Selected Issues

Abstract

Selected Issues

Credit Growth and Macroprudential Policies in the Slovak Republic1

Strong private sector credit growth has persisted for over a decade and resulted in household debt that is high relative to peers. Credit is now growing in riskier segments. Housing prices have also started to reflect pressures from strong credit growth. This paper assesses Slovakia’s household and private sector indebtedness against macroeconomic fundamentals, identifies key vulnerabilities from rapid household credit growth, assesses policy responses to date, and presents further policy options.

A. Background

1. Strong private sector credit growth has persisted in the Slovak Republic (Figure 1). In the aftermath of the global financial crisis, both mortgage and consumer loans to households have experienced double-digit growth. Household debt as a share of net disposable income is now higher than in most Central European (CE) countries. Leverage, as measured by household financial liabilities in percent of financial assets, has also soared past peers. While episodes of rapid credit growth may indeed imply increased access to credit and enable greater investment and economic growth (Levine, 2005), these episodes can also lead to vulnerabilities due to buildups of excessive leverage and asset price bubbles (IMF, 2012). Large buildups of household leverage have been shown to precede sharp drops in consumption, output, and employment during the subsequent deleveraging episodes (Mian and Sufi (2008), among others).

Figure 1.
Figure 1.

Slovak Republic: Recent Developments in Credit and Housing Markets

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

2. Credit growth in the Slovak Republic has been driven by both demand- and supply-side factors. On the demand side, Slovak households have historically been among the least indebted within the EU. Consequently, rapid lending growth could be viewed as a catch-up effect. Nevertheless, a sharp drop in lending rates, strong labor market dynamics, and a recovery in property prices also contributed to demand (Figure 1). On the supply side, two factors played a significant role. First, an extended period of accommodative monetary policy reduced net interest income and increased competition among banks in the Slovak market, putting pressures on banks to increase lending volume in an effort to stabilize profits. Second, favorable economic conditions underwrote continued improvement in the quality of the retail loan portfolio, affecting banks’ perceptions of credit risk. In the years following the crisis, a significant portion of new lending to households consisted of refinancing at lower interest rates thus limiting the pass-through to housing prices. More recently, housing prices, both existing and new dwellings, have crept upwards.

3. The rest of the paper is organized as follows. Section B describes assessments of household and private sector debt levels relative to fundamentals. Section C discusses vulnerabilities to households that result from the buildup of indebtedness. Section D describes the policy response to date and Section E assesses its effectiveness. Finally, Section F presents further policy options and concludes.

B. Household Indebtedness in the Slovak Republic: Excessive or Fundamentals-Driven?

4. Econometric analysis suggests that household indebtedness is growing at a faster pace than implied by economic fundamentals. We used a panel cointegration technique to identify any long-run relationship between the volume of retail loans and selected macroeconomic and financial variables, including interest rates and per capita income, for a set of 11 central and eastern European countries.2 If robustness tests did not reject the existence of such a relationship, regression analysis (fully modified and dynamic OLS) was used to estimate the long-run relationship between identified macroeconomic and financial variables and the stock of household loans. Estimates suggest that in Slovakia the stock of household loans is now above the level implied by economic fundamentals, with Slovakia showing the largest deviation among 11 countries in the sample. However, this approach has limitations given the small sample size.

5. To complement this analysis, we also assess overall private sector indebtedness against economic fundamentals. We use an autoregressive-distributed lag (ADL) model for a set of 36 European countries to establish a time-varying norm derived from an estimated long-term relationship between overall private sector debt (See Annex II for details). Estimates of the gap in real per capita private debt suggest that the current pace of credit growth is higher than the fundamentals-consistent growth, while the current level of real per capita private debt is lower than the fundamentals-consistent level. Results for Slovakia also reflect a significant degree of financial deepening since EU accession. An important caveat is that most countries in the sample have lower income levels than the Slovak Republic, and that interest rates may be depressed by the current ECB monetary policy stance, both factors could result in a somewhat elevated norm. Applying the standard Hodrick-Prescott (HP) filter to credit-to-GDP ratios suggests only a slight deviation from the long run trend. This exercise is subject to several limitations beyond the well-known shortcomings of the HP filter which constrain its direct use in data evaluation and policy prescription. For instance, it could lead to procyclical prudential policies because periods with large declines in GDP would suggest the presence of ‘excessive credit’ and would call for a tightening of lending standards and increases in capital requirements.

6. The credit cycle seems to have reached its post-crisis high. The credit cyclogram, compiled by the National Bank of Slovakia, is an aggregation of a set of core and supplementary variables evaluated against distributions of their own historical values to disentangle factors cyclical credit growth. According to this analysis, household indebtedness seems to be the strongest contributing factor. Growing household indebtedness increases the sensitivity of households to adverse macroeconomic shocks. The next section explores these vulnerabilities and some mitigating factors.3

A01ufig1

Credit Cyclogram

(Ratio)

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

Source: National Bank of Slovakia.

C. Vulnerabilities from Rapid Household Credit Growth and High Indebtedness

7. In addition to sustained strong growth, new lending has shifted toward riskier segments of household borrowers. The share of new loans with a loan-to-value (LTV) ratio higher than 80 percent rose from 42 percent at end-2014 to 49 percent at end-2016. During this period, and partly in response to the introduction of a binding limit on LTVs over 90 percent, the share of new loans with LTV ratios between 89 percent and 90 percent grew from 7 percent at end-2014 to 28 percent by end-March 2017. At the same time, four out of five new loans were granted to households with high debt-to-income ratios (8 and above).

