Slovak Republic: Selected Issues

International Monetary Fund. European Dept.

doi:10.5089/9781513587325.002.A001

Author:

International Monetary Fund. European Dept.

International Monetary Fund. European Dept.
Search for other papers by International Monetary Fund. European Dept. in
Current site
Google Scholar
PubMed

Publication Date:: 21 Jun 2021

eISBN:: 9781513587325

ISBN:: 9781513587325

Language:: English

Keywords:: trade cost; auto sector value; containment measure; gross export; trade pattern; COVID-19; Labor supply; Global value chains; Supply shocks

Download PDF (1.9 MB)

The Slovak auto industry is deeply integrated in regional and global value chains. While this has brought tremendous benefits, it has also made the sector susceptible to foreign shocks as exemplified by the disruptions it experienced during the COVID-19 pandemic. Going forward, the industry will need to adapt to mega trends such as potential reconfigurations of supply chains and changes in preferences and technology. This paper uses a general equilibrium global trade model to: (i) quantify the impact of the pandemic-induced labor supply shock on the auto industry in Slovakia; (ii) disentangle the spillovers of the COVID-19 shock to the Slovak auto sector through cross-border and domestic production chains; and (iii) shed light on the Slovak auto sector’s exposure to potential changes in the post-COVID global economy, namely higher trade costs with select partners.

Abstract

The Slovak Auto Sector During the Pandemic and Beyond: Model Based Evidence ^¹

A. Introduction

1. The Slovak auto industry is deeply integrated in regional and global value chains. While this integration has brought significant benefits, it has exposed the industry to foreign shocks. Such exposure became particularly salient during the COVID-19 pandemic. During the first wave of the pandemic, widespread factory shutdowns led to a historic contraction in car production. The high degree of integration of the Slovak auto industry within complex industrial value chains meant that its production was affected not only by car factory shutdowns but also by containment measures implemented up and down the value chain, both domestically and abroad.^² Understanding the relative contribution of domestic lockdowns versus spillovers from lockdowns in trading partners will be important for policy-making going forward, should the current pandemic or future shocks require similar containment measures or trigger disruptions to the supply of key inputs.

2. More broadly, the COVID-19 pandemic may lead to long-lasting changes in the global economy, some signs of which predated the crisis. In particular, a possible trend towards “deglobalization” and renationalization of production (see, for example, Chapter 1 of the April 2021 WEO) could have outsize effects on vehicle manufacturing compared to other industrial processes, since the European automotive supply chain is one of the most complex across sectors (Figure 1). Such changes could have important macroeconomic consequences in the case of Slovakia: the car manufacturing sector directly accounts for 4.0 percent of total gross value added, 3.9 percent of jobs, and about a quarter of Slovakia’s gross exports.^³ In no other country in Europe is the auto sector as dominant in terms of value added of and employment in manufacturing (Figure 2).

3. In order to shed some light on these questions, this paper employs a general equilibrium global trade model, following Bonadio et al. (2020), Huo, Levchenko and Pandalai-Nayar (2020), and Boranova et al. (forthcoming). The framework uses a static multi-country, multi-sector, multi-factor model in which countries trade both intermediate and final goods. Each sector uses labor, capital, and intermediate inputs that can come from any sector and country in the world. In the calibration, we have 64 countries from all continents and 33 economic sectors, and the main data source are the 2015 OECD Inter-country Input-Output tables.

4. The paper provides a quantification of the impact of the pandemic-induced labor supply shock on the auto industry in Slovakia. It disentangles the spillovers of the COVID-19 shock to the Slovak auto sector through cross-border and domestic production chains. Finally, it provides an illustrative simulation of how the Slovak auto sector might fare if trade costs with select partners were to sharply rise in the post-COVID global economy.

B. The Impact of the Pandemic

5. This paper follows closely Bonadio et al. (2021) and Boranova et al. (forthcoming) in modelling containment measures imposed during the COVID-19 pandemic as a labor supply shock, whose intensity depends on: (i) the stringency of the government response (captured by the Oxford Government Response Tracker; the value is set to the maximum stringency up to April 2020 to capture the peak of the first wave of the pandemic) and (ii) how amenable each sector is to working from home (captured by Dingel and Neiman (2020)’s work-from-home intensity index). ^⁴ The contraction of the gross value added of the car industry during the peak of the first wave of the pandemic is decomposed between the direct effect of containment measures in the Slovak car industry and the indirect effect of measures in other sectors of the Slovak economy and in foreign countries.

6. The model predicts that, at the peak of the first wave, gross value added in the Slovak car industry falls by 31.1 percent due to the labor supply shock. If we look instead at value added in the whole economy, we find that the labor supply shock leads to a contraction of 27.5 percent.^⁵ The stronger effect on the auto sector than in the rest of the economy is largely explained by the fact that most tasks involved in car production cannot be done remotely.

