![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
We develop a methodology for identifying financially distressed households and use it for testing the responses to shocks to the unemployment rate, the interest rate and prices of essential expenditure in the Czech Republic. We extend the approach of Johansson and Persson (2006) for Sweden and Albacete and Fessler (2010) for Austria in the literature to allow for full labour market transitions between employment and unemployment, and, due to data availability, to account for heads and spouses within households. This improvement may lead to a higher response of household distress incidence, due to the unemployment rate shock, than in both Sweden and Austria, while the effects due to the interest rate shock are of similar size as in Austria. We illustrate the use of our approach for stress testing households’ ability to pay their debts using macroeconomic scenarios from the CNB’s official forecast and from the CNB’s Financial Stability Report. The results highlight the importance of using micro-level datasets in the analysis of household distress incidence, as the impact of shocks is more pronounced among lower-income households.
Acknowledgement
The authors thank Klára Kalíšková (CERGE-EI) for the calculation of aggregate labour market flows, and Jan Frait, Michal Hlaváček, Eva Hromádková, Pirmin Fessler, Jiří Slačálek and an anonymous referee for valuable comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Frait and Komárková (Citation2012) describe how systemic risk evolves over the financial cycle and propose a framework for a bank-based economy with a simple financial sector similar to the Czech Republic with the aim of preventing systemic risk in the accumulation phase of the cycle and to mitigate its effects.
2. Credit risk resulting from impairment in debt repayments may represent the main source of risk for the financial stability of the Czech banking sector. Hence, we focus in our approach on the impact of various shocks on indebted households.
3. The model is used to simulate the impacts of 1 standard deviation shocks added to a basic scenario. The standard deviations of the variables are calculated for the period 1986–2005. The authors admit that the impact of the interest rate shock is larger because the rates were higher and more volatile in the run-up to Finland’s accession to the EMU in 1999.
4. The LGDs are not the same as those calculated by banks. The purpose is not to replicate the LGDs used by banks, but to use a risk metric that the authors are able to construct given the available data.
5. This approach abstracts from business cycle effects. In particular, interest rates and household income normally increase with more robust economic activity.
6. The household debt is concentrated among the highest income category. These households are often formed by two employed adult persons. If one adult loses his job, the other person’s income is sufficient to cover living costs and interest outlays.
7. Less than 40% of households had at least one loan in 2000–2008, while the average amount of debt among households that borrow increased from 7 to 11% of disposable income, particularly for young households with higher income levels, due to expansion of the housing loan market. Czech household indebtedness as measured by the loans-to-GDP ratio was lower than in the euro area and similar to that in other Central European economies.
8. The living minimum is declared by the Ministry of Labour and Social Affairs and covers minimum living costs in relation to the demographic composition of the household.
9. Shocks to unemployment influence not only the probability of becoming unemployed, but also unemployed persons’ chances of returning to work. Previous literature highlights the importance of unemployment duration on the level of unemployment (see for example Elsby, Hobijn, and Sahin Citation2013).
10. Moreover, the amounts of the living minimum related to household expenditures used by Bičáková, Pašaličová, and Prelcová (Citation2010) in the ADB measure were abolished in 2007, which would complicate its use in the more recent period.
11. As an alternative, we consider the amounts of the statutory living minimum (subsistence minimum amounts) as essential expenditure in equation (2).
12. The impact of changes in property prices on the financial condition of Czech households is studied in detail in Brůha, Hlaváček, and Komárek (Citation2013).
13. Exchange rate shocks affect consumer inflation through their impact on prices of imported goods. We leave the assessment of this channel of the exchange rate shock to household budgets for future research.
14. Unemployed persons lose their job skills, which is not reflected in their observed characteristics xi included in equation (3). The predicted probability of their transitions into employment is thus overestimated. On the other hand, employed individuals may be supposed to strengthen their job skills, which are unobserved to the econometrician, leading to overestimation of their transitions into unemployment.
15. On the other hand, employment to unemployment flows are more closely related to the business cycle expressed as changes in the unemployment rate (see Figure ). This is supported by simple regressions, which are available from the authors upon request. Furthermore, the evidence in Figure also suggests that the share of the inactive population in the total population was unchanged throughout the period. This justifies our assumption of a constant share of the inactive population.
16. Not calibrating the model would mean that the transition probabilities were entirely determined by the statistical significance of the relation between the covariates and the dependent variable. The tighter the relation, the lower the flows from employment to unemployment, and vice versa.
17. For a more detailed description of the Czech tax and benefit system and full simulations of changes in net household income due to transitions between employment and unemployment, see Galuščák and Pavel (Citation2012).