8. Households have sought to lengthen mortgage maturities and lock-in historically low long-term interest rates to reduce their vulnerability to shocks. Since the crisis, the average period of interest rate fixation on mortgages has increased and the share of short-term mortgages has declined considerably. This should thus act as a buffer in the face of adverse shocks. Nevertheless, it is worth noting that though the average maturity for households has increased (26 years for new loans), the marginal distribution of interest-rate fixation period of borrowers is important for assessing vulnerabilities. A higher mass of borrowers at the margin of exposure to this shock could negatively affect bank portfolio quality and asset prices and thus affect both financial and real sectors.

9. A rise in the interest rate could lead to tighter credit conditions for firms and negatively affect labor productivity. This is illustrated with impulse response functions from a panel VAR estimation of a sample of EU countries (Figure 4). A Cholesky decomposition with the following ordering is used: interest rate shock -> credit growth -> labor productivity growth -> GDP growth. The results show that a shock to the effective interest rate would lead to a persistent decline in GDP growth through its effects on credit and labor productivity growth. Specifically, a 1.25 percentage point increase in effective interest rates could cause a 0.75 percentage point drop in GDP growth after one year. Simulations imply that the shock would take roughly 5 years to dissipate. Notably, labor productivity rises on impact –reduced lending and higher financing costs cause firms to shed workers and increase capacity utilization—before falling sharply during the adjustment. Credit growth also remains negative throughout the projection period which would imply private sector deleveraging.

Figure 2.
Figure 2.

Slovak Republic: Assessing the Level and Growth of Household Indebtedness

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

Figure 3.
Figure 3.

Slovak Republic: New Loan Quality

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

Figure 4.
Figure 4.

Slovak Republic: Impulse Response Functions to Macroeconomic Shocks from an Estimated pVAR

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

10. Bank profitability has been supported by strong credit growth, but at the cost of rising sensitivity to adverse shocks. The business model of the Slovak banking sector is largely dependent on generating net interest income from loans to residents. Half of net interest income is in turn generated from lending to the household sector. To maintain profitability, banks have increased the volume of lending to households to offset the negative effect of decreasing interest margins. Banks’ reliance on continued growth in the volume of loans to households to generate profitability for shareholders is increasing banks’ sensitivity to macroeconomic shocks, specifically to shocks that either reduce demand for new loans or increase credit risk.

11. A simulation of the potential impact of macroeconomic shocks on banks’ profitability shows that interest rate and credit risks represent the largest source of risk in bank portfolios.4 The simulation uses a balance sheet approach to show that significant banks’ sensitivities to factors such as credit and interest rate risks may increase by as much as 50 percent over a three-year horizon if credit growth continues unabated. For example, a fall of the average significant bank’s net interest margin by 0.1 percentage points in 2017 could cause net profit to drop by 9 percent; whereas the bank’s net profit could decrease by as much as 13 percent if the same change in net interest margin occurs in 2020 and credit growth continues apace. Similarly, an increase in the credit risk to cost ratio would imply markedly higher losses in 2020 than in 2017 under the same conditions for less significant banks. The interest rate risk on household loans is mitigated by the short average duration of interest rate fixation (3 to 5 years). Continued vigilance and extensive supervisory stress testing, as conducted by the SSM in 2017, are nevertheless warranted.

A01ufig2

Simulation of Banks’ Profits by Bank Size

(Return on equity, percent)

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

Source: National Bank of Slovakia.

D. Policy Response to Date

12. The NBS introduced borrower-based measures in 2014,5 prompted by a growing unease about risks and imbalances in the household credit market. There were three main imbalances that motivated the decision: (i) one in four new loans being granted had a high LTV (at the level or close to 100 percent); (ii) the prevalence of borrowers taking advantage of the low interest rate environment to increase their loan value rather than to decrease their debt service payments, often without proper income and credit verification; and (iii) lack of verification of borrowers’ ability to repay their mortgages in the event of an interest rate rise. These core measures were complemented by maturity limits, interest rate sensitivity tests, and a mandatory amortization schedule for annuities. In addition, standards for real estate appraisal, income verification, and lending via financial intermediaries were introduced or tightened. By 2016, imbalances in the housing sector and vulnerabilities among mortgage lenders were mounting. A cap on pre-payment penalties (1 percent as of March 2016) further exacerbated the pressure (Figure 5). At the same time, the stock of high LTV loans at or near the regulatory boundary value (90 percent) was rising.

Figure 5.
Figure 5.

Slovak Republic: Housing Market Developments, 2016–17

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

13. These developments prompted the NBS to embark on a first revision of macroprudential policy measures in 2016–17 (Table 1).6 The objective was to promote sound and sustainable credit growth, and address risks related to vulnerabilities in the residential real estate market. Measures were transposed into binding decrees for housing loans7 as well as consumer loans.8 The scope of application was extended to all lenders (not only banks). The debt-service-to-income ratio (DSTI) became binding, albeit at a lower level (80 percent). The share of loans with an LTV of 80 to 90 percent was restricted to 40 percent of the loan book.

Table 1.

Slovak Republic: Overview of Macroprudential Policy Measures

article image
Source: National Bank of Slovakia.