7. Although the Slovak car industry is deeply integrated in regional and global value chains, the model indicates that most of the drop in value added was due to domestic containment measures (Figure 3): 69 percent of the decline is explained by the direct effect of restrictions, while 16 percent is due to restrictions in other domestic sectors, of which services account for the lion’s share. The remaining 15 percent is explained by containment measures implemented by trading partners.

8. The geographical distribution of the countries with the largest spillovers on the Slovak auto sector confirms the dominant influence of regional trading partners. Almost two thirds of the spillovers from abroad come from shocks to other European countries, reflecting the role of Slovakia as a producer relying mostly on inputs from within the European region. Outside of Europe, the main shock comes from China, accounting for 1.6 percent of the total contraction; meanwhile, the US accounts for only 0.5 percent of the total.

9. These results suggest that the reintroduction of strict lockdowns domestically to contain new waves of the virus would have a severe impact on the auto industry even if the rest of the world remained largely open. However, international spillovers, particularly those from Europe, are also important.

10. If we repeat the same exercise for other European countries, we find that on average only 11 percent of the shock to the auto industry comes from abroad. This is slightly lower than the 15 percent obtained for Slovakia, probably due to the higher degree of openness and integration into global value chains of the Slovak auto sector.

C. Reorganization of Supply Chains

11. There has recently been considerable speculation that the pandemic might lead to supply chain renationalization as a way to protect against similar shocks in the future. This exacerbates pre-existing concerns regarding global trade tensions, Brexit and potential de-globalization. To quantify the exposure of the Slovak car industry, the same model is employed to compute the change in auto industry value added in a highly stylized counterfactual scenario. In particular, this paper examines the potential impact of a complete shutdown of trade between the EU and China, and between the EU and Great Britain.^⁶ These scenarios are chosen as an illustrative tool of the channels through which the Slovak auto industry is exposed to disruptions in the patterns of international trade, rather than due to the realism of the shock. In each case, we alternatively consider a complete shutdown of trade, and a shutdown only in trade of intermediate goods and services.

12. The Slovak auto sector has a sizeable exposure to trade with China. China’s role as a destination for the value added of the Slovak auto sector has grown significantly over time, as documented by Banerjee and Zeman (2021). Suppliers, however, remain largely from within the region. In the event of a complete shutdown of trade between the EU and China, Slovakia’s auto sector value added would contract by 3.5 percent, the highest across all EU countries (Figure 4).

13. If only trade of intermediates were shut down, there would be a sizeable, but much smaller, contraction of 1.1 percent of auto sector value added. The large difference between the two scenarios reflects that China is a much more important destination for final goods than for intermediates produced by the Slovak auto industry. Other Central and Eastern European (CEE) countries, which also have large auto sectors integrated in global value chains, are also significantly exposed to trade tensions with China, but their exposure is significantly lower than that of the Slovak Republic. The reason for this is simple: China accounts for a larger fraction of Slovak auto exports (both final and intermediate) than for the exports of most other European countries.

14. Meanwhile, the exposure of the Slovak auto sector to trade with Great Britain is far smaller, albeit non-negligible. In the event of a total shutdown of trade between the EU and Great Britain, Slovak auto sector value added would contract by 1.1 percent, and by 0.3 percent when only trade of intermediate goods and services is interrupted (Figure 4). This exposure is similar to that of other CEE countries like Poland and the Czech Republic, and smaller than that of countries like Spain and Germany, for whom Great Britain is a more important export destination.

D. Conclusion

15. The auto sector is macro-critical in the Slovak Republic and is characterized by long and elaborate supply chains. Using a multi-sector and multi-country general equilibrium model, this paper presents a quantitative assessment of how the pandemic-related labor supply shocks— related to lockdowns imposed in different sectors and countries—would impact the Slovak auto sector via global value chains. Our results suggest that these labor supply shocks would have a significant adverse impact on the Slovak auto sector. While labor supply disruptions within the Slovak car sector itself account for the vast majority of the decline in value added, about 30 percent of output drop could be attributed to the effects of disruptions in supply chains within and across borders. Of those, one half come from outside Slovakia’s borders.

16. Second, we exploit the model to shed light on the exposure of the Slovak auto sector to potential reorganization of the auto supply chains. Simulations suggests that, in a highly stylized scenario in which trade costs are high enough to preclude the EU’s trade with select trading partners, the Slovak (and EU’s) auto sector would be significantly worse off. These findings present yet another piece of evidence supportive of the call to resist protectionist tendencies that could undermine productivity and growth in the post-COVID world.

