Call for applications: Klaus Liebscher ­
Economic Research Scholarship

Please e-mail applications to scholarship@oenb.at by the end of October 2021.
Applicants will be notified of the jury’s decision by end-November 2021.

The Oesterreichische Nationalbank (OeNB) invites applications for the “Klaus ­Liebscher Economic Research Scholarship.” This scholarship program gives outstanding researchers the opportunity to contribute their expertise to the research activities of the OeNB’s Economic Analysis and Research Department. This contribution will take the form of remunerated consultancy services.

The scholarship program targets Austrian and international experts with a proven research record in economics and finance, and postdoctoral research experience. Applicants need to be in active employment and should be interested in broadening their research experience and expanding their personal research ­networks. Given the OeNB’s strategic research focus on Central, Eastern and Southeastern Europe, the analysis of economic developments in this region will be a key field of research in this context.

The OeNB offers a stimulating and professional research environment in close proximity to the policymaking process. The selected scholarship recipients will be expected to collaborate with the OeNB’s research staff on a prespecified topic and are invited to participate actively in the department’s internal seminars and other research activities. Their research output may be published in one of the department’s publication outlets or as an OeNB Working Paper. As a rule, the consultancy services under the scholarship will be provided over a period of two to three months. As far as possible, an adequate accommodation for the stay in Vienna will be provided. ¹

Applicants must provide the following documents and information:

• a letter of motivation, including an indication of the time period envisaged for the consultancy
• a detailed consultancy proposal
• a description of current research topics and activities
• an academic curriculum vitae
• an up-to-date list of publications (or an extract therefrom)
• the names of two references that the OeNB may contact to obtain further information about the applicant
• evidence of basic income during the term of the scholarship (employment contract with the applicant’s home institution)
• written confirmation by the home institution that the provision of consultancy services by the applicant is not in violation of the applicant’s employment contract with the home institution

Studies

Use of loan moratoria by CESEE households: who are the users and how vulnerable
are they?

Katharina Allinger, Elisabeth Beckmann ²

Loan repayment moratoria were widely used during the COVID-19 pandemic to mitigate ­liquidity problems in the private sector and thus rapid asset quality deterioration in the banking sector. We provide novel, comparable survey evidence on the use of moratoria by households in ten Central, Eastern and Southeastern (CESEE) countries. In countries where eligible borrowers had to opt in to use moratoria, i.e. qualify and apply, 14% of borrowers did so on average; in countries where borrowers had to opt out, i.e. take action not to make use of automatically applied moratoria, take-up was 55% on average. We find that for opt-in moratoria, the main determinant of take-up is the degree to which borrowers’ finances were affected by the pandemic. Moratorium take-up is also strongly affected by the extent of indebtedness, particularly in opt-out countries. Using information on loan arrears, we show that individuals who had exited from their moratoria by fall 2020 were not more likely to be in arrears than those who never used moratoria. However, these results probably constitute the lower bound for loan defaults that may occur once all moratoria have expired. After all, we also find that borrowers whose moratoria were still active in fall 2020 were subject to more adverse shocks and exhibited a higher degree of indebtedness than borrowers who had exited moratoria.

JEL classification: G51, D14, G18, G28

Keywords: household debt relief, moratoria, loan arrears

Our paper is related to the extensive research on debt relief programs for households. A central debate in this research is whether government interventions prevent unnecessary foreclosures or generate moral hazard problems (see e.g. Guiso et al., 2013, and Agarwal et al., 2017). For CESEE, research on previous debt relief has focused on foreign currency loans (see e.g. Fischer and Yesin, 2019).

Research analyzing debt relief in the form of loan moratoria during the COVID-19 pandemic is still extremely scarce. A notable exception are Cherry et al. (2021), who analyze debt relief measures in the US based on credit registry data. They find that in contrast to previous economic crises, loan default rates did not increase in 2020 along with unemployment and that debt distress was successfully reduced by debt relief measures. Also providing evidence for the US, Haughwout et al. (2020) find that borrowers who relied on moratoria had lower credit scores and approximately 30% higher outstanding debt balances than those who did not.

So far, very few details are (publicly) known regarding the characteristics of moratorium users and the effectiveness of moratoria in preventing short-term ­liquidity and medium-term solvency issues in CESEE. At the time of writing, the Hungarian National Bank (2020, box 3) is, to our knowledge, the only CESEE-10 central bank to have published more detailed information on moratorium user characteristics.

Our study is therefore one of the first to fill an important gap in the literature on a highly policy-relevant topic. We present evidence on the use of moratoria by individuals in CESEE-10 collected in the 2020 wave of the OeNB Euro Survey, which was conducted from late September until early November. To our knowledge, our data are the only comparable cross-country evidence on the use of moratoria by individuals for such a large set of CESEE countries. In the following, we first provide some descriptive evidence on the use of moratoria and then proceed to analyze who used moratoria, differentiating between active and expired ⁶ moratoria, and provide some preliminary evidence on how moratoria affected loan defaults.

Our results show that the take-up of moratoria cannot be explained by socioeconomic characteristics of borrowers. For opt-in moratoria, the main determinant of take-up is how seriously borrowers’ finances were affected by the COVID-19 pandemic, e.g. whether borrowers were laid off or suffered some other income loss. The use of moratoria in opt-in and opt-out countries was strongly affected by the extent of indebtedness, for instance, money owed to several lenders. Being adversely affected by the pandemic increased the probability of borrowers still using moratoria at the time of the survey. Borrowers with higher debt amounts or a larger number of debt instruments were significantly more likely to continue to rely on moratoria in fall 2020. Thus, the debt overhang or difficulties servicing debts are likely going to be larger for borrowers who relied on moratoria for an extended period of time or whose moratoria were still active at the time of the survey.

When looking at arrears, we find that borrowers whose moratoria had expired by fall 2020 are not more likely to be delinquent on loans than borrowers who never took up moratoria. This finding gives rise to cautious optimism – even though our result is subject to some caveats. Moreover, our finding regarding loan arrears may not be transferable to borrowers who relied on moratoria for an ­extended period of time.

In the following, we briefly discuss the design of moratoria in CESEE in section 1. Section 2 provides an overview of the OeNB Euro Survey data and the limitations of these data. Section 3 presents the descriptive evidence. In section 4 we outline our empirical strategy before presenting analytical results and robustness analyses in section 5. Section 6 concludes.

1 The design of moratoria in CESEE

Many countries closely or loosely modeled moratoria on the European Banking Authority’s (EBA) guidelines on legislative and nonlegislative moratoria in light of the COVID-19 pandemic, published on April 2, 2020. The EBA guidelines set out the conditions for general payment moratoria which do not automatically trigger the reclassification of exposures as forborne (as defined under Article 47b of the CRR) or defaulted (as defined under Article 178 of the CRR). Over time, the EBA guidelines were extended and amended. At the time of the interviews in fall 2020, EBA/GL/2020/08 applied and most EU member states included in the OeNB Euro Survey were using EBA guideline-compliant moratoria. ⁷ In Croatia, the application deadline exceeded the one set out in the guidelines, while Romania did not notify the EBA about compliance. ⁸

However, even in the countries that provided EBA-compliant moratoria, not all moratoria were compliant to the letter of the EBA guidelines. As summarized in table 1, key design features ⁹ differed across countries, for instance the requirement for borrowers to either opt in (e.g. apply for the moratorium) or opt out (of the automatically applied moratorium). In the opt-out countries, Hungary, North Macedonia ¹⁰ and Serbia, the take-up of moratoria was much higher than in the opt-in countries. This makes sense as the opt-out regime did not require further action from borrowers such as contacting their bank or compiling application ­documents. Moreover, unlike most opt-in moratoria, the opt-out moratoria were not conditional on borrowers’ degree of affectedness by the pandemic. Some moratoria may thus also have been used ­preventively. Therefore, we should find different ­determinants of moratorium use for opt-out and opt-in countries (see section 5.1). ¹¹

Other important design differences are public (or legislative) moratoria versus private (or nonlegislative) moratoria. Private moratoria, which were largely based on guidelines issued by national banking associations, were implemented in Croatia, Poland and Bulgaria. ¹² According to information provided by the central banks of Poland and Croatia, the conditions for these moratoria varied quite widely across borrowers. Most countries provided moratoria for companies and households, and some countries applied different criteria for companies and households. Many countries accepted a broad range of credit products (e.g. credit cards, overdrafts, leasing agreements) and included nonbank lenders. However, regarding these points there was also considerable variation within the region: Czechia, for instance, was among the more restrictive countries, as the moratorium did not apply to credit products other than loans, and corporates were not allowed to defer interest payments.

Also, the deadlines for applications differed. In most countries covered by the OeNB Euro Survey, our main data source for this study, borrowers were still able to apply for moratoria at the time of the interviews in fall 2020. However, in Albania the deadline for applications had expired already in August 2020 and in Serbia the opt-out moratorium expired at the end of September and the new opt-in moratorium for distressed debtors was not announced until mid-December. North ­Macedonia is also an outlier, as borrowers were offered two opportunities to switch to more favorable loan terms, most often in the form of a postponement of repayments.

Table 1: Some key design features of moratoria in CESEE  

Country	Public/private	Application deadline	Opt-in/opt-out	Eligible borrowers/repayments
Bulgaria	Private1	March 31, 2021	Opt-in	Corporates and individuals if negatively affected by COVID-19
Czechia	Public	Oct. 31, 2020	Opt-in	Corporates (principal only), households (interest capitalized)
Hungary	Public	June 30, 20213	Opt-out	Private sector
Croatia	Private2	March 21, 2021	Opt-in	Corporates and individuals if negatively affected by COVID-19
Poland	Private	Sept. 30, 20204	Opt-in	Corporates and individuals if negatively affected by COVID-19
Romania	Public	March 15, 2021	Opt-in	Corporates and individuals if negatively affected by COVID-19 (interest accrued and (except for mortgages) capitalized)
Albania	Public	Aug. 31, 2020	Opt-in	Corporates and individuals if negatively affected by COVID-19
Bosnia and Herzegovina	Public5	Dec. 31, 2020	Opt-in	Corporates and individuals if negatively affected by COVID-19
North Macedonia	Public	Sept. 1, 20206	Opt-out (for individuals)	Corporates and individuals
Serbia	Public	Sept. 30, 20203,7	Opt-out7	Corporates and individuals
Source: Authors’ compilation.
1 The initial solution was a public moratorium, which applied for banks until May 13, 2021, and was tied to the state of emergency.
2 A public moratorium was provided for loans from the Croatian Development Bank HBOR. These make up a small portion of overall loans.
3 End date rather than application deadline for opt-out moratoria in Hungary and Serbia.
4 Extended until March 31, 2021, for negatively affected corporates.
5 Banks were required to work out adequate modalities for repayment and offered moratoria as one option.
6 Borrowers were offered two opportunities to switch to more favorable loan terms, first in March and then in September.
7 In mid-December, a new opt-in moratorium was adopted with the application deadline running until April 30, 2021, for negatively affected borrowers.

2 Data

In the 2020 wave of the OeNB Euro Survey, we included a question on moratoria, which read:

Our question clearly has some caveats when considering the different design features of moratoria discussed in section 1. The wording is necessarily a compromise that applies better to some countries than others. For instance, we decided to filter moratorium use on loan holders for simplicity, despite the fact that other credit products, e.g. overdrafts, were also eligible for moratoria in some countries. Our data thus only capture moratoria taken up by borrowers with at least one loan outstanding. ¹⁵ Moreover, we do not ask for loan amounts and therefore cannot make any assessment about the volumes of household loans affected by moratoria. These caveats are among the main reasons why we do not enter into a detailed comparison of our data with the scarce publicly available data on moratorium take-up in section 3.

The OeNB Euro Survey also elicits a rich set of information on socioeconomic characteristics; information on individual finances, including loan characteristics; beliefs and expectations; trust; financial literacy; indicators of wealth and income; income and labor shocks; as well as mitigating actions taken in response to the COVID-19 crisis. Table A1 in the annex provides a definition of the variables we use in this study and the wording of the survey questions. Table A2 shows the summary statistics of our main sociodemographic and other variables of interest for the full sample and groups of interest.