14. The risks from rising household indebtedness triggered a second round of revisions to macroprudential policies in 2018. As mentioned in the previous section, Slovakia currently has the highest level of household indebtedness among peer countries in the CEE region. Moreover, the current distribution of DTI indicates that a relatively large proportion of new loans exceed debt-to-income limits set in EU members such as Ireland, the United Kingdom, and Norway. The two proposed measures—the debt-to-income (DTI) limit of 8 and the limit of 20 percent on the share of loans with loan-to-value ratios exceeding 80—aim to bring Slovakia’s regulatory framework in line with that of peers, and reduce the risk of household indebtedness.

15. The role of the countercyclical capital buffer as a policy instrument is conceptually different from that of borrowed-based measures. First, they serve two different objectives. Borrowed-based measures only apply to new business and target actual lending practices (e.g. DSTI or LTV). In contrast, the countercyclical capital buffer builds general banking sector resilience during good times to ensure adequate loss-absorbing capacity in bad times. Second, they differ in their scope and degree of cyclicality. The countercyclical capital buffer has a broad scope because it covers all private debt, including households and non-financial corporations, and a cyclical component because the expectation is that it will be reduced in downturns to absorb credit losses. For these reasons, capital and borrower-based measures are not substitutes; and are considered complementary parts of a wider policy mix.

16. NBS decisions on the countercyclical capital buffer rate are guided by three leading indicators:

  • Private sector credit growth. A simple annual rate of private credit growth as a baseline indicator for discussions on excessive credit growth. Notwithstanding its simplicity, its clear advantage is the ease of interpretation, comparability, and widespread international use.

  • The credit gap. The NBS uses a national credit gap indicator based on the GDP trend. Its most important advantage compared to the standardized Basel formula is use of the GDP trend making the denominator more stable in both good times and bad times. Similarly, the buffer guide is calibrated based on the experience with excessive credit losses incurred by Slovak banking sector in 2009–10. Therefore, the coefficients in the buffer guide are modified as follows:
    ratet=(0.475xgapt0.95)
  • A composite indicator of financial cycle, the Cyclogram. It comprises several variables capturing not only trends and dynamics, but also stocks and levels to address the emerging character of domestic lending market (Rychtarik, 2014).

17. The NBS increased the countercyclical capital buffer rate to 0.5 percent for domestic exposures in July 2016. The countercyclical capital buffer was further increased to 1.25 percent in July 2017. While indicators of national credit gap indicator and Cyclogram provided strong quantitative guidance for these decisions, the NBS also considered the following elements: the growing level of private sector indebtedness, the loosening of lending standards, changes in average risk weights, the banking sector profitability outlook, and the general monetary policy environment. Both increases in the countercyclical capital buffer were widely anticipated by market participants following the issuance of forward guidance two quarters in advance. In both cases the decisions came into force 12 months after the decision was published in the NBS Official Journal. The total delay from trigger to implementation is close to 6 quarters, considering the lag resulting from data availability and the time necessary for legal procedures.

A01ufig3

Countercyclical Capital Buffer: Guides and Measures

(Percent)

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

Source; National Bönk of Slovakia

E. Effectiveness of the Policy Response

18. In practice, the ongoing evaluation of the effectiveness of macro-prudential policy is mainly based on detailed monitoring of a standard set of indicators of credit standards (LTV, DSTI, DTI, maturity). Other indicators such as rate of growth of housing or consumer loans, indicators related to the residential real estate market, or indicators of changes in credit quality in the retail loan portfolio of banks are closely monitored as well, although they are not direct policy objectives.

19. We conduct an event study to assess the impact of macroprudential measures. The event study follows the approach of Kuttner and Shim (2013). It shows that, for the most part, the impact of macroprudential policy measures was short-lived in both Slovakia and selected peer countries (Netherlands, Czech Republic, and Poland) following the global financial crisis. In Slovakia, for example, the impact tended to dissipate in the third quarter after the macroprudential measure came into effect. In Poland, the impact dissipated after one quarter. The event study highlights the shortcomings of the initial rounds of tightening of macro-prudential policy measures in Slovakia, given that the framework for borrower-based measures had to be built from the ground up and that macroprudential policy operated with a lag between 4–6 quarters in Slovakia (Figure 6). This meant that the reaction time in the years following the crisis was slower than desired because of macroprudential limits, in many cases, started from market averages and were slowly phased-in.9

Figure 6.
Figure 6.

Slovak Republic: Event Study of Housing Prices and Housing Credit

(T= Quarter of Tightening Episode)

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

20. We also conduct an empirical analysis broadening our comparator set of countries. Several new datasets—including Kuttner and Shim (2016) and Cerutti et al (2015)—eased the data constraint on analyzing the impact of non-interest rate policies on housing cycles and provided new insights into the most impactful policy levers for stabilizing housing markets. In addition, several studies underscored the impact of LTV ratios identified by Crowe et al (2011) and Cerutti et al (2015), both of which highlighted the key role of loan to value limits in curbing real estate booms. Kuttner and Shim (2016) also stress the importance of debt service to income ratios; they estimate that a policy action to reduce the maximum debt-service-to-income ratio could reduce real credit growth by as much as 4–7 percentage points over the subsequent 4 quarters. Interestingly, Kuttner and Shim (2016) find a significant, cumulative effect of housing-related taxation on housing prices when they limit their dataset to the 18 countries that they define as active users of macroprudential policies.