References

Banerjee, Biswajit and Juraj Zeman (2021). “The Motor Vehicles Industry in Slovakia, 2005–2015.” National Bank of Slovakia Occasional Paper 1/2021.
- Search Google Scholar
- Export Citation
Bonadio, Barthelemy, Zhen Huo, Andrei Levchenko and Nitya Pandalai-Nayar (2020). “Global Supply Chains in the Pandemic,” NBER Working Papers 27224.
- Search Google Scholar
- Export Citation
Boranova, Vizhdan, Raju Huidrom, Ezgi Ozturk, Ara Stepanyan, and Shihangyin (Frank) Zhang (forthcoming). “Cars in Europe: Supply Chains and Spillovers during Covid Times.” IMF Working Paper.
- Search Google Scholar
- Export Citation
Dingel, Jonathan I. and Brent Neiman (2020). “How Many Jobs Can be Done at Home?” Covid Economics, Vetted and Real-Time Papers 1:16–24.
- Search Google Scholar
- Export Citation
Fally, Thibault (2012). “Production Staging: Measurement and Facts,” unpublished manuscript, University of Colorado-Boulder.
- Search Google Scholar
- Export Citation
Huo, Zhen, Andrei Levchenko and Nitya Pandalai-Nayar (2020). “International Comovement in the Global Production Network,” NBER Working Papers 25978.
- Search Google Scholar
- Export Citation

Annex I. Model Setup

This annex provides more details on the setup of the general equilibrium model. The world economy consists of N countries (labelled by n and m), and there are J sectors (indexed by j and i) in each country. Each country n is populated by a representative household. The household consumes the final good available in country n and supplies labor and capital to firms. International trade is subject to iceberg trade costs τ_mnj to ship good j from country m to country n (the first subscript denotes the source or exporter, and the second denotes the destination or importer).

Households. There is a continuum of workers in a representative household who gain utility from the common consumption bundle. The household’s utility maximization problem is

\max_{F_{n}, {H_{n j}}} U (F_{n} - \frac{ψ}{1 + ψ} \underset{j}{Σ} ξ_{n j} H_{n j}^{1 + \frac{1}{ψ}})

Subject to

P_{n} F_{n} \equiv \sum_{m, j} P_{m n j} F_{m n j} = \sum_{j} W_{n j} H_{n j} + \sum_{j} R_{n j} K_{n j}

On the supply side, H_nj is the total labor hours supplied to sector j, and K_nj is the amount of Installed capital, which is assumed to be exogenous:

K_{n j} = {\bar{K}}_{n j}

Each unit of labor supply collects a sector-specific wage W_nj, and capital is rented at the sector-specific price R_nj. ξ_nj is a (negative) preference shock on the labor supply of sector j. This preference shock is used to capture the pandemic-induced containment measures. ψ is the Frisch elasticity that governs the responsiveness of labor supply. This type of preference gives an especially simple isoelastic labor supply curve that only depends on the real wage:

\begin{matrix} H_{n j} = {(\frac{W_{n j}}{ξ_{n, j} P_{n}})}^{ψ} & (1) \end{matrix}

On the demand side, F_n is the consumption of final goods with corresponding aggregate consumer price index P_n, F_mnj is the final use in n of sector j goods coming from country m and P_mnj is the corresponding price. The final use F_n in the economy is a Cobb-Douglas aggregate across sectoral final composites, where each sectoral final composite aggregates up country-specific absorptions:

\begin{matrix} F_{n} = Π_{j = 1}^{J} {(F_{n j})}^{ω_{n j}}, F_{n j} = {[\underset{m}{Σ} ϑ_{m n j}^{\frac{1}{ρ}} F_{m n j}^{\frac{ρ - 1}{ρ}}]}^{\frac{ρ}{ρ - 1}} & (2) \end{matrix}

The corresponding price indices can be also expressed by the following CES aggregations:

\begin{matrix} P_{n} = Π_{j = 1}^{J} {(\frac{P_{n j}}{ω_{n j}})}^{ω_{n j}}, P_{n j} = {[\underset{m}{Σ} ϑ_{m n j} F_{m n j}^{1 - ρ}]}^{\frac{1}{1 - ρ}} & (3) \end{matrix}

The final expenditure share of a particular good from country m and sector j that is imported by country n is given by

\begin{matrix} \frac{P_{m n j} F_{m n j}}{Σ_{k . l} P_{k n l} F_{k n l}} \equiv π_{m n j}^{f} = ω_{n j} \frac{ϑ_{m n j} P_{m n j}^{1 - ρ}}{Σ_{k} ϑ_{k n j} P_{k n j}^{1 - ρ}} = ω_{n j} π_{m n j}^{c} & (4) \end{matrix}