3 Descriptive results on moratorium use

To cross-check our data, we compared them with the scarce publicly available data on moratorium use by households from national sources, EBA (2020) and EIB (2021, 2020). However, most of the available public evidence refers to moratorium use in terms of shares of portfolios rather than shares of borrowers, severely limiting comparability. Moreover, where information on the share of borrowers is available, the type of credit is often not disclosed. Whether the data include all credit products or just loans, all participating institutions or just (a selection of) banks, all moratoria or just EBA-compliant moratoria, makes a substantial difference in the reported shares. This also helps to explain the publication of seemingly conflicting evidence on moratorium use for individual countries (see table A3 in the annex). Overall, we find little evidence that would lead us to believe that there are grave and systematic errors in our data. The country where our data seem most out of sync with the available public data is Hungary – we cannot verify the reason, but different data populations (households versus individuals) and/or possibly undersampled consumer loans, which were most affected by moratoria in Hungary (Hungarian National Bank, 2020 and 2021), could be a factor. Generally, credit registers, public authorities or banking associations can provide more precise and more granular (but often confidential) national data than the OeNB Euro Survey. ¹⁶ The strength of our data lies in the cross-country comparison and the combination with a rich microeconomic dataset, which allows us to tackle the research questions detailed in section 4.

4 Empirical strategy

We thus investigate three main questions in this study: (1) Which individual characteristics have the highest significant correlation with the use of moratoria? We are particularly interested in finding out whether sociodemographic characteristics and preferences, loan features or the degree of affectedness by the pandemic have the highest impact and whether differences can be traced back to moratorium design. (2) Are borrowers who benefited from moratoria during 2020 but had ­exited by fall different from those whose moratoria were still ­active in fall 2020? (3) How is moratorium use related to loan arrears and default? The combination of these questions is particularly interesting from a policy perspective. After all, we form our expectations regarding credit quality deteriorations over the next months partially from the evidence on whether moratorium users resumed repayment once their moratoria ended. If the characteristics of borrowers with active moratoria were to differ from those of borrowers with ­expired moratoria, then these extra­polations could be quite erroneous.

Our key dependent variable on moratorium use is filtered on borrowers with one or more loans outstanding from a bank or nonbank financial institution. Estimates that do not take into account that the selection into the credit market may be biased downward (Mocetti and Viviano, 2017). Therefore, we considered a Heckman (1979) model. In principle, we have variables available that are likely good instruments for dealing with selection bias, e.g. the distance to banks and bank concentration as well as some sociodemographic characteristics, such as age, being married, having children, owning a car. Regressions confirm previous research that these variables have a strong impact on the probability of having been granted a loan (Beckmann et al., 2012; Costa and Farinha, 2012; Nguyen, 2007), but no ­theoretical or statistical impact on moratorium use. However, the coefficient rho, which assesses selection bias, is insignificant in all our specifications and the ­coefficients of the Heckman probit and an ordinary probit model are almost indistinguishable – we therefore opted for simplicity and use a probit model for our estimation:

In this model, M is the observed outcome of moratorium use, which depends on M*, the underlying, continuous variable determining the take-up of the moratorium. We might think of M* as the utility the borrower gets from using the moratorium, e.g. in terms of alleviating his/her financial distress. M* is determined by a set of variables X, that are derived from theoretical and empirical research (see also tables A1 and A2 in the annex). We split these into four subsets:

To analyze how borrowers with expired moratoria differ from borrowers with active moratoria, we define three outcomes for our dependent variable: no moratorium, expired moratorium, active moratorium. Theoretically it is possible to view these outcomes as ordered, which would imply that if the underlying latent variable M* is above a certain threshold, borrowers are more likely to have active moratoria rather than expired moratoria. We consider this a plausible assumption for those countries where application deadlines ran beyond the time of the survey. We therefore estimate a generalized ordered logit model (Williams, 2006), as the parallel regression assumption of the ordered logit model does not hold for all ­explanatory variables according to a Brant test and an approximate likelihood-ratio test (Brant, 1990; Wolfe, 1997; Long and Freese, 2014).

As our outcome variable does not have a natural and undisputable ordering, we also estimate a multinominal logit – which assumes no ordering of outcomes. The resulting estimates are unbiased if possibly less efficient.

To assess the impact of moratorium use on loan repayment, we again estimate probit models where the dependent variable indicates whether borrowers are or have been in loan arrears. The explanatory variables are informed by previous ­research: socioeconomic characteristics, loan characteristics and adverse shocks (see Duygan-Bump and Grant, 2009; Guiso et al., 2013).

All regressions include country fixed effects and standard errors clustered at the country level.

5 Results

5.1 Determinants of moratorium use

More important explanatory factors than sociodemographic variables are the number of loans and the degree of affectedness by the COVID crisis. Borrowers with more than four forms of debt (including nonbank and informal debts) are
9 percentage points more likely to have relied on moratoria than other borrowers. In ­CESEE-10, the probability of moratorium take-up increases with the number of mitigating actions borrowers had to take in response to the crisis (e.g. reduce consumption or savings). The probability of moratorium take-up is 13 percentage points higher among those who had to take more than five mitigating actions. Also, individuals who experienced an income shock are significantly more likely to take up moratoria. It should be noted that the importance of sociodemographic variables barely changes if crisis variables are dropped. Results for the US also show that households who suffered from COVID-19-related shocks are more likely to use moratoria (Cherry et al., 2021).

In contrast to the US (Cherry et al. 2021), CESEE-10 borrowers who live in regions that were more affected economically by COVID-19 are not more likely to use moratoria: Local economic shocks (as proxied by nightlight following Henderson et al., 2012 ¹⁹ ) are not a significant determinant of moratorium use ²⁰ – individual affectedness is the main determinant.

Column 2 and column 3 show the estimations for opt-in and opt-out countries separately. We split the sample, as the decision to take out repayment moratoria is likely governed by very different factors, depending on this fundamental difference in design. In line with the discussion in section 1 we expect that COVID-19 affectedness should be the strongest predictor in the opt-in countries, as moratoria were explicitly targeted towards these borrowers. In opt-out countries, where moratoria were not targeted to specific borrowers, we expect a more diverse set of determinants. These expectations are confirmed by our estimations. In the opt-in countries, the variables with the largest significant impact on the decision to use moratoria are indeed the COVID-affectedness variables. ²¹

On the contrary, in the opt-out countries many affectedness variables are insignificant – this corresponds to our expectation that affectedness was less important than other factors and that moratoria reached a broad range of people with diverse motivations for take-up. It is interesting to note that the coefficient for those working in industries strongly affected by lockdowns, e.g. tourism, arts and personal services, is very high and significant in the opt-out countries. This could mean that beyond affectedness, these borrowers might also have had a precautionary motive to safeguard finances amidst income uncertainty.

Compared to opt-in countries, the number of debt instruments is more important among the opt-out countries. People with two or three different types of debt are almost 10 percentage points more likely to have taken up moratoria, while those with four or more forms of debt are 14 percentage points more likely. This makes sense as moratorium relief may have been a welcome opportunity for some borrowers for reasons of distress (if they were overindebted already before the pandemic) or for precautionary reasons (if they were unsure about the impact of the pandemic on their job and income). Borrowers with foreign currency loans are significantly less likely to have taken up moratoria. Generally, there are quite some differences to the opt-in countries in other sociodemographic variables as well – e.g. being self-­employed reduces the likelihood of take-up by 13 percentage points in the opt-out countries, while increasing the probability in the opt-in countries, even though the coefficient is only mildly significant. Being in charge of household finances increases the probability by almost 16 percentage points in the opt-out countries. Similar to results from the US, where borrowers had to apply for moratoria (Cherry et al., 2021), mortgage holders in opt-in countries are significantly more likely to use moratoria than borrowers with consumption or car loans.

Finally, if we only look at countries with private moratoria (Bulgaria, Croatia and Poland), ²² findings are similar to opt-in countries with one notable exception: Trust in banks significantly increases the probability of moratorium use by 4 percentage points – highlighting that private moratoria often require renegotiation between borrowers and lenders.

Clearly, country-specific factors that may not be captured by country fixed ­effects could play an important role, and the results for country groupings may differ from results for individual countries. For example, a recent analysis by the Hungarian National Bank (2021) shows that in Hungary moratorium use is negatively correlated with income. However, due to low numbers of observations per country we do not present results at a more disaggregate level. ²³

Table 2: Determinants of moratorium use  

Dependent variable	Moratorium use
Sample	All countries	Opt-in countries	Opt-out countries
Female	0.015	0.028**	–0.012
	(0.014)	(0.013)	(0.021)
Size of household	–0.009	–0.017***	0.012
	(0.008)	(0.005)	(0.023)
Manages household finances	0.041	0	0.151***
	(0.030)	(0.033)	(0.003)
Unemployed	0.006	0.023	–0.068
	(0.052)	(0.040)	(0.123)
Self-employed	0	0.035*	–0.129***
	(0.030)	(0.019)	(0.034)
Retired	–0.109***	–0.085*	–0.137***
	(0.025)	(0.043)	(0.034)
Income: refused answer	–0.003	0.038	–0.059
	(0.036)	(0.032)	(0.067)
Income: low	–0.032	–0.01	–0.102
	(0.043)	(0.027)	(0.150)
Income: medium	0.006	0	0.033
	(0.018)	(0.023)	(0.028)
Condition of residence: poor	0.004	–0.054***	0.112***
	(0.029)	(0.021)	(0.043)
Own secondary residence	–0.001	–0.007	0.014
	(0.012)	(0.011)	(0.033)
Receives remittances	–0.049*	–0.023	–0.119**
	(0.026)	(0.025)	(0.055)
No savings	0.044**	0.048*	0.054*
	(0.020)	(0.027)	(0.028)
Financial literacy=1	0.004	0.002	–0.002
	(0.018)	(0.016)	(0.028)
Financial literacy=2	0.034	0.019	0.062
	(0.031)	(0.027)	(0.083)
Financial literacy=3	–0.016	–0.03	0.03
	(0.033)	(0.024)	(0.070)
Trust banks	–0.023	–0.018	–0.037
	(0.023)	(0.028)	(0.062)
Risk-averse	–0.037**	–0.039*	–0.040**
	(0.017)	(0.023)	(0.016)
Mortgage	0.029	0.045**	–0.002
	(0.020)	(0.021)	(0.053)
Foreign currency loan	–0.019	0.017	–0.081***
	(0.026)	(0.024)	(0.020)
2–3 further forms of debt	0.025	–0.013	0.096***
	(0.030)	(0.037)	(0.028)
4 or more further forms of debt	0.093***	0.065**	0.143***
	(0.023)	(0.026)	(0.050)
Industry of occupation: public sector/ICT	–0.043***	–0.055***	–0.026
	(0.016)	(0.015)	(0.018)
Industry of occupation: tourism, arts, personal services, other	0.028	–0.004	0.107***
	(0.024)	(0.026)	(0.003)
Mitigating actions: low	0.046***	0.035***	0.078***
	(0.008)	(0.014)	(0.016)
Mitigating actions: medium	0.093***	0.092***	0.071
	(0.024)	(0.016)	(0.059)
Mitigating actions: high	0.131***	0.124***	0.131
	(0.034)	(0.024)	(0.101)
Experienced income shock	0.088***	0.089***	0.089
	(0.023)	(0.018)	(0.065)
Country fixed effects	Yes	Yes	Yes
Log-L	–937.8	–490.2	–418.1
Pseudo-R2	0.23	0.17	0.15
N	2,206	1,498	708
P(DepVar=1)	0.24	0.13	0.47
Source: Authors’ calculations.
Note: Average marginal effects from probit regressions. Standard errors are clustered at the country level. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.

5.2 Differences between borrowers with expired and active moratoria in fall 2020

The increase in the probabilities for the crisis variables is generally higher for active moratoria than for expired moratoria. For instance, in columns 4 to 6, experience of a shock increases the probability of initial take-up by roughly 3 percentage points, while it increases the probability of ongoing moratorium use by 5 percentage points. As expected, when considering the efficiency of the multinomial logit and generalized ordered logit, the results are stronger when ordering is assumed in the model, but also hold in the non-ordered case for the crisis variables. Both Albania and Serbia have particularly high shares of moratorium users with expired moratorium (above 70%) (see chart 1). This is likely due to the design and expiration of the moratoria in these countries, as prolonging existing moratoria or applying for new moratoria was no longer an option. The absence of choice for borrowers dilutes the strength of our analysis in table 3. Repeating estimations without Albania and ­Serbia, the finding that a higher degree of affectedness by COVID-19 and higher debt volumes (indicated by the number of debts and type of loan) contribute to continued moratorium use is confirmed even more clearly. ²⁵ This implies that the repayment behavior and vulnerability of borrowers with expired moratoria might not provide good guidance on what to expect once extended moratoria expire. Large-scale cliff-effects are likely mitigated by the fact that moratorium design largely included an extension of loan terms and no rise in post-moratorium installment rates. Nonetheless, for borrowers with post-pandemic solvency rather than liquidity issues, targeted solutions need to be found (e.g. in the form of loan restructurings) once moratoria expire to avoid the social and economic costs associated with sharply increasing borrower default rates.