21. Our empirical analysis assesses the relationship between housing prices and mortgage credit growth, and non-interest rate policy instruments (Annex III). We extend Kuttner and Shim’s (2016)’s panel data set of 57 emerging and advanced economies from 1980: Q1 through 2016: Q4. The data on non-interest rate policy instruments includes both reserve requirements and macroprudential policy instruments. As control variables we use structural features of the mortgage industry (type of interest rate, term to maturity), and macroeconomic variables (real GDP growth, CPI growth and the policy interest rate). The results are discussed below (Table 2).

Table 2.

Slovak Republic: Impact of Non-Interest Rate Policies on Housing Prices and Housing Credit Growth1

article image

A *** next to a coefficient signifies significance at the 1 percent level; ** at the 5 percent level; and * at the 10 percent level. See Annex III for regression specification and definitions.

See Annex III for variable definitions and discussion of robustness tests.

  • Non-interest rate policies and housing prices. Among macroprudential instruments, the maximum loan-to-value ratios and limits on exposure to residential real estate have a statistically significant impact on housing price appreciation. DSTI ratios, risk weights, and provisioning do not have a statistically significant effect on prices. We also find the effect of housing-related taxation to be statistically insignificant in our augmented dataset, even when the sample size is restricted to the 18 countries defined as active users of macroeconomic policy by Kuttner and Shim (2016).

  • Non-interest rate policies and mortgage credit growth. Of the set of non-interest rate policies evaluated, caps on debt-service-to-income ratios appear to be the most effective policy tool for curbing mortgage credit growth. DSTI ratios are unchanged by housing price increases, which means that their effect is undiluted by housing price appreciation during real estate booms. As is the case with housing prices, our findings are consistent with previous studies that find changes in the maximum LTV ratio and limits on exposures also have a negative impact on housing credit growth. However, changes in reserve requirements have no statistically significant impact on housing credit growth since they do not affect borrowers directly.

22. We also undertake an impact analysis of the recently approved revisions to macroprudential policy. Using detailed household data from EU SILC (2016) for a sample of 5,738 representative households, we estimate the maximum potential household indebtedness based on both demand and supply factors. The simulation suggests that after tightening DSTI and LTV limits, households’ maximum indebtedness would still increase with further growth in household income or declines in interest rates, but the rate of growth will converge to the rate of growth in household income over the long-term. In addition, while the DSTI limit in Slovakia (80 percent) is higher than the average among peer countries such as Estonia, Latvia, Hungary and Slovenia (50 percent), our calculations reveal that the calibration applied in Slovakia is at least as strict as in other countries for households earning median income, and even stricter for those with lower incomes but less binding for households with higher incomes. Therefore, the proposed debt-to-income limit (DTI) would be complementary since it would mainly curb the risk of high indebtedness among higher income households.

Table 3.

Slovak Republic: Impact on the Change in the Maximum Potential Household Indebtedness

article image
Sources: National Bank of Slovakia; EU SILC; and Household Finance and Consumption Survey.

23. The results suggest that these borrower-based measures do not represent a significant one-off impact on the market and that, in the medium-term, the accumulation of other risks would be significantly reduced. Indeed, the importance of the DTI limit for the mitigation of the risk of household credit growth is mainly applicable in a scenario where interest rates fall even lower and income growth is linked to the rising risk of overheating in the labor market and the economy. Under such a scenario, the implicit DTI limit will be less binding for an increasing number of households.

24. Our impact analysis also shows that the pace of credit growth gradually decreases with the implementation of measures. We assume that the majority of loans affected by the tightened limits would still be granted, albeit at a lower volume. The tightening of both the DTI and LTV limits might cause a drop household credit growth of about 0.5–1.4 p.p., and a decrease in the volume of new loans approximately by 8 percent each. This estimation however depends on some assumption, notably (i) the overlap between groups of clients affected by these limits, (ii) the proportion of clients that will give up the loan application if full requested amount cannot be granted and (iii) the proportion of clients who will partially replace the reduced loan volume due to the tightening of the LTV limits by consumer credit financing. Under current regulations, approximately a third of the debt subject to the stricter LTV limits can instead be financed through consumer credit, while maintaining the level of monthly repayments.

25. As the preceding analysis shows, the authorities’ current and planned macro-prudential measures are slowing down household credit growth at the margin, and supporting the aim of keeping household debt in line with incomes and debt servicing capacity.

  • The event study highlights the need to take a longer-term view to assess the full impact of measures, but also underscores the potential for impacts to dissipate thus necessitating ongoing vigilance.

  • The empirical analysis shows that in Slovakia, the DSTI measure is appropriately calibrated for median households. However, the proposed DTI limit would complement the existing macroprudential framework by discouraging high indebtedness among higher income households for whom the DSTI limit is less binding.

  • While the share of loans with high LTV ratio is declining, the impact analysis makes a case that further tightening of both LTV and DTI limits would be effective in mitigating the build-up of imbalances with no significant downside in terms of credit market development.

The analysis underscores the importance of proceeding with the second revision of macroprudential policies as planned. While merited, the proposed introduction of LTV and DTI limits might not be enough to avoid an unsustainable run-up in household credit. Therefore, the NBS has communicated the possibility of further tightening of macroeconomic policy, if warranted. However, it is also important to consider what more can be done beyond tightening macroprudential policy.