Firms. A representative firm in country n and sector j operates a Cobb-Douglas production function

\begin{matrix} Y_{n, j} = Z_{n j} {(K_{n j}^{α_{j}} H_{n j}^{1 - α_{j}})}^{η_{j}} X_{η_{j}}^{1 - η_{j}} & (5) \end{matrix}

Where the TFP is denoted by Z_nj, K_nj and H_nj are the corresponding capital and labor supply from the household, and X_nj is the intermediate input usage that aggregates inputs from all potential countries and sectors:

\begin{matrix} X_{n j} = {(\sum_{i} \sum_{m} μ_{m i, n j, n j}^{\frac{1}{ε}} X_{m i, n j}^{\frac{ε - 1}{ϵ}})}^{\frac{ε}{ε - 1}} & (6) \end{matrix}

where X_mi,nj is the usage of inputs coming from sector i in country m in production of sector j in country n, and μ_mi,nj is the intermediate taste shifter. Similar to the final good price index, we can derive the price index of this intermediate input bundle:

\begin{matrix} P_{n j}^{x} = {(\sum_{i} \sum_{m} μ_{m i, n j} P_{m i, n j}^{1 - ϵ})}^{\frac{1}{1 - ε}} & (7) \end{matrix}

where P_{mj ,nj} be the price paid in country n, sector j for inputs from country m, sector i.

Let P_nj denote the price of output produced by sector j in country n. No arbitrage in shipping implies:

\begin{matrix} P_{m i, n j} = P_{m n i} = τ_{m n i} P_{m i} & (8) \end{matrix}

The cost minimization decision implies that the payments to primary factors and intermediate inputs are:

\begin{matrix} R_{n j} K_{n j} = α_{j} η_{j} P_{n j} Y_{n j} & (9) \end{matrix}

\begin{matrix} W_{n j} H_{n j} = (1 - α_{j}) η_{j} P_{n j} Y_{n j} & (10) \end{matrix}

\begin{matrix} P_{m i, n j} X_{m i, n j} = π_{m i, n j}^{x} (1 - η_{j}) P_{n j} Y_{n j} & (11) \end{matrix}

where $π_{m i, n j}^{x}$ is the share of intermediates from country m, sector i in total intermediate spending by country n, sector j, given by:

\begin{matrix} \frac{P_{m i, n j} X_{m i, n j}}{Σ_{k, l} P_{k i, n l} X_{k i, n l}} \equiv π_{m i, n j}^{χ} = \frac{μ_{m i, n j} {(τ_{m n i} P_{m i})}^{1 - ε}}{Σ_{k, l} μ_{k l, n j} (τ_{k n l} P_{k l}) 1 - ε} & (12) \end{matrix}

Equilibrium conditions. An equilibrium in this economy is a set of goods and factor prices {P_nj, W_nj, R_nj}, factor allocations {K_nj, H_nj}, and goods allocation {Y_nj}, {F_mnj, X_{mi, nj}} for all countries and sectors such that given all exogenous realization of labor supply shocks {ξ_nj}:

(i) households maximize utility by satisfying (1)-(4).
(ii) firms maximize profits through (5)-(12).
(iii) all markets clear.

More specifically, market clearing conditions for sectoral goods should satisfy:

\begin{matrix} P_{n j} Y_{n j} = + \underset{f i n a l u s e}{\underset{︸}{\sum_{m} \sum_{i} η_{i} P_{m i} Y_{m i} ω_{m j} π_{n m j}^{c}}} + \underset{i n t e r m e d i a t e u s e}{\underset{︸}{\sum_{m} \sum_{i} (1 - η_{i}) P_{m i} Y_{m i} π_{n j, m i}^{x}}} & (13) \end{matrix}

On the other hand, the labor supply should meet the corresponding labor demand from representative firms in each country and sector:

\begin{matrix} H_{n j} = {(\frac{W_{n j}}{ξ_{m j} P_{n}})}^{Ψ} = \frac{{(1 - α_{j})}_{η_{j}} P_{n j} Y_{n j}}{W_{n j}} & (14) \end{matrix}

Here we want to highlight another important feature of this framework: like other quantitative models in the international trade literature, the model equilibrium nominal sectoral output P_njY_nj can be solved up to a normalization with information on only the value-added shares η_i and the final and intermediate demand systems $ω_{m j} π_{n m j}^{c} a n d π_{n j, m i}^{x}$ . Given this nice property, we can compute the equilibrium changes of the variables of interest that only requires the calibration of a limited number of parameters. This greatly reduces the burden of computation as the number of countries and sectors increase and makes this framework more tractable to match the data.

Prepared by Mariano Spector, drawing on the forthcoming working paper by Boranova et al. (2021). We thank Barthelemy Bonadio and Andrei Levchenko for sharing the codes of their model and for useful discussions and advice.