18. The identification of the model is based on nonlinearity in the selection equation in equation (4').
19. Benczúr et al. (Citation2014) estimate the effect of income taxation on labour force participation using the Hungarian Household Budget Survey. In the Heckman (Citation1979) selection model they introduce gains to work, reflecting the fact that non-labour income accounts for lost transfers due to taking up a job. The same methodology is used in Galuščák and Kátay (Citation2014) to compare the effects of the Czech and Hungarian tax and benefit systems on labour supply. We leave the use of their methodology for stress testing the Czech household sector for future research.
20. The set of explanatory variables xi used in equations (4) and (4') is the same as in equation (3). Hence, we predict the probability of unemployment as unity minus the predicted values from equation (4') in our simulations instead of the predicted values from equation (3). The use of predicted participation values from the Heckman selection model accounts for the wage equation in model (4). Nevertheless, we leave the description of the unemployment probability model in this section as well as reporting its estimates in Section 5 for expository purposes.
21. We assume that within households, potential transitions between labour market states are independent.
22. This approach is equivalent to performing Monte Carlo simulations. Johansson and Persson (Citation2006) and Albacete and Fessler (Citation2010) employ Monte Carlo simulations, as they consider the unemployment shock to affect employed individuals only, while unemployed individuals remain unemployed after the shock.
23. The average weighted residual maturity for all loan types is thus set at 13 years.
24. By long-term effects we mean full refixation of loans and by short-term effects we mean partial refixation.
25. For housing loans, the most often used fixation periods are 1, 3 and 5 years. All interest rates in the first category of loans, one-third in the second category of loans and one-fifth in the last category are thus refixed within a 1-year period. If we assume that all three types of fixations are equally frequent, then on average 51.1% of interest rates are refixed within a 1-year horizon. As a robustness check, we show alternative results with a different ratio of refixed interest rates.
26. We neglect the effect of increased capital income from deposits due to higher interest rates. According to aggregate statistics, the flows from deposits are relatively small compared with the outflows due to loan repayments. Moreover, interest rates on deposits are less sensitive to movements in the general interest rate.
27. The effects of partial price shocks are rather short-term, so we neglect the potential transmission of higher prices into wages through collective bargaining.
28. Rental expenditure is included in the other goods category in Dybczak, Tóth, and Voňka (Citation2010), which besides rents consists of materials for the maintenance of dwellings and other services. The price elasticity is –0.25, while the income elasticity stands at 0.87. As the category of other goods is too heterogeneous, we assume that rents are price and income inelastic.
29. In Table we also show the average values of the statutory subsistence minimum amount, which we use as an alternative definition of essential living costs.
30. The assumption of normal distribution of shocked variables is not fully realistic, as the empirical distributions on the one hand are bell-shaped but on the other hand have heavier tails than the normal distribution. However, the probability of large shocks remains very low.
31. In the stress testing results reported in CNB (Citation2013b), the HBS dataset for 2012 was not available, so the 2012 data were approximated using wage and other income growth rates consistent with the available macroeconomic evidence. See Hlaváč, Jakubík, and Galuščák (Citation2013) for details on how they aged the 2011 HBS data into 2012.
32. We investigate how the results are sensitive to the definition of essential expenditure when we include expenditure on food, energy, health and rent. In Figure A1 in the Appendix we show the impact of the unemployment and interest rate shocks if we use the statutory subsistence minimum amounts as essential expenditure in the financial surplus defined in equation (2). We observe that the incidence of distress is lower in comparison with our baseline definition of essential expenditure, but the impact of shocks is very similar to the evidence in Figure . Next, we consider the alternative share of refixed loans in the short-term effects of the interest rate shock. The results – in Figure A2 in the Appendix – suggest that the percentage of distressed households is very similar at refixed rates of 50% or less. On the other hand, a higher share of distressed households is observed at a refixed rate of 50% for the most sizeable shock of 3 standard deviations.
33. The datasets contain weights that may be used for the estimation and aggregation of the results. The use of weights in the estimation and in the calculation of the aggregate indicators has a negligible effect on the results, so we rely on unweighted regressions and averages.
34. Based on the statistics in Table , the ratio of instalments to net household income is 13%. This is the upper bound for the interest ratio, as it contains the interest as well as principal repayments.
35. The quintiles are determined according to household income in the whole sample of households. Figure shows households with debt only, so the number of households is not the same across quintiles.
36. The results hold if we consider the percentage of distressed households with respect to all households (see Figure A6 in the Appendix).