F. What More Can Be Done?

26. Fiscal policies could perhaps play a complementary role in amplifying the effects of macroprudential policies on housing price and mortgage credit growth. Several studies suggest that property taxes can be used as an effective tool to dampen house price volatility and to curb excessive mortgage credit growth. Wolswijk (2005) noted that greater fiscal subsidization was associated with higher mortgage credit to GDP ratios in EU countries. Using data on effective property tax rates in the United States, Poghosyan (2016) finds that an increase of 0.5 percent in property tax rates can reduce housing price volatility by 0.5–5.5 percent. These findings suggest that increasing recurrent property taxation and reducing mortgage interest deductibility could mitigate incentives for debt-financed home ownership. Most recently, Fatica and Prammer (2017) find that tax benefits in European Union countries reduce the cost to homeowners of housing capital by nearly 40 percent on average compared to the efficient level under neutral taxation.

27. Compared to peers, the role of housing-related taxation in curbing housing prices and mortgage credit growth in Slovakia could be significantly expanded. Property taxes in Slovakia are among the lowest in the European Union; property taxes account for 1.4 percent of total tax revenue compared to 6.8 percent on average in the EU. As a percentage of GDP, property taxes are the lowest of any EU country. The effective tax rate is just 0.10 percent, among the lowest in the EU, one-third of the EU average and one-tenth that of France.

28. The tax treatment of owner-occupied housing in Slovakia distorts the incentives of home ownership. Compared to the tax-neutral benchmark, in Slovakia the tax on owner-occupied housing is just 58 percent of what the tax-neutral benchmark would suggest. In other words, the favorable tax treatment of owner-occupied housing in Slovakia reduces the user cost of housing capital by one percent of the house value per year, on average, compared to the tax-neutral treatment. Poterba (1995) suggests that to achieve real estate market equilibrium the cost of owning and maintaining your own house should be equal to the cost of renting a comparable property. A tax treatment substantially below the tax-neutral benchmark—such as the one in Slovakia—acts as a subsidy distorts the market by pushing the cost of owner-occupied housing below its equilibrium level. In Slovakia, the average tax subsidy is comprised of a tax exemption of imputed rent (23 percent of the tax-neutral benchmark) and untaxed capital gains (18 percent of the tax-neutral benchmark) (Figure 7). Since January 1, 2018 young borrowers also benefit from a mortgage interest deduction, not yet reflected in the data.10

Figure 7.
Figure 7.

Slovak Republic: Tax Treatment of Owner-Occupied Housing

Citation: IMF Staff Country Reports 2018, 242; 10.5089/9781484370940.002.A001

29. Increasing housing-related taxes could help dampen demand for household credit and provide additional revenue for the budget. As a first step, the average subsidy on untaxed capital gains could be reduced by about one-third or 6 percent of the tax-neutral benchmark, to bring it in line with the EU average. Slovakia’s tax rate on capital gains is currently 19 percent, compared to the EU top rate of 35 percent (for Malta), and main residences are exempt from capital gains after tenure exceeds 2 years. As a second step, reducing the tax exemption on imputed rent would likely require a significant increase in the effective tax rate on property over the long-run. This would require a combination of phased-in increases in recurrent property tax rates as well as enhanced tax administration efforts. Increasing property tax revenue to the EU average could yield up to 1.2 percent of GDP annually—and would represent an important contribution to the budget.

References

  • Abrigo and Love, 2015, Estimation of Panel Vector Autoregression in Stata: a Package of Programs. University of Hawaii working paper.

  • Borio, C., Drehmann, M., Gambacorta, L., Jiménez, G., & Trucharte, C. (2010). Countercyclical capital buffers; exploring options (Vol. 317). BIS Working Paper.

    • Search Google Scholar
    • Export Citation
  • Drehmann, M., Borio, C. E., & Tsatsaronis, K. (2011). Anchoring countercyclical capital buffers: the role of credit aggregates (Vol. 355). BIS Working Papers

    • Search Google Scholar
    • Export Citation
  • Cerutti, Eugenio, Stijn Claessens, and Luc Laeven, (2017). The use and effectiveness of macroprudential policies: New evidence, Journal of Financial Stability, 28, (C), 203224.

    • Search Google Scholar
    • Export Citation
  • Kuttner, Kenneth N. & Shim, Ilhyock, 2013. Can Non-Interest Rate Policies Stabilize Housing Markets? Evidence from a Panel of 57 Economies. Kenneth N. Kuttner, Ilhyock Shim. NBER Working Paper No. 19723. Issued in December 2013.

    • Search Google Scholar
    • Export Citation
  • Kuttner, Kenneth N. & Shim, Ilhyock, 2016. “Can non-interest rate policies stabilize housing markets? Evidence from a panel of 57 economies,” Journal of Financial Stability, Elsevier, vol. 26 (C), pages 3144.

    • Search Google Scholar
    • Export Citation
  • Levine, Ross, 2005. “Finance and Growth: Theory and Evidence” in Handbook of Economic Growth, Volume 1A. Edited by Philippe Aghion and Steven N. Durlauf. Elsevier Press.

    • Search Google Scholar
    • Export Citation
  • Mian, Atif and Amir Sufi, 2008. “Summary of “the consequences of mortgage credit expansion”,” Proceedings, Federal Reserve Bank of Chicago, issue May, pages 129132.

    • Search Google Scholar
    • Export Citation
  • Národná Banka Slovenska (2014-2017): Financial Stability Reports

  • Rychtárik, S. (2014). Analytical background for the counter-cyclical capital buffer decisions in Slovakia. Biatec, 22 (4), 1015.