Banerjee and Zeman (2021) provide a comprehensive analysis of the evolution of the motor vehicles sector in the Slovak Republic, its integration with global production chains, and backward and forward linkages. See also IMF Staff Report 2019.

Data for direct value added and jobs corresponds to 2019, while the data for gross exports corresponds to 2018 (source: Eurostat). Throughout this paper, we proxy car manufacturing by sector C29 in the European classification system NACE Rev. 2.

The model does not include any “preference” shock against travelling during the pandemic nor nominal rigidities. Therefore, the pandemic shock is modeled purely as a supply shock, although demand is endogenously affected through income effects.

Note that the labor supply shock is calibrated to the peak intensity of the containment measures, so it is not intended to match the GDP contraction in any given quarter.

In these scenarios, there are no labor supply shocks, only a reconfiguration of trade patterns (modeled as an infinitely large increase in iceberg trade costs). The labor supply shock discussed before is intended to capture the short-run impact of the pandemic, while the change in trade structure is interpreted as a long-term shock.

References

Andrle, Michal and Miroslav Plasil (2019), “Assessing House Prices with Prudential and Valuation Measures”. IMF Working Paper 19/59.
- Search Google Scholar
- Export Citation
Bloor, C., and Lu, B. (2019), “Have the LVR restrictions improved the resilience of the banking system?”, Reserve Bank of New Zealand Analytical Note Series, May.
- Search Google Scholar
- Export Citation
EBA (2021), “Risk parameters statistics by country of the counterparty for IRB banks” PD-adjusted for Slovak IRB banks on retail exposures secured on real estate property.
- Search Google Scholar
- Export Citation
EBA (2020), “Final guidelines on the appropriate subsets of sectoral exposures to which competent or designated authorities may apply a systemic risk buffer in accordance with Article 133 (5) (f) of Directive 2013/36/EU”, September.
- Search Google Scholar
- Export Citation
EBA (2016), “Template for notifying national macroprudential measures not covered by CRR/CRD by National Bank of Slovakia”.
- Search Google Scholar
- Export Citation
Ebeke, C., Jovanovic, N., Valderrama, L. and Zhou, J. (2021), “Corporate Liquidity and Solvency in Europe during COVID-19: The Role of Policies”, IMF Working Paper 21/56.
- Search Google Scholar
- Export Citation
ESRB (2014), “Recommendation of the European Systemic Risk Board on guidance for setting countercyclical buffer rates”.
- Search Google Scholar
- Export Citation
Gornicka, L., and Valderrama, L. (2020), “Stress Testing and Calibration of Macroprudential Policy Tools”, IMF Working Paper 20/165, August.
- Search Google Scholar
- Export Citation
Harrison, O., Jurča, P., Rychtarisk, S., and Yackovlev, I. (2018), “Credit growth and macroprudential policies in the Slovak Republic”, Slovak Republic Selected Issues, IMF Country Report 18/242, July.
- Search Google Scholar
- Export Citation
Harrison, I., and Mathew, C., (2008), “Project TUI: A Structural Approach to the Understanding and Measurement of Residential Mortgage Lending Risk,” Reserve Bank of New Zealand.
- Search Google Scholar
- Export Citation
International Monetary Fund (2019), “Downside Risks to House Prices”, Global Financial Stability Report, April.
- Search Google Scholar
- Export Citation
Jurča, P., Klacso, J., Tereanu, E., Forletta, M. and Gross, M. (2020), “The Effectiveness of Borrower-Based Macroprudential Measures: A Quantitative Analysis for Slovakia”, IMF Working Paper 20/134.
- Search Google Scholar
- Export Citation
Moody’s (2021), “Looking beyond the pandemic: CEE’s credit challenges and ESG risks”.
- Search Google Scholar
- Export Citation
National Bank of Slovakia (2020), “Financial Stability Report”, November.
- Search Google Scholar
- Export Citation
National Bank of Slovakia (2021), “Financial Stability Report”, June.
- Search Google Scholar
- Export Citation
Reserve Bank of New Zealand (2021), “Financial Stability Report”, May.
- Search Google Scholar
- Export Citation
Valderrama, L, (forthcoming), “Macroprudential Policy Calibration in Slovakia”, IMF Working Paper.
- Search Google Scholar
- Export Citation

Annex I. A Structural Model to Measure Mortgage Risk

The Model: Projecting Credit Risk, Probability of Default and Loss Given Default

1. We use a structural model by risk and vintage buckets to define the loss event. The modeling approach follows Gornicka and Valderrama (2020), which is itself based on Harrison and Mathew (2008). A borrower defaults if she fails the two insolvency tests in the Slovak insolvency regime: (i) ‘the cash-flow test’ which indicates that the borrower cannot pay the debt as it comes due and the bank can open insolvency proceedings; and (ii) ‘the balance sheet test’ whereby a borrower becomes insolvent if the value of her assets fall below the value of liabilities. This assumption implies that a borrower in financial distress could avoid default, as long as she has positive home equity that be extracted to refinance the loan or repay the debt.^¹ This approach differs from Jurca et al (2020) who assume default if a borrower does not have sufficient cash to repay debt net of subsistence consumption.^²