Table 3: Determinants of moratorium use  

Dependent variable	No moratorium	Expired moratorium	Active moratorium	No moratorium	Expired moratorium	Active moratorium
Sample	All countries
Model	Multinomial logit			Generalized ordered logit
Mortgage	–0.031	–0.014	0.045***	–0.028	–0.02	0.048***
	(0.020)	(0.017)	(0.015)	(0.018)	(0.013)	(0.018)
Foreign currency loan	0.021	–0.036***	0.014	0.021	–0.038***	0.017
	(0.024)	(0.012)	(0.023)	(0.025)	(0.013)	(0.020)
2–3 further forms of debt	–0,026	0,008	0,019	–0,027	0,01	0,017
	(0.030)	(0.026)	(0.015)	(0.026)	(0.010)	(0.016)
4 or more further forms of debt	–0.096***	0,028	0.069**	–0.083***	0.031***	0.051***
	(0.021)	(0.031)	(0.027)	(0.026)	(0.010)	(0.016)
Mitigating actions: low	–0.047***	0,031	0,016	–0.050***	0.022***	0.028***
	(0.008)	(0.019)	(0.015)	(0.006)	(0.003)	(0.003)
Mitigating actions: medium	–0.092***	0.048***	0,044	–0.089***	0.037***	0.052***
	(0.027)	(0.017)	(0.028)	(0.032)	(0.013)	(0.019)
Mitigating actions: high	–0.130***	0.059***	0.071***	–0.126***	0.050***	0.076***
	(0.036)	(0.020)	(0.025)	(0.040)	(0.016)	(0.025)
Experienced income shock	–0.089***	0.039**	0.050***	–0.083***	0.031***	0.052***
	(0.024)	(0.017)	(0.014)	(0.026)	(0.009)	(0.016)
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Additional controls as in table 2	Yes	Yes	Yes	Yes	Yes	Yes
Log-L	–1,237.5	–1,237.5	–1,237.5	–1,246.2	–1,246.2	–1,246.2
N	2,206	2,206	2,206	2,206	2,206	2,206
Source: Authors’ calculations.
Note: Average marginal effects from multinomial logit regression (columns 1–3) and generalized ordered logit regression (columns 4–6). Standard ­errors are clustered at the country level. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.

5.3 Moratorium use and loan arrears

It is beyond the scope of this paper to provide a full-fledged analysis of the ­determinants of loan arrears. Therefore, the following results should not be taken as a causal analysis and be interpreted with some caution: First, the number of ­observations the analysis is based on is low. Second, we do not attempt to model selection into the credit market and our analysis does not take into account supply affects. Finally, the analysis is based on self-reported moratorium use and self-­reported arrears and may thus be subject to some measurement error. Many factors, e.g. cultural factors and the type of moratorium, likely influence whether borrowers think they are in arrears while actually having postponed payments in agreement with the lender. Most of these limitations, however, will bias estimates downward, which in turn lends some support to using the following results as some preliminary insights.

Over the 12 months preceding the survey, between 10% (Bosnia and Herzegovina) and 40% (Albania) of borrowers reported to have been in loan arrears at least once (chart 2, left panel). When comparing results from 2020 with previous survey waves on loan arrears, we find that the percentage of borrowers in arrears has, on average across countries, increased slightly – a finding which is in contrast to what Cherry et al. (2021) report for the US, where household delinquencies in 2020 are below pre-pandemic levels. Looking at loan arrears which would qualify as “nonperforming loans,” i.e. more than 90 days overdue, the percentages are much lower than the percentages for loan arrears of less than 90 days overdue at 10% in North Macedonia and Albania and just above 1% in Croatia. Loan arrears are correlated with moratorium use (chart 2, right panel). Every second borrower with an active moratorium in fall 2020 reports having been in loan arrears at least once over the past 12 months. For those borrowers who never relied on moratoria the percentage is much lower at 22%.

It is important to remember that borrowers who were in loan arrears before the pandemic were not eligible for moratoria in most countries. Nevertheless, loan arrears are self-reported and it may well be the case that some (although clearly not all, as illustrated by chart 2, right panel) borrowers perceive using the moratorium as equivalent to being in loan arrears.

Table 4 presents selected results on the determinants of loan arrears and the degree of arrears. ²⁶ In line with previous research on the determinants of arrears, income and adverse shocks are significant determinants of loan arrears (Duygan-Bump and Grant, 2009). We further confirm that owing debt to utility providers or nonbank private lenders such as payday lenders is also associated with a higher propensity of loan arrears (Gerardi et al., 2018; Allinger and Beckmann, 2021).

Turning to moratorium use and its association with loan arrears, we find a ­positive and significant correlation with previous but not with current arrears (columns 1 and 2, table 4). To partially address our concerns that moratoria may be perceived as equivalent to arrears, we first reduce the sample to borrowers without active moratoria (columns 1 to 4). However, including them in a second step does not change the results substantially. Compared to borrowers who never relied on moratoria, borrowers with expired moratoria are 16 percentage points more likely to have been in arrears once over the past 12 months and 6 percentage points more likely to have been in arrears twice or more over the past 12 months. But borrowers with expired moratoria were not more likely to be in ­arrears at the time of the survey than those who never took up moratoria. This may indicate that moratoria were indeed successful in tiding borrowers over the worst adverse shocks. For borrowers with expired moratoria there does not seem to be an indication that they are particularly prone to fall into loan arrears. However, the fact that we do not know whether borrowers who did not use the moratorium option but are in arrears were eligible for moratoria or already in arrears before the pandemic is a serious caveat.

Indeed, the positive correlation between previous use of moratoria and past loan arrears could be of concern for policymakers and might be interpreted as a sign that moratoria were in fact not successful in preventing nonperforming loans that emerge due to temporary shocks. We investigated several possible economic explanations for the positive correlation. For example, borrowers may have several loans only some of which may be eligible for moratoria. ²⁷ Based on these additional analyses we conclude that the positive correlation may be a survey response phenomenon, as respondents may perceive moratoria as being equivalent to loan arrears and as both categories are self-reported. Thus, the main message from table 3 seems to be that borrowers with COVID-19 moratoria did not fall into loan arrears when resuming payment after their moratoria had expired (columns 3, 4, 7 and 8). ²⁸

However, table 3 also indicates that borrowers who had exited moratoria ­significantly differ from those who continued to rely on moratoria in fall 2020. Thus, the outlook for loan arrear developments once moratoria expire for all borrowers may yet be less optimistic than the findings summarized in table 4 suggest. Our results are broadly in line with information released by several large banking groups active in CESEE, which noted in their earnings calls for Q4 2020 that ­default rates for loans exiting moratoria in their portfolios were very low (Seeking Alpha, 2021a, 2021b and 2021c). At the time of writing, to our knowledge, regional central banks had not released similar information yet, likely due to the fact that moratoria were still largely ongoing when the fall/winter financial stability reports were being drafted. While central banks only projected moderate increases in nonperforming loans, they highlighted uncertainties related to these projections (see e.g. Czech National Bank, 2020; Hungarian National Bank, 2020).

Table 4: Effect of moratoria on loan arrears  

Dependent variable	Loan arreas in past 12 months: once	Loan arreas in past 12 months: twice or more	Current loan arrears	Current loan arrears: more than 90 days	Loan arreas in past 12 months: once	Loan arreas in past 12 months: twice or more	Current loan arrears	Current loan arrears: more than 90 days
Sample	Borrowers with no or expired moratorium				All borrowers
Active moratorium					0.201***	0.165***	0.152***	0.083***
					(0.037)	(0.034)	(0.027)	(0.023)
Expired moratorium	0.172***	0.072***	0.041	0.002	0.164***	0.064***	0.037	–0.002
	(0.028)	(0.025)	(0.025)	(0.013)	(0.028)	(0.024)	(0.023)	(0.011)
Income shock	0.120***	0.062***	0.036***	0.004	0.130***	0.076***	0.051***	0.017**
	(0.027)	(0.013)	(0.011)	(0.007)	(0.026)	(0.016)	(0.011)	(0.008)
Mortgage	0.006	0.005	0.018	0.015	–0.002	0.007	0.019	0.016
	(0.034)	(0.018)	(0.018)	(0.010)	(0.032)	(0.020)	(0.021)	(0.012)
Foreign currency loan	0.071***	–0.003	0.026*	0.007	0.078***	–0.001	0.031*	0.02
	(0.026)	(0.015)	(0.015)	(0.009)	(0.029)	(0.015)	(0.017)	(0.013)
Debt payday, pawnshop, private, internet loan	0.232***	0.123***	0.072***	0.044***	0.217***	0.102***	0.076***	0.051***
	(0.025)	(0.014)	(0.015)	(0.006)	(0.035)	(0.016)	(0.019)	(0.014)
Utility bill arrears	0.157***	0.088***	0.080***	0.030***	0.153***	0.093***	0.083***	0.041***
	(0.032)	(0.023)	(0.022)	(0.012)	(0.026)	(0.025)	(0.025)	(0.015)
Owe money to family	0.063*	0.050*	0,023	0,012	0.066*	0.053*	0,025	0.024**
	(0.036)	(0.027)	(0.019)	(0.010)	(0.034)	(0.027)	(0.020)	(0.011)
Income: refused answer	–0.045*	–0,019	–0.037**	–0,019	–0,041	–0,004	–0,013	0,01
	(0.027)	(0.012)	(0.017)	(0.016)	(0.025)	(0.022)	(0.025)	(0.020)
Income: low	0.075***	0.030*	0.041**	0.032**	0.088***	0.050***	0.052***	0.040**
	(0.024)	(0.018)	(0.017)	(0.016)	(0.024)	(0.018)	(0.014)	(0.017)
Income: medium	0.034***	0.025***	0.028***	0,006	0.031**	0.026***	0.036***	0,007
	(0.013)	(0.007)	(0.011)	(0.007)	(0.013)	(0.008)	(0.013)	(0.008)
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sociodemographic controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Log-L	–856.5	–540.8	–398	–162.1	–999.2	–682.4	–540.1	–260.8
N	1,925	1,925	1,930	1,693	2,172	2,172	2,179	1,934
P(DepVar=1)	0.24	0.11	0.07	0.03	0.27	0.13	0.1	0.05
Source: Authors’ calculations.
Note: Average marginal effects from probit regressions. Standard errors are clustered at the country level. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.

5.4 Robustness analyses

As pointed out above, we estimated all probit models also as Heckman probit models to check for selection bias. Using Heckman selection models, the magnitude and significance of sociodemographic variables is hardly affected. This also holds for the variables indicating affectedness by the crisis and experiencing adverse shocks. Regarding indicators of indebtedness, we find that results from a Heckman selection model indicate a slightly stronger and highly significant effect.

To focus on the main variables of interest, we only presented one specification for all models with respect to control variables. Some of these control variables are correlated and might introduce multicollinearity issues to our estimates. Table A4 in the annex presents a specification where we drop financial literacy, personal characteristics and beliefs and specify socioeconomic characteristics in a more parsimonious manner. In addition to the specification presented in table A4 we repeat estimations including only income or only labor market status as controls. Results for the variables on loan characteristics or the COVID-19 impact do not change in terms of sign or significance. In alternative specifications we also address the concern of network effects. Guiso et al. (2013) find that knowing people who defaulted strategically increases the propensity for strategic default. The question on moratoria allows us to control for network effects. We find that knowing other moratorium users increases borrowers’ propensity to use moratoria by 19 percentage points. When we differentiate between active and expired moratoria (table 3), knowing other moratorium users increases the likelihood of reporting an “expired moratorium” by 6 percentage points and the likelihood of reporting an “active moratorium” by 9 percentage points. Controlling for network effects does not change our baseline results. ²⁹ Network effects regarding moratoria do not have an effect on loan arrears.

Regarding our generalized logit estimation, some of the probabilities estimated are below zero. As this concerns few observations and we provide the multinominal logit results as a comparison, we are not too concerned by this fact. For the multinominal logit we tested the independence from irrelevant alternatives assumption (IIA) – the test developed by Weesie (2000) and the Small-Hsiao test (Small and Hsiao, 1985) do not reject the null hypothesis that the IIA assumption holds. The test by Hausman and McFadden (1984) also does not reject the null hypothesis, but has a negative test statistic for one outcome, which is not uncommon and likely the result of some shortcomings of this test. In either case, the similarity of the estimation results across the two models (and also the ordered logit) makes us confident that our conclusions regarding the difference between borrowers with active moratoria and borrowers with expired moratoria at the time of the interview are sound.