  • Rychtárik, S. Krcmár, M. [2013] Dynamics of housing affordability in Slovakia and its regions on the background of macroeconomic development Volume 21, 9/2013.

    • Search Google Scholar
    • Export Citation
  • Vandenbussche, Jerome, Vogel, Ursula, & Detragiache, Enrica. 2012. Macroprudential Policies and Housing Prices—A New Database and Empirical Evidence for Central, Eastern and Southeastern Europe. Working Paper 12/303. International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Shim, Ilhyock, Bogdanova, Bilyana, Shek, Jimmy, & Subelyte, Agne. 2013. Database for Policy Actions on Housing Markets. BIS Quarterly Review, September 2013, 8395.

    • Search Google Scholar
    • Export Citation

Annex I. Estimation of the Retail Credit Level Consistent with Fundamentals1

1. The relationship between retail debt and macroeconomic fundamentals is estimated using panel cointegration model. The model has the following form:

lnLit=αi+ΣjβjXijt+εit(1)
  • Lit : stock of retail loans,

  • Xijt : macro-financial variables,

  • i : country index, t – time index, j – index of variable.

The vector of macro-financial variables includes a combination of the following variables:

  • the volume of GDP (in natural logarithms),

  • income (in natural logarithms),

  • retail interest margin,

  • employment rate,

  • unemployment rate,

  • net effective exchange rate,

  • property price index (in natural logarithms).

The analysis was performed on eleven countries: Slovakia, the Czech Republic, Poland, Hungary, Slovenia, Croatia, Estonia, Latvia, Lithuania, Romania and Bulgaria. The analysis used nominal variables, since it was their impact on the nominal stock of loans which was examined. The time series (with quarterly frequency) were adjusted for seasonal variations.

2. Respective equations including different combinations of the macro-financial variables were estimated fully modified ordinary least squares (FMOLS) method and the dynamic ordinary least squares (DOLS) method. Based on the estimation, six specifications of cointegrating equations are selected, using the following combination of explanatory variables:

  • income, interest margin and unemployment rate;

  • GDP, income and unemployment;

  • GDP, interest margin and unemployment;

  • property prices, GDP and employment rate;

  • property prices, GDP, income and unemployment rate;

  • property prices, income, interest margin and unemployment.

All specifications were first tested using panel cointegration tests to confirm that a long-run relationship indeed exists between lending volumes and a possible combination of explanatory variables. As all equations are estimated using both the method (DOLS and FMOLS), it gives in total 12 possible cointegrating equations.

Annex II. Estimation of the Private Sector Debt Level Consistent with Fundamentals

1. The relationship between private sector debt and its main determinants is estimated using an Arellano-Bond specification on a panel dataset. The specification was chosen to address endogeneity issues in the dynamic panel structure which render fixed effects estimators inconsistent. A broad sample of 36 European countries during 1995–2016 is used under the admittedly strong assumption that all countries share the same long-term elasticities with respect to fundamentals.

2. The relationship between private sector debt and its main determinants is cast as a single equation, autoregressive-distributed lag (ADL) model. The latter can be interpreted as a stylized, reduced-form, demand and supply system expressed in semi-loglinear form:

lnDitPit=αi+Σj=12βjlnDitjPitj+Σj=01γjlnYitjPitj+Σj=01δjRitj+εi,t(2)
  • DtPt : Per capita private sector debt stock in thousands of 2005 purchasing-power-parity U.S. dollars (see note to Figure 7 for details);

  • YtPt : Per capita GDP in thousands of 2005 purchasing-power-parity U.S. dollars (source: IMF’s World Economic Outlook database), used as a measure of debt-servicing capacity that affects positively (+) both the demand and supply of credit;

  • Rt : nominal interest rate on private sector debt (fraction)1, which has opposite effects on demand (-) and supply (+);

  • i : country index, t – time index.

The demand-side effect of changes in interest rates is expected to dominate the supply-side impact in the reduced-form equation, in line with the findings in the existing literature (Cottarelli, Dell’Ariccia, and Vladkova-Hollar, 2003; Schadler and others, 2005; Iossifov and Khamis, 2009). Lack of data on private sector net worth for CESEE countries outside the EU prevents us from including that variable in the regional regressions.

3. The long-run relationship between private sector debt and its main determinants is then given by the long-run solution of the ADL model, under the stability condition (0<Σj=12βj<1)

dit*=α1Σj=12βj+Σj=01γj1Σj=12βjγit*+Σj=01δj1Σj=12βjRit*,where(3)
  • lowercase variables are expressed in natural logarithm of per capita quantities in thousands of 2005 purchasing power parity U.S. dollars; asterisk indicates long-run value;

All variables entering equation (2) are assumed to be either (trend) stationary or integrated of order one and cointegrated (that is, there is a linear combination of the variables in levels that is stationary). In the latter case, the long-run coefficients inferred from the short-run regression specification lie in the cointegration space of the dependent and explanatory variables (Hendry, 1995).