2. We project credit risk by vintage and aggregate risk at the portfolio level. This is required because borrower-based measures are applied to the new production of mortgages, whereas sectoral capital buffers are applied to the outstanding portfolio. The volume of earlier vintages changes over time as borrowers repay the principal and refinance their mortgages at lower lending rates. Crucially, the risk characteristics of mortgages also transition over time due to the full and partial repayment of principal, the dynamics of real estate prices, the evolution of household income, and changes to interest rates. Following Bloor and Lu (2019), the stock of outstanding mortgages at time t satisfies:

\begin{matrix} S t o c k_{b, t} = S t o c k_{b, t - l} - P a y m e n t s_{b, t} - T r a n s i t i o n s_{b, t} + F l o w s_{b, t} & (1) \end{matrix}

where b captures the risk bucket (LTV, DSTI, DTI). Solving the system recursively and taking into account the maturity of mortgage loans in Slovakia, we trace the migration of mortgage loans (payment, transitions, and flows) from 1999Q1 to 2020Q2. We then aggregate risk at the vintage/risk bucket level to project the loss rate of the outstanding portfolio.

3. The probability that a borrower i gets into financial distress in period t is driven by:

\begin{matrix} \Pr (F D_{i, t}) = Φ (D S T I_{i, t - 1}) \cdot D_{i} + β_{1} \cdot {Δ D S T I}_{i, t - 1}^{γ} + Φ (D S T I_{i, t - 1}) \cdot (β_{2} \cdot U_{t - 1} + β_{3} \cdot Δ U_{t - 1}^{α}) & (2) \end{matrix}

which is a function of: (1) the borrower’s affordability risk measured by the debt service-to-income ratio (DSTI) in the previous period; (2) the change in the debt servicing capacity since the last period (ΔDSTI); (3) the likelihood of being unemployed measured by the unemployment rate (U) in the previous period, and the change in the unemployment rate (ΔU), which are a function of the current DSTI ratio of the borrower; and (4) a demographic factor (D) which captures shocks to household composition.

4. A default occurs when the borrower is in distress and cannot repay the loan to the bank early by selling the collateral at market price $\tilde{P}$ net of transaction costs C^³:

\begin{matrix} {\tilde{P}}_{i, t} - C < N P V (L_{i, t}, r_{t}^{t y p e, M}, r_{t}^{f}, T_{s, t}) & (3) \end{matrix}

where the net present value of the loan (NPV) at time t consists of two elements: (1) the outstanding loan amount L_i,t, and (2) the penalty for early prepayment, which is assumed to equal the discounted value of future interest payments. This is a function of $r_{t}^{t y p e, M}$ which stands for the current interest rate of the mortgage (that differs from the rate at issuance depending on the type of the mortgage and its resetting schedule M); T_s,t which denotes the remaining maturity at time t of a loan issued at time s, and $r_{t}^{f}$ is the 10-year Slovak sovereign yield used to discount the value of future interest payments at time t. The amortization schedule (for each period j over its remaining lifetime) is proxied by a linear amortization scheme.

5. Conditional on defaulting, a bank’s loss given default (LGD) on a mortgage is driven by the discounted sale price of the collateral:

\begin{matrix} {L G D}_{i, t} = \max {0, N P V (L_{i, t}, r_{t}^{t y p e, M}, r_{t}^{f}, T_{s, t})} - (1 - δ) \cdot \frac{{\tilde{P}}_{i, t + n}}{{(1 + r_{t}^{f} + s p r e a d)}^{n}} & (\begin{matrix} 4 \end{matrix}) \end{matrix}

where δ denotes the foreclosure discount at which the bank sells the repossessed collateral at time t+n, where n denotes the time needed to realize the proceeds of the sale, and spread is the risk-adjusted spread used to discount the value of the house. From the perspective of affordability risk, the price of collateral affects the likelihood of financial distress as well as the loss given default for the bank.