6 Conclusion

Our study aims to close this gap. We show evidence on moratorium use collected in the 2020 wave of the OeNB Euro Survey in the CESEE-10 countries. Our analytical results on moratorium use highlight substantial differences between the opt-in and opt-out countries. In opt-in countries, eligibility was often tied to pandemic-related constraints, so that being adversely affected by the crisis significantly increases moratorium use. In the opt-out countries, certain sociodemographic variables, such as the sector of employment, the condition of respondents’ residence and the number of debt instruments held, are strong predictors, while the impact of the crisis on individual borrowers does not play a role. For the full sample of countries, we further find that borrowers whose moratoria had expired by fall 2020 were able to resume loan repayments and did not fall into arrears. However, borrowers whose moratoria were still active in fall 2020 were more ­affected by the pandemic and also held more debt than borrowers with expired moratoria. This finding also indicates that the encouraging result of borrowers not developing repayment problems after exiting moratoria may not hold for those borrowers who relied on moratoria for an extended period of time.

Our results provide some preliminary insights for policymakers. It is reassuring that moratoria indeed reached borrowers who were more vulnerable and less resilient, contributing to reducing a surge in loan defaults during 2020. At the same time, ongoing moratorium use could be an indicator for solvency rather than ­liquidity issues. Thus, it would likely be overly optimistic to assume that results from borrowers who exited moratoria could be extended to borrowers with active moratoria. Going forward, banks’ and policymakers’ focus will need to shift from short-term liquidity support to sustainable restructuring solutions for struggling borrowers. Lessons from the global financial crisis of 2008 should be heeded. For instance, McCann et al. (2020) use the example of mortgage distress in Ireland after the global financial crisis to illustrate the importance of long-term sustainable restructurings rather than short-term fixes. They point toward a clearly positive correlation between deeper up-front repayment cuts and the likelihood of a successful restructuring. Amromin et al. (2020) reviewed the US experience after the global financial crisis and point out that there are many ways to ease borrower distress with differing costs to borrowers and lenders: reducing payment-to-income ratios, forbearance, loan refinancing and debt forgiveness. Besides, they also point out that borrower assistance should not be over-engineered and excessively complex, and that social safety nets and income support can be important complementary factors to ensure debt sustainability in the medium and long term.

Our results also point out some interesting avenues for future research: While valuable and a good starting point, our results strongly highlight the need for collecting further and more detailed evidence, for example, on debt overhang and especially on moratorium-induced debt overhang. Cross-country surveys might not be the ideal tool to collect hard data on debt overhang. However, survey data may provide insights on how moratoria affected the relationship between lenders and borrowers. Survey data may especially provide valuable insights into whether government-led debt relief intervention could have adverse effects going forward, such as creating moral hazard problems or inducing credit constraints. With a history of banking crises during transition in some of the CESEE-10 countries, government-led intervention in debt relief may have a strong impact on trust in banks. Our results also indicate that the country-specific differences in moratorium design matter as well. Future research may provide insights into whether opt-in or opt-out moratoria proved to be more successful in achieving the different policy goals related to borrower welfare and financial stability. Last but not least, unwinding will have to be country-specific and would benefit from case studies for individual countries.

References

Agarwal, S., G. Amromin, I. Ben-David, S. Chomsisengphet, T. Piskorski and A. Seru. 2017.
Policy intervention in debt renegotiation: Evidence from the Home Affordable Modification Program. In: Journal of Political Economy 125(3). 654–712.

Allinger, K. and E. Beckmann. 2021. Prevalence and determinants of nonbank borrowing in CESEE: evidence from the OeNB Euro Survey. In: Focus on European Economic Integration Q1/21. OeNB. 7–35.

Amromin, G., J. Dokko and K. Dynan. 2020. Helping homeowners during the Covid-19­ ­pandemic: Lessons from the Great Recession. Chicago Fed Letter 443.

Association of Serbian Banks. 2020. The Second Moratorium Worth Almost One Billion ­Euros. Association of Serbian Banks Press Release. August 21.

Beckmann, E., J. Fidrmuc and H. Stix. 2012. Foreign Currency Loans and Loan Arrears of Households in Central and Eastern Europe. OeNB Working Papers 181.

Brant, R. 1990. Assessing Proportionality in the Proportional Odds Model for Ordinal Logistic Regression. In: Biometrics 46(4). 1171–1178.

Cherry, S. F., E. X. Jiang, G. Matvos, T. Piskorski and A. Seru. 2021. Government and private household debt relief during Covid-19. NBER Working Paper 28357.

Costa, S. and L. Farinha. 2012. Households’ indebtedness: a microeconomic analysis based on the results of the households’ financial and consumption survey. In: Financial Stability Report. Banco de Portugal.

Croatian National Bank. 2020. Macroprudential Diagnostics: third quarter of 2020. In: Croatian National Bank Macroprudential Diagnostics V/12. October.

Czech National Bank. 2020. Risks to financial stability and their indicators – December 2020.

Duygan-Bump, B. and C. Grant. 2009. Household debt repayment behavior: What role do institutions play? In: Economic Policy 24(57). 107–140.

EBA. 2020. First evidence on the use of moratoria and public guarantees in the EU banking sector. Thematic note. EBA/Rep/2020/31.

EIB. 2021. CESEE Bank Lending Survey. Spring.

EIB. 2020. CESEE Bank Lending Survey. Autumn.

Fischer, A. M. and P. Yesin. 2019. Foreign currency loan conversions and currency mismatches. CEPR Discussion Papers 13923

Gerardi, K., K. F. Herkenhoff, L. E. Ohanian and P. S. Willen. 2018. Can’t pay or won’t pay? Unemployment, negative equity, and strategic default. In: The Review of Financial Studies 31(3). 1098–1131.

Guiso, L., P. Sapienza and L. Zingales. 2013. The Determinants of Attitudes towards Strategic Default on Mortgages. In: Journal of Finance 68(4). 1473–1515.

Guiso, L. and P. Sodini. 2013. Household Finance: An Emerging Field. In: Stulz, M. (ed.) Handbook of the Economics of Finance. Elsevier. 1397–1532.

Haughwout, A. F., D. Lee, J. Scally and W. van der Klaauw. 2020. Following Borrowers through Forbearance. Federal Reserve Bank of New York Liberty Street Economics. November 17.

Hausman, J. A. and D. L. McFadden. 1984. Specification tests for the multinomial logit model. In: Econometrica 52(5). 1219–1240.

Heckman, J. J. 1979. Sample Selection Bias as a Specification Error. In: Econometrica 47(1). 153–161.

Henderson, J. V., A. Storeygard and D. N. Weil. 2012. Measuring economic growth from outer space. In: American Economic Review 102(2). 994–1028.

Hungarian National Bank. 2021. Financial Stability Report. June.

Hungarian National Bank. 2020. Financial Stability Report. November.

Long, J. S. and J. Freese. 2014. Regression Models for Categorical Dependent Variables Using Stata, 3rd Edition. Stata Press.

Lusardi, A. and P. Tufano. 2015. Debt literacy, financial experiences, and overindebtedness,
In: Journal of Pension Economics and Finance 14(4). 332–368.

McCann, F. and T. O’Malley. 2020. Resolving mortgage distress after COVID-19: some lessons from the last crisis. In: Central Bank of Ireland Financial Stability Notes 2020/7.

Mocetti, S. and E. Viviano. 2017. Looking behind mortgage delinquencies. In: Journal of Banking and Finance 75. 53–63.

Nguyen, C. H. 2007. Access to credit and borrowing behaviour of rural households in a Transition Economy. In: FAO International Conference on Rural Finance Research. March 19–21.

Polish National Bank. 2020. Financial stability report. December.

Piskorksi, T. and A. Seru. 2018. Mortgage market design: Lessons from the Great Recession. Brookings Paper on Economic Activity. 429–513.

Republic of North Macedonia Ministry of Finance. Economic Reform Program 2021–2023.
https://finance.gov.mk/wp-content/uploads/2021/01/Economic-Reform-Programme-2021-2023.pdf

Romanian National Bank. 2020. Financial stability report. June.

Seeking Alpha. 2021a. UniCredit SpA (UNCFF) CEO Jean-Pierre Mustier on Q4 2020 Results – Earnings Call Transcript. February 11.

Seeking Alpha. 2021b. Raiffeisen Bank International AG (RAIFF) CEO Johann Strobl on Q4 2020 Results – Earnings Call Transcript. March 12.

Seeking Alpha. 2021c. Erste Group Bank AG (EBKOF) CEO Bernd Spalt on Q4 2020 Results – Earnings Call Transcript. February 26.

Small, K. A. and C. Hsiao. 1985. Multinomial Logit Specification Tests. In: International Economic Review 26. 619–627.

Weesie, J. 2000. Seemingly unrelated estimation and the cluster-adjusted sandwich estimator.
In: Stata Technical Bulletin 9.

Williams, R. 2006. Generalized Ordered Logit/Partial Proportional Odds Models for Ordinal Dependent Variables. In: The Stata Journal 6(1). 58–82.

Wolfe, R. 1997. OMODEL: Stata modules to perform tests on ordered probit and ordered logit models. Statistical Software Components S320901. Boston College Department of Economics.

Annex

Table A1: Variable definitions  

Variable	Definition
Female	Dummy variable that is 1 for female respondents, else zero.
Size of household	Number of household members permanently living in the household, including household members that are ­temporarily absent (e.g. students or persons doing military service).
Manages household finances	Dummy variable that is 1 for respondents who state they are either personally or together with someone else in charge of managing household finances.
Labor market status (employed, self-employed, unemployed, retired)	Dummy variable coded as 1 if respondent belongs to the selected occupational category. Omitted category: ­employed. Students are excluded from the sample.
Income (high, medium, low, refused answer)	Dummy variables that take the value 1 for each net household income tercile (high, medium, low). Sample values are used to construct terciles. For those respondents who did not give an answer, an additional dummy variable is defined (income refused).
Condition of residence: poor	Dummy variable defined by interviewer based on the answer to the following question “Could you describe the ­condition of the dwelling? Excellent and well maintained; good, needs some minor repairs; poor, needs major work; very poor, some walls, ceilings need replacement.” Categories “poor, needs major work” and “very poor, some walls, ceilings need replacement” defined as 1, else zero.
Own secondary residence	Dummy variable that takes the value 1 if the respondent or someone living in the same household owns a secondary residence.
Financial literacy	Categorical variable ranging from 0 to 3 depending on the number of correct answers to the following 3 questions: (1) Suppose you had 100 [LOCAL CURRENCY] in a savings account and the interest rate was 2% per year. ­Disregarding any bank fees, how much do you think you would have in the account after 5 years if you left the ­money to grow: more than 102, exactly 102, less than 102 [LOCAL CURRENCY]? / Don’t know / No answer. (2) Suppose that the interest rate on your savings account was 4% per year and inflation was 5% per year. Again disregarding any bank fees – after 1 year, would you be able to buy more than, exactly the same as, or less than today with the money in this account? / Don’t know / No answer. (3) Suppose that you have taken a loan in EURO. Then the exchange rate of the [LOCAL CURRENCY] depreciates against the EURO. How does this change the amount of local currency you need to make your loan installments? The amount of local currency… increases / stays exactly the same / decreases / Don’t know / No answer. “Don’t know” responses are counted as incorrect answers, “No answer” responses missing.
Industry of occupation	Dummy variables based on a question regarding industry of occupation. The original question covers 16 industries, which we group into 3 categories that are little, moderately and highly affected by lockdowns. Industry of occupation: ICT, public sector: Dummy variable that takes the value 1 if respondent’s profession is in (i) Banking, financial and insurance activities, (ii) Information and communication technology, (iii) Professional, ­ scientific, technical, administrative and maintenance services, (iv) Education (kindergarten, school, university, etc.), (v) Human health care, nursing and social work activities, (vi) Public administration, justice, police, defence, trade unions, religious organizations, else zero. Industry of occupation: tourism, arts, personal services, other: Dummy variable that takes the value one if the ­respondent’s profesison is in (i) Tourism, accommodation, restaurant, café, bar, (ii) Arts, culture, entertainment and recreation, (iii) Personal services (hairdresser, beauty treatment, ­funeral, etc.), (iv) other sector, else zero.
Source: OeNB Euro Survey.