The equilibrium-correction (EC) model isomorphic to the reduced-form demand and supply system (1) is then given by:

Δdit=γ0Δyit+δ0ΔRit(1Σj=12βj)(dt1dit1*)(4)

Annex III. Estimation of the Impact of Non-Interest Rate Policies on Housing Prices and Housing Credit Growth

1. To estimate the relationship between non-interest rate policies and housing prices or housing credit growth, we use simple ordinary least squares with fixed effects and robust standard errors. We estimate two equations, one for each dependent variable, that take the following forms:

lnPit=αi+β1lnPit1+ΣjβjMijt+ΣjβjXijt+ΣjβjXijt1+εit(5)
  • Pit : housing prices,

  • Mit : non-interest rate macroprudential policies,

  • Xit : macro-financial variables,

  • i : country index; t : time index; j : variable index.

lnCit=αi+β1lnCit1+ΣjβjMijt+ΣjβjXijt+ΣjβjXijt1+εit(6)
  • Cit : housing credit growth,

  • Mit : non-interest rate macroprudential policies,

  • Xit : macro-financial variables,

  • i : country index; t : time index; j : variable index.

The panel dataset includes observations for 57 emerging and advanced economies spanning the period from the first quarter of 1980 through the fourth quarter of 2014. The set of macro-financial control variables is derived from the literature and comprises:

  • Interest type on mortgage loans

  • Policy rate

  • Harmonized CPI

  • Real GDP growth

The independent variables capture non-interest rate macroprudential policies compiled from a variety of sources, notably ESRB and BIS databases, and databases from IMF working papers:

  • LTV ratio

  • DSTI ratio

  • Risk-weights

  • Provisioning

  • Exposure limits

As a robustness check, we ran a Hausman test to determine whether the fixed effects specification was appropriate. We compare random and fixed effects models and test whether the unique errors (ui) are correlated with the regressors, the null hypothesis is they are not. The p-value was significant (0.032) so fixed effects was deemed the appropriate model in this case. To address possible issues of non-stationery time series data, we take the first differences of those variables (see Table 1).

Table 1.

Variable Definitions and Sources

article image

1

Prepared by Olamide Harrison, Pavol Jurca, Štefan Rychtárik, and Irene Yackovlev.

2

More countries could not be included due to lack of disaggregated data on household debt, see Annex I for details.

3

Private sector credit growth in the Slovak Republic has been high and sustained, but lower than seen in previous credit boom episodes. Deviations from a backward-looking rolling cubic trend and ad-hoc thresholds suggest that the current episode of private sector credit growth does not meet all the technical criteria to be classified as a credit boom. As defined in the IMF’s 2012 Staff Discussion Note on Credit Booms (SDN/12/06), an episode of credit growth is defined as a boom if the deviation of credit-to-GDP from the estimated trend is greater than 1.5 times its standard deviation and the growth rate of credit-to-GDP exceeds 10 percent; or the growth rate exceeds 20 percent.

4

The analysis is described in detail in FSR 11/2017 (p. 32-35).

5

Recommendation No 1/2014 of NBS of 7 October 2014 in the area of macroprudential policy on risks related to market developments in retail lending: full text and comprehensive summary.

6

The revision of the measures is explained in FSR 05/2017 (p. 44-45; for housing loans) and FSR 11/2017 (p. 45-46; for consumer loans).

7

Decree No 10/2016 of Národná banka Slovenska of 13 December 2016 laying down detailed provisions on the assessment of borrowers’ ability to repay housing loans: full text and comprehensive summary. The revision of the housing loan measures is explained in FSR 05/2017 (p. 44-45).

8

Decree No 10/2017 of Národná banka Slovenska of 14 November 2017 laying down detailed provisions on the assessment of borrowers’ ability to repay consumer loans: full text (in SK only) and comprehensive summary. The revision of the consumer loan measures is explained in FSR 11/2017 (p. 45-46).

9

It is possible that other factors –such as frontloading before implementation, legislative changes, and decreasing interest rates –also had a dampening effect yet are not captured by the event study methodology.

10

Since January 1, 2018, young borrowers up to 35 years of age with incomes below 1.3 times the average can deduct up to 50 percent of their mortgage-related interest expenses from their tax liability for the first five years after purchasing their home. The deduction is capped at €400 per year, and the mortgage volume is capped at €50,000. This replaced a previous interest rate subsidy for young borrowers.

1

The authors are thankful to Ján Klacso who has done this part of econometric analysis.

1

For EU countries, the implicit interest rate is calculated using sectoral accounts data as the ratio of interest payments (including financial intermediation services indirectly measured) over the average of the beginning and end-period combined stock of debt of firms and households. For other countries, data are mostly for the lending rate, published in the IMF’s International Financial Statistics database, with gaps in country coverage filled with data for the short-term interest rate published in the OECD’s Economic Outlook database and from national data sources.

References

  • Allen, J. and van der Velder, R. (2001), ‘Educational mismatches versus skill mismatches-effects on wages, job satisfaction and on-the-job search’, Oxford Economic Papers 53 (3), 434452.

    • Search Google Scholar
    • Export Citation
  • Andrews, D., A. Caldera Sánchez and Å. Johansson (2011), Housing Markets and Structural Policies in OECD Countries, OECD Economics Department Working Papers, No. 836.

    • Search Google Scholar
    • Export Citation
  • Braconier, H., G. Nicoletti and B. Westmore (2014), Policy Challenges for the next 50 years, OECD Economics Department Policy Papers, No. 9.

    • Search Google Scholar
    • Export Citation
  • Brunello, G., P. Garibaldi and E. Wasmer (2007), Education and Training in Europe, Oxford University Press.

  • Desjardins, R. and K. Rubenson (2011), An Analysis of skill mismatch using direct measures of skills, OECD Education Working Papers, No. 63.