6. We simulate probability of default (PD) and LGD for each LTV-vintage bucket of mortgages using Monte Carlo simulation techniques. In practice, we divide the portfolio of mortgages into buckets based on their key credit quality indicators and perform simulations for each bucket separately. We consider buckets by quarter of mortgage origination, and by the LTV (“vintage-LTV buckets”). The vintage of the mortgage matters, as it allows us to calculate the impact of interest shocks on affordability risk, the remaining time to maturity, the outstanding value of the loan and the market value of collateral. For a given vintage-LTV bucket of mortgages, a number N of borrowers already in financial distress is considered. For ea ch o f the N borrowers a house price draw is generated from a distribution with a mean equal to the average house price level in the tail risk scenario. For each of the house price draws, the model determines whether condition (3) is satisfied (i.e. if the borrower defaults) and—if so—the LGD in equation (4). In the next step, the bucket-specific PD is calculated as the total number of defaults divided by the number of draws, N, and multiplied by the bucket-specific probability of financial distress from equation (2). To reduce noise, this simulation process is repeated K times for each bucket. The final bucket-specific PDs and LGDs are calculated as the averages across K iterations. In the applications presented we set N=2000 and K= 10,000. In the final step, the ultimate outputs, i.e. portfolio-wide PD, LGD, and the loss rate are calculated by combining separately estimated outputs for the vintage-LTV buckets and weighting them by the share in the total portfolio.

7. We use a comprehensive supervisory dataset covering all issuances of housing loans in Slovakia starting in 1999Q1 which were still outstanding in 2020Q2. The rich dataset provides volumes of issuance at origination, outstanding amounts as of 2020:Q2, repricing schedules by vintage and remaining time to next refixation period, maturity profile, and lending rates by type of mortgage. Also, it provides loan risk characteristics by DSTI, LTV, and DTI bucket at origination by vintage, as well as the time series of matrices including LTV-DTI and LTV-DSTI starting in 2017Q1. The granularity of reporting is very high with 8 buckets for DSTI, 12 buckets for LTV, and 10 buckets for DTI. Repricing schedules are reporting for 6-time buckets. As the data captures the risk profile at origination, we need to compute risk characteristics Point-in-Time (PiT) to capture the law of motion of house prices, interest rates, and income growth (see Part II of Annex I). For housing loans issued in 2020;Q2, the largest concentration is in the [70–80] LTV / [50–60] DSTI bucket. Overall, 20 percent of mortgages are issued with LTV greater than 70 percent and DSTI exceeding 50 percent.

8. The model is calibrated using historical loan losses on mortgage exposures in Slovakia during 2008–2010 and matching predicted values to IRB banks’ 1-year default probability under baseline conditions. During 2008–2010, real estate prices contracted by 23 percent, unemployment rose by 4.5 percentage points and annual disposable income growth slowed down from 11 to 0 percent. At the same time, lending rates declined by 60 basis points providing some relief to stretched households. We generate a rate of financial distress in equation (2) from the increase in NPLs in mortgage loans from 2.2 to 3.7 percent and using the experience of the UK real estate crisis in the 1990s to back out the share of distressed households. We allocate the share of financial stress due to the rise in unemployment (80 percent) and the change to mortgage rates (20 percent). We distribute the aggregate stressed sales by DSTI bucket to match that a 50 percent increase in the debt servicing ratio for a typical mortgage issued in 2020;Q2 generates an increase in financial distress by 6 percentage points following Harrison and Mathew (2008). Further details of the calibration are found in Valderrama et al (forthcoming).

9. From Origination to Point-in-Time (PiT) Risk Parameters

We calculate the Po int-in-Time (PiT) risk parameters of loans issued from 1999;Q1 through 2020;Q2 as of 2020;Q2 using information on the risk characteristics of the loans, and data provided by NBS on household disposable income, interest rates at origination, and real estate prices. We denote the time of issuance as s, the current period as t, and the maturity of the loan as T.

First, we compute the PiT LTV ratio by backing out the outstanding principal of the loan net of repayments at time t and repricing the mortgage collateral:

\begin{matrix} L T V_{s, t}^{P i T} = \frac{(L T V_{s}^{O r i g} \cdot P_{s}) \cdot (\frac{T - t}{T - s})}{P_{t}} & (5) \end{matrix}

Then, we compute the income PiT using information extracting income at origination from DSTI and lending rates at origination $i_{s}^{O r i g}$ , and quarterly income growth g:

\begin{matrix} I n c o m e_{s} = \frac{(\frac{1}{T} + i_{s}^{O r i g}) \cdot (L T V_{s}^{O r i g} \cdot P_{s})}{D S T I_{t}} & (6) \end{matrix}

\begin{matrix} I n c o m e_{t} = I n c o m e_{s} \cdot {(1 + g)}^{t - s} & (7) \end{matrix}