Table A1 continued: Variable definitions  

Variable	Definition
Receives remittances	Dummy variable that takes the value 1 if the respondent receives remittances from abroad.
No savings	Dummy variable that takes the value 1 if the respondent does not have any savings, zero if respondent has savings.
Trust banks	Dummy variable based on the following question “Please tell me how much trust you have in the following institutions: (…) domestically owned banks (…) foreign owned banks (...). For each of the institutions, please tell me if you tend to trust it or tend not to trust it. 1 means “I trust completely,” 2 means “I somewhat trust,” 3 means “I neither trust nor distrust,” 4 means “I somewhat distrust” and 5 means “I do not trust at all.” Answers 1 and 2 are coded as 1, else zero.
Risk averse	Dummy variable based on the following question “Please tell me whether you agree or disagree on a scale from 1 (strongly agree) to 6 (strongly disagree) with the following statement: “In financial matters, I prefer safe investments over risky investments.” Answers “strongly agree” and “agree” are defined as “risk averse”, else zero.
Mortgage	Dummy variable that takes the value 1 if the respondent has a mortgage, zero otherwise (consumption loan, loan for education or other purposes).
Foreign currency loan	Dummy variable that takes the value 1 if the respondent has a loan denominated in foreign currency, zero if ­respondent has a loan denominated in local currency.
# further forms of debt	Categorical variable ranging from 0 to 3 depending on the number of debts the respondent owes from the following list: overdraft, bank loan, credit card debt, purchase from a store or company using installment credit, leasing ­contract, a utility provider by delaying payment of bills, internet loan, payday loan, pawnshop, money owed to ­employer, money owed to family, relatives or friends, money owed to another private lender, other debt. Categories are defined as: 0 debt owed, 1 debt owed, 2-3 debts owed, 4 or more debts owed. Base category: 1 debt owed (Sample does not include respondents without debt).
Debt payday, pawnshop, private, internet loan	Dummy variable that takes the value 1 if respondent owes money to a payday lender, pawnshop, private lender or ­internet loan provider, else zero.
Utility bill arrears	Dummy variable that takes the value 1 if respondent owes money to a utility provider, else zero.
Owe money to family	Dummy variable that takes the value 1 if respondent owes money to family or friends, else zero.
Mitigating actions	Categorical variable ranging from 0 to 3 depending on the number of mitigating actions the respondent had to take from the following list: reduce amount spent on everyday expenses, reduce or postpone larger expenditures, reduce money set aside for savings, reduce help to friends or relatives whom I helped before, utilize savings or sold possessions, take out a loan from a bank, over-draft bank account, borrow money family and friends or from another source, delay payment of rent or other bills, forced to move. 1-2 mitigating actions are defined as “low”; 4 mitigating actions are defined as “medium”; 5 or more mitigating actions are defined as “high”. Base category: zero mitigating ­actions.
Experienced income shock	Dummy variable that takes the value 1 if respondents had to reduce work hours and received a reduced salary, were laid off from a job or if households experienced an unexpected significant reduction of their income over the past 12 months; zero if none of the above apply.
Source: OeNB Euro Survey.

Table A2: Summary statistics  

	Minimum	Maximum	Mean
Variable			Borrowers	Moratorium users	Moratorium users: opt-in countries	Moratorium users: opt-out countries	Expired moratoria	Active moratoria
Female	0	1	0.49	0.53	0.54	0.52	0.53	0.51
Size of household	1	5	3.08	3.19	3.03	3.30	3.25	3.13
Manages household finances	0	1	0.88	0.89	0.88	0.90	0.90	0.88
Unemployed	0	1	0.10	0.11	0.13	0.09	0.10	0.13
Self-employed	0	1	0.11	0.11	0.17	0.06	0.09	0.13
Retired	0	1	0.11	0.05	0.04	0.06	0.05	0.06
Income: refused answer	0	1	0.22	0.22	0.18	0.24	0.22	0.19
Income: low	0	1	0.15	0.13	0.18	0.10	0.12	0.16
Income: medium	0	1	0.28	0.31	0.34	0.29	0.32	0.29
Condition of residence: poor	0	1	0.09	0.12	0.10	0.13	0.12	0.10
Own secondary residence	0	1	0.10	0.10	0.12	0.09	0.09	0.11
Financial literacy=1	0	1	0.20	0.18	0.18	0.19	0.19	0.18
Financial literacy=2	0	1	0.31	0.36	0.37	0.34	0.33	0.38
Financial literacy=3	0	1	0.34	0.33	0.26	0.38	0.34	0.30
Works in public sector/ICT	0	1	0.30	0.26	0.21	0.30	0.28	0.25
Works in tourism, arts, personal services, other	0	1	0.13	0.17	0.19	0.16	0.16	0.18
Receives remittances	0	1	0.09	0.06	0.09	0.05	0.05	0.08
No savings	0	1	0.53	0.61	0.56	0.65	0.63	0.61
Trust banks	0	1	0.34	0.31	0.30	0.32	0.33	0.28
Risk averse	0	1	0.66	0.66	0.58	0.72	0.68	0.65
Mortgage	0	1	0.48	0.48	0.49	0.47	0.44	0.55
Foreign currency loan	0	1	0.24	0.25	0.22	0.28	0.26	0.24
2–3 further forms of debt	0	1	0.43	0.44	0.39	0.47	0.47	0.40
4 or more further forms of debt	0	1	0.12	0.21	0.25	0.17	0.19	0.22
Debt payday, pawnshop, private, internet loan	0	1	0.10	0.13	0.19	0.08	0.10	0.17
Utility bill arears	0	1	0.07	0.14	0.10	0.16	0.15	0.13
Owe money to family	0	1	0.13	0.17	0.22	0.14	0.15	0.20
Mitigating actions: low	0	1	0.34	0.29	0.25	0.33	0.31	0.27
Mitigating actions: medium	0	1	0.28	0.35	0.37	0.34	0.37	0.34
Mitigating actions: high	0	1	0.15	0.26	0.32	0.22	0.23	0.30
Experienced income shock	0	1	0.41	0.57	0.74	0.46	0.56	0.60
Source: Authors’ calculations.

Table A3: Public information on moratorium use by households in CESEE  

Source	OeNB Euro Survey	EBA (2020)	EIB (2020)	EIB (2021)	Central banks and banking associations1
Indicator	% of respondents with loan	% of household loan portfolio	% of affected portfolio based on answers from banks in intervals – average = share of banks in ­interval multiplied with mid-point of interval		Varying
Period	Sept.-Oct. 2020	June 30, 2020	Sept./Oct. 2020	March/April 2021	Varying
Reporting entity	Individuals in CESEE	1–3 large banks per country	15 international banking groups and 85 (EIB, 2020)/ 90 (EIB, 2021) local subsidiaries/ banks (~ 50% of regional banking assets)		Varying
Bulgaria	19.7%	7% (mortgages: 7.6%)	Average: 10.1%	Average: 9%
Czechia	17.8%	Not included	Average: 15%	Average: 10%	As of October 31, 2020: 15% of household loan portfolio (Czech National Bank, 2020)
Hungary	29.7%	21.9% (mortgages: 25.3%)	Average: 47.5%	Average: 41%	As of June 2020: 60% of households with loans, and ~50% of household loan portfolio (Hungarian National Bank, 2020)
Croatia	9.3%	7.1% (mortgages: 8.1%)	Average: 7%	Average: 5%	As of August 31, 2020: 6.7% of household loan portfolio (Croatian National Bank, 2020)
Poland	17.1%	8.5% (mortgages: 8.5%)	Average: 15%	Average: 18.5%	As of end Q3: 5% of household loan portfolio (Polish National Bank, 2020)
Romania	9%	13.0% (mortgages: 8.9%)	Average: 13.5%	Average: 13.7%	As of June 5, 2020: ~8% of households with loans (Romanian National Bank, 2020)
Albania	19.9%	Not included	Average: 22.5%	Average: 17.5%
Bosnia and Herzegovina	6.3%	Not included	Average: 15.5%	Average: 5%
North Macedonia	67%	Not included	Average: 21.5%	Average: 11.5%	As of August 2020: terms eased for 54.8% of household loan portfolio (mainly postponed repayment) (Republic of North Macedonia Ministry of Finance, 2021)
Serbia	68.4%	Not included	Average: 87%	Average: 70%	As of mid-August 2020: take-up of second moratorium: 82% of households (including moratoria on credit cards, overdrafts) (Association of Serbian Banks, 2020)
Source: Authors’ compilation.
1 Only information where household loans were reported separately and either as a percentage of a portfolio or a percentage of loan holders is included.

Table A4: Robustness: determinants of moratorium use  

Dependent variable	Moratorium use
Sample	All countries	Opt-in countries	Opt-out countries
Female	0.024	0.037**	–0.001
	(0.016)	(0.015)	(0.036)
Size of household	–0.005	–0.014**	0.01
	(0.008)	(0.007)	(0.027)
Manages household finances	0.024	–0.012	0.108***
	(0.029)	(0.031)	(0.011)
Employed	0.056*	0.031	0.113*
	(0.029)	(0.025)	(0.063)
Income: refused answer	0.005	0.049*	–0.066
	(0.038)	(0.026)	(0.069)
Income: low	–0.02	–0.002	–0.097
	(0.043)	(0.021)	(0.161)
Income: medium	0.009	0	0.033
	(0.017)	(0.020)	(0.032)
Condition of residence: poor	–0.006	–0.056**	0.093**
	(0.029)	(0.024)	(0.043)
No savings	0.051***	0.049**	0.062***
	(0.017)	(0.022)	(0.024)
Mortgage	0.034*	0.043**	0.016
	(0.018)	(0.017)	(0.055)
Foreign currency loan	–0.021	0.011	–0.075***
	(0.022)	(0.023)	(0.027)
2-3 further forms of debt	0.018	–0.011	0.084***
	(0.028)	(0.037)	(0.020)
4 or more further forms of debt	0.087***	0.068***	0.139***
	(0.023)	(0.024)	(0.048)
Industry of occupation: public sector/ICT	–0.042**	–0.056***	–0.025
	(0.017)	(0.016)	(0.028)
Industry of occupation: tourism, arts, personal services, other	0.034	0.005	0.096***
	(0.024)	(0.028)	(0.010)
Mitigating actions: low	0.051***	0.033**	0.096***
	(0.009)	(0.015)	(0.028)
Mitigating actions: medium	0.090***	0.089***	0.074
	(0.022)	(0.020)	(0.059)
Mitigating actions: high	0.145***	0.131***	0.146
	(0.030)	(0.019)	(0.103)
Experienced income shock	0.095***	0.099***	0.086
	(0.022)	(0.018)	(0.063)
Country fixed effects	Yes	Yes	Yes
Log-L	–1,000.1	–532.9	–441.8
Pseudo-R2	0.22	0.15	0.14
N	2,336	1,592	744
P(DepVar=1)	0.24	0.13	0.47
Source: Authors’ calculations.
Note: Average marginal effects from probit regressions. Standard errors are clustered at the country level. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.

What do people in CESEE think about
public debt?

Markus Eller, Branimir Jovanovic, Thomas Scheiber ³⁰

This paper investigates public attitudes toward public sector debt in ten Central, Eastern and Southeastern European (CESEE) countries. Unique data from a special module of the 2018 OeNB Euro Survey wave indicate that people in CESEE have, in general, negative attitudes toward public debt. Most respondents believe that higher public debt compromises the opportunities of future generations and implies higher taxes or lower social benefits in the future. Beliefs that higher public debt allows for higher investments today are also widespread but less common. Econometric analysis reveals that wealthier individuals and more disadvantaged ­societal groups (particularly people who have experienced economic hardship) tend to be most concerned about public debt. The finding that the worse off are more debt averse contrasts with existing studies for advanced economies that have found that it is primarily the better off who are more skeptical. This difference may be explained by the comparatively lower level of social spending and the predominance of regressive tax systems in CESEE, which could make disadvantaged groups of society believe that the burden of higher debt must eventually be shouldered by them.

JEL classification: C42, D78, E62, H63, P35

Keywords: public debt, public preferences, survey data, CESEE

Given the increased role of public indebtedness in the region (already before the COVID-19 pandemic), this paper studies how individuals tend to perceive public debt and whether their beliefs reflect their socioeconomic background, economic hardship experiences, financial literacy levels or political attitudes. Gaining such insights is important also from a macroeconomic perspective. After all, the profile of individuals who are more skeptical or supportive about public debt is meaningful information for policymakers, as the macroeconomic stabilization function of fiscal policy during a crisis or the political feasibility of post-crisis fiscal consolidation rely on people’s trust in the government’s capacity to manage taxpayer money conscientiously. This reasoning is backed by a series of papers that have studied household-level responses to fiscal policy shocks to improve our understanding of fiscal multipliers (e.g. Johnson et al., 2006; Shapiro and Slemrod, 2009; D’Acunto et al., 2018), that have identified support of individuals as an important factor for successful fiscal consolidation (e.g. Abbas et al., 2010; Ardanaz et al., 2020), or that have more generally studied the impact of public trust in government on macroeconomic outcome variables (e.g. Zak and Knack, 2001; Horvath, 2013; Sangnier, 2013).

The literature using household survey data to determine people’s attitudes ­toward fiscal policy measures and public indebtedness has grown in recent years. The reviewed papers (see the next section) focus mainly on single countries rather than employing cross-country surveys. However, cross-country surveys are important to ensure measurement equivalence for a larger set of countries and to detect correlations with individual attitudes that go beyond country-specific institutional and cultural factors. Therefore, the 2018 wave of the OeNB Euro Survey included a series of questions aimed at revealing people’s attitudes toward public debt and public sector governance. The paper at hand builds on this survey information and investigates the views people shared about public debt ³³ and its effects, addressing several individual-specific factors that are related to expressing such views. While country-specific factors are controlled for by fixed effects, we exploit information from the whole sample rather than focusing on the explanation of cross-country differences in this paper. ³⁴

The remainder of the paper is organized as follows: Section 1 reviews the literature on survey-based papers that examined individual perceptions of fiscal policies in general and public debt in particular. Section 2 introduces the data and provides a descriptive analysis of individuals’ public debt attitudes. Section 3 guides through the econometric analysis, studying the background of people who expressed particular attitudes on debt. Finally, section 4 concludes and discusses some policy implications.

1 Related literature

Pitlik et al. (2011) investigate whether opinions on fiscal policy issues are driven more by ideology or self-interest, using a telephone survey of around 1,000 Austrian voters conducted in 2008. They find that perceived self-interest – the expectation that a proposed policy measure may have personally adverse consequences – is at least as important for the acceptance of policy measures as ideological conviction.

Stix (2013) investigates the determinants of preferences for public debt consolidation in Austria, using a survey conducted on 2,000 randomly selected voters in 2010. He investigates the relevance of several factors, including self-interest, intergenerational distributional preferences, preferences for distributional fairness among the current generation and the credibility of medium-term fiscal policy plans. He finds that all these factors matter, which has clear implications for the design of debt consolidation plans.

Heinemann and Henninghausen (2012) investigate determinants of public debt attitudes in Germany, using a representative survey of 1,000 respondents conducted in 2011. They consider several explanatory variables, such as individual credit constraints, intergenerational considerations (having children), ideology and trust. They find that Ricardian motives and ideology and trust are important for explaining the individual heterogeneity of debt consolidation preferences.

Hayo and Neumeier (2019) investigate determinants of individual attitudes toward taxes, public debt and public spending in Germany, using a representative survey conducted on 2,000 respondents in 2013. They find that economic well-being, confidence in politicians, economic knowledge, and time and party preferences are all statistically significantly related to public finance preferences. According to these results, a stronger preference for fiscal consolidation can be identified for respondents that are particularly forward-looking, well informed about debt-related economic variables or better off financially.

Roth et al. (2020) examine how beliefs about the government debt-to-GDP ratio affect people’s attitudes toward government spending and taxation, using a series of experiments on more than 4,000 respondents from the USA. They find that most people underestimate the debt ratio and turn less supportive about government spending once they are made aware of the actual amount of debt, but do not substantially alter their attitudes toward taxation.

If we expand our focus, we can also refer to a large related body of the recent political economy literature. One group of literature focused on the impact of fiscal austerity on public opinion and voting choices, providing evidence mostly for advanced economies (e.g. Arias and Stasavage, 2019; Hübscher et al., 2020; Kalbhenn and Stracca, 2020; Bansak et al., 2021) and only rarely for emerging market economies (e.g. Ardanaz et al., 2020, for Latin American economies). Another group of studies investigated the acceptance of fiscal rules by politicians (incumbents versus opposition) using survey methods; several of them focus on Germany (e.g. Heinemann et al., 2020; or Blesse et al., 2021). This type of literature is of interest for our research question in so far as it can be expected that people who support debt rules are also more likely to express debt-averse perceptions.

While it is obvious that there is already some literature on public attitudes toward public sector debt, most of the existing studies refer to advanced economies. However, to the best of our knowledge, no study has so far investigated this issue for CESEE countries. Our study aims to fill this gap. Moreover, the cross-country sample we study is larger than that of the reviewed papers, as it includes respondents from ten different countries and a variety of survey questions, allowing us to investigate a broad set of factors that are potentially related to individual public finance attitudes.

2 Data and variables

The survey questionnaire elicits a rich set of information on socioeconomic characteristics, indicators of wealth and finances, individual beliefs, expectations and trust. The questionnaire is composed of a core set of questions regarding the extent of euroization and individuals’ loan and saving decisions. In the 2018 survey wave, we added a special module of questions aimed at capturing individual attitudes toward public sector governance, respondents’ knowledge and perception of public debt, and their preferences with regard to public spending priorities (building on Hayo and Neumeier, 2019; and Stix, 2013). Lastly, we extended the survey data by merging satellite nightlight data at the level of the primary sampling unit as a proxy for local economic activity (Henderson et al., 2012).

The key variable for our analysis, which captures individuals’ views of public debt and its impact, is based on the following survey question:

The idea behind the question is to find out to what extent respondents agree with different statements about the implications of public debt. The first two statements expressing the belief that higher public debt compromises the chances of future generations and allows for higher investments today address the more general impact of public debt, while the views that higher debt implies higher taxes and lower pensions/benefits in the future (third and fourth statement) capture individual-­level affectedness. People’s perception about the benefit-related impact is also relevant for considering the role of implicit public debt, i.e. obligations for future government expenditures resulting from current regulations. As a caveat, the four statements are not well-balanced in the sense that positive and negative statements are represented equally strongly: With reference to debt, only one statement put to respondents in the survey carries a positive connotation (“makes it possible to conduct necessary investments today”), while the other three statements are framed negatively. This may prime respondents to adopt a more critical position on public debt. However, before confronting interviewees with these statements, the survey introduced an analogy between public debt and individual debt (see above). Moreover, not all the negatively framed items are blocked together. Finally, the six-point Likert response scale is intended to reduce the likelihood of choosing neutral options and thus ­encourage respondents to make a clear choice between agreeing and disagreeing.

Chart 2 shows the (weighted) percentage of respondents who agree or strongly agree with the four statements on public debt and its effects. Based on unconditional averages across all observations in the sample (see the “CESEE” bars), nearly 70% of respondents agreed that higher public debt compromises the opportunities of future generations (north-west panel). In contrast, only about 40% of respondents agreed that higher public debt levels allow for higher investments today, ranging from a share of 30% in Croatia to 55% in North Macedonia (north-east panel). Finally, roughly 60% of respondents agree that higher public debt implies paying more taxes and receiving lower social benefits in the future (southern panels). The three panels with the skeptical views on public debt exhibit a similar ranking of countries: The countries with the lowest shares of skepticism are Czechia and Albania, while the highest shares can be found in Bulgaria, Croatia, Romania and North Macedonia. ³⁶

Interestingly but not surprisingly, knowledge of the size of public debt as a percentage of GDP is rather limited in the investigated CESEE countries (see chart 3). On average, 36% of respondents provide a “don’t know” answer, ranging from 16% in Albania to even around 60% in Bulgaria and Romania. The share of respondents giving appropriate answers about the debt ratio varies from 17% in Hungary and Albania to about 40% in Croatia and North Macedonia (indicated by the patterned bar sections in the chart). ³⁷

When quality-checking the investigated survey data, we noticed some straight-lining of answers, i.e. a certain tendency among respondents to provide the same answers in blocks of questions, or a certain failure to adequately differentiate between different items in the questionnaire. In the case of the four questions on public debt attitudes shown in chart 2, straight-liners accounted for nearly 25% of respondents. The prevalence of straight-lining introduces some measurement errors. However, the background analysis provided in an online supplement shows that this is an issue predominantly in Albania and, to a lesser extent, in Hungary, where we identified some major cases of multiple straight-lining. ³⁸ For an analysis of the extent to which our results are affected by the inclusion of (major) straight-­lining cases, see the robustness analysis (section 3.3).

3 Econometric analysis of factors associated with people’s attitudes toward public debt

3.1 Econometric set-up and research hypotheses

As for the explanatory variables, we follow the literature overviewed in section 1 and include several groups of variables. ³⁹ The first group refers to general sociodemographic and socioeconomic characteristics of individuals and households (XSoc). Some of these characteristics are standard controls used in the related literature. It is noteworthy that we also include wealth indicators here to test the hypothesis whether wealthier individuals have a more critical view of public debt as they might fear – in a Ricardian manner – that their assets will be negatively affected in the future by higher debt levels, e.g. via higher taxes, higher inflation or a sovereign debt crisis.

The second group of variables captures individuals’ general attitudes (XGen) that are usually included as explanatory variables in a debt-related context. Following Heinemann and Henninghausen (2012), Stix (2013), or Hayo and Neumeier (2019) for instance, we include variables that measure time preference, discomfort about owing money, and general trust in institutions, expecting that those who have a stronger preference for the present, who are willing to incur debt themselves or are ­generally more trustworthy are also more inclined to see public debt in a positive light.

The third class of explanatory variables captures economic hardship experiences (XPer). Here, the basic hypothesis would be that those who are/were subject to economic hardship are likely more skeptical with regard to the crisis mitigation capacities of government (see Denisova et al., 2010; Eller and Scheiber, 2020) and therefore are also more critical in their assessment of public debt. In line with ­existing literature (e.g. Heinemann and Henninghausen, 2012; Stix, 2013; Hayo and Neumeier, 2019; Roth et al., 2020), we include as economic hardship variables unemployment, the necessity of basic consumption cutbacks in the past, net household income, local economic activity, respondents’ health status and access to emergency borrowing.

The fourth group of variables measures financial literacy and knowledge of public debt (XFinlit). There is already a broad literature showing that financial literacy influences the borrowing behavior of individuals (e.g. Beckmann and Stix, 2015; or Lusardi and Tufano, 2015). Analogously, financial literacy might also be relevant for people’s views about public debt. Persons who understand, or are better informed about, the cost of public indebtedness may also be more skeptical as regards the future burden of public debt (as argued for public spending preferences in Hayo and Neumeier, 2019). As empirical proxies, we include interest in politics, a financial literacy dummy, knowledge of the size of public debt, and backward-oriented debt concerns.

The fifth and final group of variables captures individuals’ general political ­attitudes (XPol), as included in several of the surveyed studies. In essence, a more positive view of current policies can be expected to be correlated positively with tolerance for public debt. This view would be backed by political economy ­approaches that point to rent-seeking, self-interested behavior of politicians eventually undermining trust and questioning a government’s capacity of managing taxpayer money diligently (for respective overviews, see Alesina et al., 1997; or Persson and Tabellini, 2000). As empirical proxies, we include variables capturing individuals’ satisfaction with public services, trust in government, the belief that the state is wasting taxpayer money, and preference for redistribution.

For choosing the concrete combination of regressors in the individual specifications, we face the following tradeoff: Basically, we attempt to include as many explanatory variables as possible in order to exploit the rich survey information at hand and to identify interesting profiles of individuals expressing skeptical versus trustful views about public debt. At the same time, by including a large set of ­regressors, we face econometric challenges such as multicollinearity and overfitting. To address these issues at least to some extent, we opted for a specific-to-general approach, starting from the following parsimonious specification of the baseline model:

(1)

Our baseline model incorporates (1) the basic sociodemographic and socioeconomic characteristics Xsoc, which we expand by adding block-wise (2) individuals’ debt-relevant general attitudes XGen and (3) economic hardship experiences XPer. We then estimate two extended models. The first extension adds the variables on financial literacy or debt knowledge XFinlit. The second extension enhances the baseline model by adding, one by one, the variables capturing political attitudes XPol.

(2.1)

(2.2)

Potential multicollinearity issues were addressed by theoretical considerations as well as consulting bivariate correlations (see online supplement). Model selection is based on statistics of model fit and explanatory power. The selection of the presented models is supported unanimously by three different measures: the Schwarz criterion, or Bayesian information criterion (BIC); McFadden’s adjusted pseudo R2; and Tjur’s D (coefficient of discrimination).

3.2 Basic results

The estimated average marginal effects for the baseline specification are shown in table 1. Among the variables capturing sociodemographic and socioeconomic characteristics, wealth indicators indeed turn out to be important. We include two related indicators: one that captures real wealth in the form of the condition of respondents’ homes and another that captures financial wealth in the form of savings. Compared to individuals who live in a dwelling that was assessed by the interviewer to be of medium condition, individuals living in dwellings both of poor and excellent condition ⁴⁰ are more likely to associate higher debt levels with higher investments today as well as with negative effects for the future. The positive connotation of public debt for poorer individuals regarding investment benefits may stem from the belief that public investment can improve public infrastructure and thus their living standards, too. Turning to financial wealth, respondents who ­possess savings are more concerned about the future impact of public debt than those without savings. This finding, together with the previous one on individuals living in dwellings of excellent condition, would be in accordance with the existing literature for advanced economies and corroborate Ricardian arguments. ⁴¹ Finally, to control to some extent for per capita wealth, we also include household size. Larger households apparently are less likely to believe that higher public debt allows for higher investments today, probably because they are confronted with cramped living conditions and would therefore support more (residential) investment. ⁴²

Besides the explanatory variables discussed so far, we add in a second step individuals’ debt-relevant general attitudes (see medium panel in table 1). Our results indicate that people who have a present bias (i.e. are either impatient or impulsive or both; constituting 45% of the sample) express a positive connotation of public debt across all the specifications. This is as expected, given that these respondents are less likely to identify themselves with future repayment obligations. In contrast, respondents who are more debt averse themselves, i.e. those that feel uncomfortable if they owe money to others, have very strong concerns about the possible future negative effects of higher public debt. Finally, general trust in institutions (an index excluding government) matters as well: Those who are in general more trustful also see public debt in a more positive light. ⁴³

In a third step, we include additional explanatory variables that capture economic hardship respondents may have experienced (see lower panel in table 1). ⁴⁴ First, respondents that were unemployed at the time of the survey tend to agree that public debt compromises the opportunities of future generations and implies lower social benefits in the future. This can probably be explained by the fact that the unemployed are more vulnerable and, due to this, do not believe that the system offers sufficient solidarity. Second, people who had to cut back on basic consumption during 2008–2018 hold skeptical views about public debt across all the four specifications. Having to cut down basic consumption is well remembered even after several years and thus an indicator for severe economic hardship (see EBRD, 2006; Denisova et al., 2010). These people’s negative views of public debt might be due to a loss of trust in the crisis mitigation capacities of public institutions (see Eller and Scheiber, 2020) and/or the fear that higher public debt levels could make them go through a similar hardship experience again. Unemployment and past ­basic consumption cuts, alongside the explanatory variables mentioned above, constitute the “baseline controls” referred to in the remainder of this section. ⁴⁵ Third, we add other variables step by step that might be related to the attitudes of people who suffer(ed) economic hardship (only specifications where these variables are statistically significant are reported). Similarly to the results for housing conditions, individuals who belong to both low- or high-income households are more likely to agree that higher public debt compromises the opportunities of future generations. ⁴⁶ At the same time, respondents are more likely to support the view that higher debt allows for higher investments today if they live outside the capital city (probably due to the thinner infrastructure there) but also in a neighborhood characterized by comparatively stronger economic activity (proxied by nightlight intensity), or assess their personal health status as favorable (to control also for a less economically oriented hardship variable). Finally, access to liquidity in an emergency situation dampens people’s skeptical debt views: these individuals are less likely to think that they have to pay higher taxes and will receive lower social benefits in the future as a result of higher public debt today.

To provide a first summary picture, chart 4 depicts the relative importance of the most important baseline regressors by comparing for each dependent variable the estimated average marginal effects of these regressors. Interestingly, individuals’ public debt attitudes are strongly associated with other personal attitudes; especially aversion to incurring personal debt appears crucial with the by far largest average marginal effects in the three specifications indicating “debt skepticism” (i.e., to reiterate, the understanding that public debt compromises the opportunities of future generations, implies higher taxes, and implies lower pensions and benefits). Finally, belonging to more disadvantaged groups of society (poor dwelling conditions, unemployed, experience of basic consumption cutbacks) also yields very ­significant effects in most specifications.

Table 1: Baseline estimation: main factors associated with attitudes toward public debt  

	Binary dependent variables
Outcome variable: agreement with statement (0/1)	Higher public debt compromises ­opportunities of future ­generations		Higher public debt allows for higher investments today			Higher public debt ­implies higher future taxes		Higher public debt ­implies lower future pensions and benefits
	1	2	3	4	5	6	7	8	9
	Average marginal effects
Respondents’ sociodemographic and ­socioeconomic characteristics
Aged 19 to 34 years	–0.004	–0.007	–0.007	–0.005	–0.012	0.003	–0.000	–0.004	–0.005
	(0.013)	(0.012)	(0.013)	(0.013)	(0.013)	(0.013)	(0.012)	(0.013)	(0.013)
Aged 55+ years	–0.004	–0.002	–0.002	0.001	0.011	–0.011	–0.012	–0.021	–0.024*
	(0.013)	(0.013)	(0.014)	(0.014)	(0.014)	(0.013)	(0.013)	(0.014)	(0.013)
Female	–0.017*	–0.008	–0.004	–0.001	0.001	0.003	0.004	–0.007	–0.002
	(0.010)	(0.010)	(0.010)	(0.010)	(0.010)	(0.010)	(0.009)	(0.010)	(0.009)
Dwelling is well maintained	0.030**	0.024*	0.035***	0.032**	0.029**	0.028**		0.023*
	(0.012)	(0.012)	(0.013)	(0.013)	(0.013)	(0.013)		(0.013)
Dwelling is poor, needs major repair	0.041**	0.053***	0.050**	0.048**	0.054***	0.006		0.043**
	(0.018)	(0.017)	(0.020)	(0.019)	(0.019)	(0.018)		(0.019)
Respondent has accumulated savings	0.033***	0.026**	0.024*	0.025*	0.024*	0.028**		0,020
	(0.012)	(0.012)	(0.014)	(0.014)	(0.014)	(0.013)		(0.013)
Refused to reveal extent of savings	–0.025	–0.016	–0.026	–0.024	–0.016	0.008		–0.012
	(0.032)	(0.031)	(0.039)	(0.037)	(0.038)	(0.032)		(0.034)
2-person household	–0.020	–0.006	–0.042**	–0.039**	–0.042**	–0.036**	–0.032*	–0.012	–0.017
	(0.016)	(0.017)	(0.018)	(0.018)	(0.018)	(0.017)	(0.017)	(0.017)	(0.017)
3-plus-person household	–0.033*	–0.016	–0.051***	–0.043**	–0.050***	–0.043**	–0.034**	–0.022	–0.019
	(0.017)	(0.018)	(0.018)	(0.018)	(0.018)	(0.017)	(0.017)	(0.017)	(0.017)
Respondents’ general attitudes
Present bias (index)	–0.025***	–0.026***	0.032***	0.031***	0.031***	–0.024***	–0.023***	–0.014***	–0.013**
	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)
Discomfort of owing money	0.105***	0.106***	0.041**	0.033*	0.034*	0.144***	0.153***	0.131***	0.137***
	(0.016)	(0.016)	(0.018)	(0.018)	(0.018)	(0.017)	(0.016)	(0.017)	(0.016)
General trust in institutions (index)	–0.011	–0.012*	0.024***	0.028***	0.024***	–0.019***	–0.016**	–0.023***	–0.021***
	(0.007)	(0.007)	(0.008)	(0.008)	(0.008)	(0.007)	(0.007)	(0.007)	(0.007)
Respondents’ economic hardship experiences
Currently unemployed	0.047***		0.022			0.023		0.041**
	(0.015)		(0.016)			(0.016)		(0.016)
Had to cut back consumption (2008–2018)	0.057***		–0.039***			0.066***		0.049***
	(0.013)		(0.014)			(0.013)		(0.013)
Household net income: 1st quartile		0.037**
		(0.015)
Household net income: 4th quartile		0.033**
		(0.015)
Capital city resident				–0.066**
				(0.033)
Local economic activity (nightlight brightness)				0.016*
				(0.009)
Self-reported health status (index)					0.024***
					(0.008)
Access to emergency borrowing (index)							–0.111***		–0.091**
							(0.040)		(0.041)
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Log likelihood	–5,200.3	–5,458.0	–5,732.2	–6,000.4	–5,976.6	–5,278.8	–5,537.5	–5,446.8	–5,705.2
Pseudo R-squared (McFadden)	0.07	0.06	0.03	0.03	0.03	0.10	0.10	0.07	0.07
Pseudo R-squared (McKelvey & Zavoina)	0.13	0.12	0.06	0.06	0.06	0.19	0.18	0.14	0.14
Probability > Chi squared (df_m)	345.42 (23)	314.09 (24)	182.54 (23)	182.56 (23)	187.20 (22)	507.24 (23)	471.12 (18)	406.43 (23)	362.63 (18)
Number of observations	8,907	9,315	8,687	9,086	9,046	8,765	9,155	8,652	9,038
BIC	10,618.9	11,144.6	11,682.00	12,219.5	12,162.8	10,775.6	11,248.3	11,111.3	11,583.5
Adjusted pseudo R-squared (McFadden)	0.06	0.06	0.02	0.02	0.02	0.09	0.09	0.07	0.07
Tjur’s D	0.09	0.08	0.04	0.04	0.04	0.12	0.12	0.09	0.09
P(DepVar=1)	0.68	0.68	0.41	0.42	0.42	0.61	0.61	0.59	0.59
Source: Authors’ calculations based on OeNB Euro Survey 2018.
Note: Average marginal effects estimated with probit regressions with country fixed effects, using data from the OeNB Euro Survey 2018; robust standard errors are adjusted for clustering at the PSU level and reported in parentheses. ***, **, * denote that the average marginal effect is statistically different from zero at the 1%, 5% and 10% level, respectively. For a definition of the variables, see annex table A1. P(DepVar=1) denotes the unconditional sample probability of the respective dependent variable. By construction, positive (negative) average marginal effects imply that respondents are more (less) likely to agree with the statements underlying the dependent variables. Base categories are: 35 to 54 years old; main residence is good, only needs minor repair; reports to have no savings; 1-person household; Czech resident. For specification 2, the base category consists of the 2nd and 3rd household net income quartile. The additional dummy for refusing to report household net income is not shown. The sample comprises all ten OeNB Euro Survey countries.

In a fourth step, we expand the baseline specification by variables capturing financial literacy levels or debt knowledge (in line with equation 2.1). Table 2 shows that the financial literacy dummy ⁴⁷ is related positively to the first three ­dependent variables, suggesting that people who understand the basic concepts related to ­finance tend to be more skeptical about the future impact of higher public debt, while agreeing that higher debt levels allow for higher investments today (conditional on that these respondents are not concerned about debt dynamics). Respondents who are concerned about public debt dynamics during the past ten years have similar views (with a substantially larger average marginal effect) as financially literate persons. They weigh potential future drawbacks of public debt more heavily but also agree – despite their backward-oriented concerns – that higher public debt levels allow for higher investments today. Including an interaction term between the two variables reveals that individuals who are both financially literate and concerned about past debt dynamics are less likely to see public debt positively as a catalyst for investment, compared with respondents who are not financially literate and unconcerned about the debt history, turning the positive view into a skeptical one. This brings us to studying the role of debt knowledge more closely. Financial literacy and debt knowledge correlate only slightly, and their correlation with debt attitudes differs as well. Knowing (or guessing) the size of public debt is related negatively to the dependent variables, albeit not in a highly statistically significant manner. The most convincing case can be made for debt knowledge being associated with a less pronounced belief that there is a nexus between today’s public debt level and the future tax burden. However, knowing the current debt level and ­being at the same time concerned about past debt developments (see the results for the respective interaction terms) amplifies both skeptical views about the future impact of public debt and the belief that higher debt allows for higher investments today. Finally, we include in table 2 also self-assessed interest in politics to capture individuals that are likely better-informed about public finance issues: they also share the positive view that higher public debt levels allow for higher investments today. ⁴⁸

Table 2: Extension 1: association of attitudes toward public debt with financial knowledge  