    • Search Google Scholar
    • Export Citation
  • Estevão, M. and E. Tsounta: 2013. “Has the Great Recession Raised U.S. Structural Unemployment?’. IMF working paper (WP/11/105).

  • Forth, J. and G. Mason (2006), “Do ICT Skill Shortages Hamper Firms’ Performance? Evidence from UK Benchmarking Surveys”, National Institute of Economic and Social Research Discussion Papers, No. 281.

    • Search Google Scholar
    • Export Citation
  • Hartog, J. (2000), “Overeducation and Earnings: Where are We, Where We Should Go?”, Economics of Education Review, Vol. 19, pp. 131147.

    • Search Google Scholar
    • Export Citation
  • Kampelmann, S. and Rycx, F. (2012), ‘The impact of educational mismatch on firm productivity: Evidence from linked panel data’, Economics of Education Review 31 (6), 918931.

    • Search Google Scholar
    • Export Citation
  • McGowan, M. and Andrews, D. (2015), ‘Labour Market Mismatch and Labour Productivity’, OECD Economics Department Working Papers (1209).

    • Search Google Scholar
    • Export Citation
  • Montt, G. (2015), “The Causes and Consequences of Field-of-study Mismatch: An Analysis Using PIAAC”, OECD Social, Employment and Migration Working Papers, No. 167, OECD Publishing, Paris, http://dx.doi.org/10.1787/5jrxm4dhv9r2-en.

    • Search Google Scholar
    • Export Citation
  • OECD (2016), Getting Skills Right: Assessing and Anticipating Changing Skill Needs, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264252073-en.

    • Search Google Scholar
    • Export Citation
  • OECD, 2017, “Getting Skills Right: Skills for Jobs Indicators”, OECD Publishing, Paris.

  • Pica, G. and J. Rodriguez Mora (2005), Who is Afraid of a Globalised World? FDI and the Allocation of Talent, CSEF Working Papers, No. 184.

    • Search Google Scholar
    • Export Citation
  • Podpiera, J., Raei, F., and A., Stepanyan, (2017), “A Fresh Look at Potential Output in Central, Eastern, and Southeastern European Countries”, International Monetary Fund, Working Paper WP/17/37, Washington DC.

    • Search Google Scholar
    • Export Citation
  • Quintini, G. (2011), “Right for the Job: Over-qualified or Under-skilled?”, OECD Social, Employment and Migration Working Papers (120).

    • Search Google Scholar
    • Export Citation
  • Van der Vlist, A.J., C. Gorter, P. Nijkamp and P. Rietveld (2002), Residential mobility and local housing market differences, Tinbergen Institute Discussion Papers, No. 3.

    • Search Google Scholar
    • Export Citation
  • Verhaest, D. and E. Omey (2006), “The Impact of Overeducation and its Measurement”, Social Indicators Research, Vol. 77, pp. 419448.

    • Search Google Scholar
    • Export Citation

Annex I. Skills Mismatch Index

1. We constructed the index of skill mismatch using the framework presented in Estevao and Tsounta (2011). The index measures the difference between skill supply and demand, where the supply is based on the education attainment of the labor force and the demand on the employment by education. Following Estevao and Tsounta (2011), the skills mismatch index for each country i at time t is constructed using the following equation:

Skillsmismatchindexi,t=Σj=13(Si,j,tMi,j,t)2(1)

in which j is the skill level; Si,j,t is the percentage of the population with skill level j at time t in country i (skill level supply), and Mi,j,t is the percentage of employees with skill level j at time t in country i (skill level demand).

  • Educational attainment data from Eurostat are used to construct skill level supply using less than primary, primary and lower secondary education (as low skilled), upper secondary and post-secondary non-tertiary education (as semi-skilled), and tertiary education (as high skilled). Skill level demand is approximated by the percentage of employees with less than primary, primary and lower secondary education (to proxy low-skilled demand), with upper secondary and post-secondary non-tertiary education (for semi-skilled demand), and with tertiary education (for high-skilled demand).1

2. One component of the skills mismatch index that measures the difference between the share of tertiary educated workers in employment and the share of tertiary educated people in the labor force could be treated as an index of skills shortage/excess. However, skill supply would be significantly influenced by the changes in the education quality. To account for the differences in the education quality overtime and cross-country, we used education quality index from the World Economic Forum.

Highskillsupplyt,i=LF0,ih+Σt=1NΔLFt,ih*EDQt,i(2)

3. Where, LF0,it is initial level of the share of labor force with tertiary education for country i, ΔLFt,ih is the change in labor force with tertiary education at time t for country i, and EDQt,i is the index of education quality from World Economic Forum for country i at time t. the index for education quality is normalized to the 90th percentile of the distribution data for all European countries, so that countries at 90th percentile have value 1.

4. The new data from high skilled labor supply produced by the equation 2 is used to calculate difference in relative shares of high skill labor supply and demand as an index for skills shortage/excess index.

1

Prepared by Ara Stepanyan.

2

Relative to peers, the Slovak Republic has relatively low scores on Transparency International’s Corruption Perception Index, World Governance Indicator’s Control of Corruption Index, and World Economic Forum’s Judicial Independence index.

1

Although the Estevao and Tsounta (2011) method of estimating skill supply could be reasonably robust based on educational attainment, the measures of skill demand and skill intensity does have some weaknesses.

Slovak Republic: Selected Issues
Author: International Monetary Fund. European Dept.