This allows us to compute DSTI and DTI PiT as:

\begin{matrix} D S T I_{s, t}^{P i T} = \frac{(L T V_{s}^{O r i g} \cdot P_{s}) \cdot (\frac{1}{T - s} + i_{s, t}^{s + k})}{I n c o m e_{i}} & (8) \end{matrix}

\begin{matrix} D T I_{s, t}^{P i T} = \frac{(L T V_{s}^{O r i g} \cdot P_{s}) \cdot (\frac{T - t}{T - s})}{I n c o m e_{i}} & (9) \end{matrix}

where $i_{s, t + j}^{s + k}$ is the lending rate as of t of a mortgage issued in s and with the last re-setting period of interest rate in s+k. During the stress testing horizon at time t+j, we compute the shock to DSTI as :

\begin{matrix} Δ D T I_{s, t + j}^{P i T} = \frac{(L T V_{s}^{O r i g} \cdot P_{s}) \cdot (\frac{1}{T - s} + i_{s, t + j}^{s + λ})}{I n c o m e_{t} \cdot (1 + {s h o c k}_{j})} & (10) \end{matrix}

where $i_{s, t}^{s + k}$ is the lending rate as of t+j of a mortgage issued in s and with the last re-setting period of interest rate in s + λ.

Prepared by Laura Valderrama (EUR). The author thanks Marek Licak, Pavol Jurča, Ján Klasco, and colleagues at the National Bank of Slovakia for sharing the data and for insightful comments and remarks,and Lucyna Górnicka, Peter Harvan, and Erlend Nier for very fruitful discussions.

The rise in house prices in Slovakia mirrors trends in global asset prices supported by accommodative financial conditions

House price growth is regressed on one-year lag of afford ability (detrended housing prices/GDP per capita), GDP per capita growth, employment rates growth, and changes in construction costs. The estimation is conducted using a panel approach for CESEE countries over 2006–2020. A key driver of the hosing price gap is the recent decline in GDP, which may not fully capture long-term income effects or households’ debt servicing capacity.

National Bank of Slovakia (2021), “Financial Stability Report”, May.

Banks are allowed to exceed the maturity limit for ten percent of new loans but in practice have made limited use of this option.

Five percent of new loans can exceed the DTI limit of 8.

Exemptions included five percent of new loans granted with DSTI up to 70 percent; and an additional five percent of new consumer loans granted below 70 percent provided their maturity does not exceed five years.

While consumer credit started decelerating in January 2019, the pandemic has reinforced this trend. Factors underlining the consumer credit contraction include lower demand related to the general decline in consumption during the pandemic,some consolidation of debts into housing loans,and binding regulatory limits.

In Slovakia,the average maturity of new housing loans issued in 2020 was around 22½ years, with around 54 percent of loans between 20 and 30 years. However, the time to repricing is shorter at around 4¼ years with one third of new issuances having a remaining time to re-setting period between 2 and 3 years.

For a household formed by two adults and two children and net income of EUR 2,000, a DSTI ratio of 40 percent in peer countries is equivalent to 80 percent under NBS DSTI definition.

The large production is partly driven by the high share of refinancing loans which increased from 50 percent in 2019 to 55 percent in 2021 (NBS, 2021).

The baseline scenario includes income and lending rate projections consistent with WEO forecasts as of April 2021. For ease of comparison with NBS stress test results, the scenario includes projections for unemployment based on NBS baseline scenario (FSR, November 2020). To benchmark baseline results against Slovak banks’ credit risk projections for mortgage exposures booked under IRB, the house price shock is consistent with a downturn scenario as prescribed by Basel for the calculation of capital requirements for IRB exposures.

See Slovakia staff report (2021), Box 2: “The corporate sector”, June; and Ebeke et al (2021).

The adjusted PD on retail exposures secured on real estate property reported by Slovak IRB banks ranges between 0.33 percent (median) and 0.88 percent (weighted average). See EBA (2021).

The assessment of potential of leakages from the migration of housing loans outside the scope of the macroprudential tool as a result of tightening of LTV limits is outside the scope of the paper.

See Jurča et al. (2020) for an analysis of the relative effectiveness of LTV and DSTI limits on the probability of default using Slovak household surveys.

For a related discussion, see also Andrle and Plasil (2019).

This figure is marginally lower than the NBS estimate of the sectoral SyRB needed to cover stressed losses incurred by banks.

The differentiation of macro prudential risk weights by risk bucket is applied in Belgium and New Zealand.

This finding is, however, contingent on the modeling framework, which does not measure the potential migration of housing loans to the consumer portfolio to top up LTV-constrained mortgages.

By contrast, the model does not consider “strategic defaults”, i.e. a situation where the borrower decides to stop repayments once the value of the underlying collateral falls below the value of the loan. Incentives to do so might exist in the case of non-recourse loans, which is not the case in Slovakia.

Their analysis is also based on projecting credit risk for mortgage flows rather than for the outstanding portfolio.

The transaction cost is likely to increase during a crisis, given the high stock of foreclosed properties.

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/002/2021/134/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully