand, 67 percent originated between 2003 and 2008, the years in which house prices were at their most overvalued, credit standards at their most liberal in Ireland (Kelly and others, 2015) and equity release or “top up” loans at their most prevalent;

  • The distribution of default across the LTV categories highlights stark differences in default propensity. In the PDH segment, loans with an LTV under 90 generally have default stocks between 7 and 11 percent. However, in the categories in negative equity, default stocks rise from 20 percent to 27 percent to 70 percent for loans between 100–120, 120–150 and 150+ LTV ratio; and

  • Finally, fixed rate loans are shown to have a far lower default propensity (4.6 percent) than variable rate mortgages (14 percent among SVR and 11.9 among tracker mortgages).

Table 10.

Ireland: Default stocks by Loan Group

(Percent, at end-2014)

article image
Source: CBI. All default rates are on an unweighted count basis.