	Binary dependent variables
Outcome variable: agreement with statement (0/1)	Higher public debt compromises ­opportunities of future generations		Higher public debt ­allows for higher ­investments today		Higher public debt ­implies higher future taxes		Higher public debt ­implies lower future pensions and benefits
	1	2	3	4	5	6	7	8
	Average marginal effects
Respondents’ financial literacy level and debt knowledge
Interested in politics	0.005	0.007	0.051***	0.047***	0.009	0.013	–0.001	0.006
	(0.012)	(0.012)	(0.013)	(0.013)	(0.012)	(0.012)	(0.013)	(0.013)
Financially literate	0.090***		0.099***		0.086***		0.042
	(0.031)		(0.035)		(0.032)		(0.033)
Concerned about development of public debt	0.189***	0.146***	0.177***	0.084***	0.237***	0.176***	0.230***	0.201***
	(0.020)	(0.016)	(0.024)	(0.019)	(0.020)	(0.017)	(0.021)	(0.017)
Interaction: financially literate AND concerned about development of public debt	0.007		–0.123***		–0.038		0.008
	(0.032)		(0.037)		(0.034)		(0.035)
Correctly identified size of public debt		–0.031		–0.055*		–0.072**		–0.061*
		(0.029)		(0.032)		(0.032)		(0.033)
Failed to identify size of public debt		0.037***		–0.030*		0.026*		0.011
		(0.014)		(0.016)		(0.015)		(0.015)
Interaction: correctly identified size of public debt AND concerned about its development		0.083***		0.068**		0.084**		0.052
		(0.031)		(0.034)		(0.033)		(0.034)
Other baseline controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Log likelihood	–4,716.0	–5,014.3	–5,351.7	–5,612.3	–4,796.9	–5,071.7	–4,956.7	–5,226.5
Pseudo R-squared (McFadden)	0.10	0.09	0.04	0.04	0.13	0.12	0.10	0.10
Pseudo R-squared (McKelvey & Zavoina)	0.18	0.17	0.03	0.03	0.12	0.12	0.10	0.09
Probability > Chi squared (df_m)	531.27(27)	494.14(28)	249.82(27)	241.93(28)	689.10(27)	674.40(28)	601.19(27)	608.72(28)
Number of observations	8,381	8,788	8,206	8,579	8,276	8,661	8,166	8,543
BIC	9,685.0	10,292.0	10,955.8	11,487.3	9,846.4	10,406.3	10,165.6	10,715.5
Adjusted pseudo R-squared (McFadden)	0.09	0.08	0.08	0.07	0.24	0.23	0.19	0.19
Tjur’s D	0.13	0.11	0.05	0.05	0.16	0.15	0.13	0.13
P(DepVar=1)	0.68	0.68	0.41	0.41	0.62	0.62	0.60	0.59
Source: Authors’ calculations based on OeNB Euro Survey 2018.
Note: Average marginal effects estimated with probit regressions with country fixed effects, using data from the OeNB Euro Survey 2018; robust standard errors are adjusted for clustering at the PSU level and reported in parentheses. ***, **, * denote that the average marginal effect is statistically different from zero at the 1%, 5% and 10% level, respectively. For a definition of the variables, see annex table A1. P(DepVar=1) denotes the unconditional sample probability of the respective dependent variable. By construction, positive (negative) average marginal effects imply that respondents are more (less) likely to agree with the statements underlying the dependent variables. Base categories are: 35 to 54 years old; main residence is good, only needs minor repair; reports to have no savings; 1-person household; Czech resident. The sample comprises all ten OeNB Euro Survey countries.

In a fifth and final step, we expand the baseline specification by several attitudes about politics (in line with equation 2.2). However, the issue of endogenous regressors applies particularly to these political preference variables, and the interpretation in the sense of conditional correlations should therefore be reiterated. We cover four different attitudes about politics and include each of them in a separate regression. Accordingly, table 3 shows how each dependent variable is associated with political preferences. The following observations can be made. First, people who are more satisfied with the delivery of public services are more confident about higher public debt – very consistently across the four specifications. The same holds true for ­individuals who think that the state manages tax revenues conscientiously (except for the specification depicting public debt as a catalyst for ­investment, where the relation is insignificant). Second, people who believe that the state should reduce income differences between the rich and the poor are less likely to agree that higher public debt compromises the opportunities of future generations or implies higher future taxes. Thus, people who support income redistribution by the state tend to be less skeptical about the future impact of public debt, possibly because they have in mind that public spending enabled by higher debt would benefit the poor relatively more. Finally, people who have greater trust in government tend to be more confident about higher public debt levels, too. ⁴⁹

Table 3: Extension 2: association of attitudes toward public debt with political attitudes  

	Specifications for attitudes toward politics
	Satisfaction with public services ­(index)	State is wasting taxpayer money	Preference for income ­redistribution	Trust in central government
	1	2	3	4
	Average marginal effects
Dependent variable: higher public debt compromises opportunities of future ­generations
General trust in institutions (index)	0.001	–0.001	–0.013*	–0.000
	(0.007)	(0.007)	(0.007)	(0.007)
Respective political attitudes	–0.045***	0.114***	–0.029**	–0.045***
	(0.010)	(0.013)	(0.013)	(0.015)
Other baseline controls	Yes	Yes	Yes	Yes
Country fixed effects	Yes	Yes	Yes	Yes
Log likelihood	–5,082.5	–5,046.7	–5,128.2	–5,149.8
Pseudo R-squared (McFadden)	0.07	0.08	0.07	0.07
Pseudo R-squared (McKelvey & Zavoina)	0.14	0.15	0.13	0.13
Probability > Chi squared (df_m)	353.87(24)	425.62(24)	352.94(24)	345.93(24)
Number of observations	8,710	8,716	8,795	8,822
BIC	10,391.7	10,320.1	10,483.4	10,526.8
Adjusted pseudo R-squared (McFadden)	0.07	0.08	0.07	0.06
Tjur’s D	0.09	0.10	0.09	0.09
P(DepVar=1)	0.68	0.68	0.68	0.68
Dependent variable: higher public debt ­allows for higher ­investments today
General trust in institutions (index)	0.013	0.024***	0.024***	0.018**
	(0.008)	(0.008)	(0.008)	(0.008)
Respective political attitudes	0.048***	0.011	0.022	0.030*
	(0.011)	(0.014)	(0.015)	(0.016)
Other baseline controls	Yes	Yes	Yes	Yes
Country fixed effects	Yes	Yes	Yes	Yes
Log likelihood	–5,590.4	–5,604.9	–5,659.2	–5,677.7
Pseudo R-squared (McFadden)	0.03	0.03	0.03	0.03
Pseudo R-squared (McKelvey & Zavoina)	0.06	0.06	0.06	0.06
Probability > Chi squared (df_m)	196.41(24)	186.10(24)	184.74(24)	183.58(24)
Number of observations	8,491	8,501	8,580	8,608
BIC	11,407.0	11,435.9	11,544.8	11,581.9
Adjusted pseudo R-squared (McFadden)	0.03	0.02	0.02	0.02
Tjur’s D	0.04	0.04	0.04	0.04
P(DepVar=1)	0.42	0.41	0.41	0.41
Source: Authors’ calculations based on OeNB Euro Survey 2018.
Note: Average marginal effects estimated with probit regressions with country fixed effects, using data from the OeNB Euro Survey 2018; robust standard errors are ­adjusted for clustering at the PSU level and reported in parentheses. ***, **, * denote that the average marginal effect is statistically different from zero at the 1%, 5% and 10% level, respectively. For a definition of the variables, see annex table A1. P(DepVar=1) denotes the unconditional sample probability of the respective ­ dependent variable. By construction, positive (negative) average marginal effects imply that respondents are more (less) likely to agree with the statements underlying the dependent variables. The sample comprises all ten OeNB Euro Survey countries.

Table 3 continued: Extension 2: association of attitudes toward public debt with political attitudes  

	Specifications for attitudes toward politics
	Satisfaction with public services ­(index)	State is wasting taxpayer money	Preference for income ­redistribution	Trust in central government
	1	2	3	4
	Average marginal effects
Dependent variable: higher public debt implies higher future taxes
General trust in institutions (index)	–0.010	–0.010	–0.021***	–0.001
	(0.007)	(0.007)	(0.007)	(0.008)
Respective political attitudes	–0.043***	0.121***	–0.028**	–0.081***
	(0.011)	(0.013)	(0.013)	(0.015)
Other baseline controls	Yes	Yes	Yes	Yes
Country fixed effects	Yes	Yes	Yes	Yes
Log likelihood	–5,164.5	–5,104.1	–5,206.2	–5,206.6
Pseudo R-squared (McFadden)	0.10	0.11	0.10	0.10
Pseudo R-squared (McKelvey & Zavoina)	0.19	0.21	0.19	0.19
Probability > Chi squared (df_m)	503.17(24)	609.56(24)	513.88(24)	532.70(24)
Number of observations	8,578	8,573	8,655	8,681
BIC	10,555.5	10,434.5	10,639.1	10,640.0
Adjusted pseudo R-squared (McFadden)	0.09	0.10	0.09	0.10
Tjur’s D	0.13	0.14	0.13	0.13
P(DepVar=1)	0.61	0.61	0.61	0.62
Dependent variable: higher public ­debt ­implies lower future pensions and benefits
General trust in institutions (index)	–0.013*	–0.014*	–0.024***	–0.003
	(0.007)	(0.007)	(0.007)	(0.008)
Respective political attitudes	–0.042***	0.107***	–0.013	–0.083***
	(0.011)	(0.014)	(0.014)	(0.015)
Other baseline controls	Yes	Yes	Yes	Yes
Country fixed effects	Yes	Yes	Yes	Yes
Log likelihood	–5,333.8	–5,289.2	–5,380.9	–5,369.4
Pseudo R-squared (McFadden)	0.07	0.08	0.07	0.07
Pseudo R-squared (McKelvey & Zavoina)	0.14	0.16	0.14	0.15
Probability > Chi squared (df_m)	399.03(24)	478.73(24)	405.63(24)	434.69(24)
Number of observations	8,464	8,464	8,545	8,569
BIC	10,893.7	10,804.5	10,988.1	10,965.2
Adjusted pseudo R-squared (McFadden)	0.07	0.07	0.07	0.07
Tjur’s D	0.09	0.10	0.09	0.10
P(DepVar=1)	0.59	0.59	0.59	0.59
Source: Authors’ calculations based on OeNB Euro Survey 2018.
Note: Average marginal effects estimated with probit regressions with country fixed effects, using data from the OeNB Euro Survey 2018; robust standard errors are ­adjusted for clustering at the PSU level and reported in parentheses. ***, **, * denote that the average marginal effect is statistically different from zero at the 1%, 5% and 10% level, respectively. For a definition of the variables, see annex table A1. P(DepVar=1) denotes the unconditional sample probability of the respective ­ dependent variable. By construction, positive (negative) average marginal effects imply that respondents are more (less) likely to agree with the statements underlying the dependent variables. The sample comprises all ten OeNB Euro Survey countries.

3.3 Robustness analysis

First, the prevalence of highly-significant country fixed effects in all specifications indicates that cross-country differences in the various institutional and economic backgrounds are important for determining individual public debt perceptions. Therefore, we execute a type of jackknife test, re-estimating the baseline specifications and the extensions by excluding one country at a time. The main effects of the regressors on general attitudes, economic hardship, financial literacy and political attitudes are robust with respect to the exclusion of individual countries. Yet, we identify some robustness issues regarding the marginal effects of the regressors capturing real wealth, savings and knowledge about the size of public debt. For the baseline model and extension 2, the reported highly significant average marginal effects of wealth and savings turn out to be robust, while the less significant average marginal effects of the wealth and savings regressors vary with the inclusion of some countries, and they lose significance if Albania is excluded. For extension 1, the category dwelling well-maintained turns out robust, while the poor-wealth and savings dummies vary with the inclusion of some countries. Moreover, the average marginal effects on savings, general trust in institutions or unemployment occasionally even increase and become more significant if particular countries are excluded. Finally, the household-size regressors turn out to be sensitive with ­respect to the inclusion of Czechia, Poland and sometimes of Albania.

Second, the analysis of the correlates of straight-lining (see online supplement) concludes that nondifferentiation is mainly influenced by individual characteristics and rushing through the interview, while the rather few cases of multiple straight-lining pile up mainly in Albania and appear to be influenced by interviewer effects on response styles, too. To assess the impact of straight-lining on the robustness of the presented results, we re-estimate the baseline specification and the extensions by dropping (a) all respondents who straight-lined at least once,
(b) all respondents who straight-lined their responses on the dependent variables and (c) by dropping only the major cases of multiple straight-lining (see the results for the latter in the online supplement). Excluding all cases with straight-lining (or only with respect to the dependent variables) shrinks the sample size by 58% (or 24%) and would presumably exclude many (some) respondents who simply have pronounced views on the specific topic of public debt. Regression results show for all specifications that the exclusion of all straight-lining respondents leads to insignificant average marginal effects of the regressors for wealth, savings and household size; but confirm the robustness of the effects of all other explanatory variables. ⁵⁰ This implies that the majority of straight-lining cases are not measurement errors but rather assemble respondents with pronounced views associated with relatively richer and poorer respondents. Excluding the major cases of multiple straight-lining (i.e. 4% of the sample size, mostly concentrated on Albania and Hungary) yields smaller and less significant average marginal effects only for the poor-wealth dummy, confirming the robustness of all other regressors. ⁵¹

Third, given that the binary dependent variables used thus far are derived from naturally ordered raw data, we alternatively apply an ordered probit and a generalized ordered probit estimator. The results confirm the main findings concerning excellent housing conditions, general attitudes and economic hardship (see online supplement for the baseline and further econometric details). But the regressors for poor housing condition and savings turn out insignificant for all the four dependent variables. The average marginal effects as well as the threshold parameters change sign between the categories “strongly agree” and “agree” for the statement that higher debt compromises the opportunities of future generations, and between “somewhat agree” and “agree” in the other cases – substantiating the coding rule for the binary dependent variables. As another robustness check with respect to defining the dependent variable, we rerun the probit regressions with the broader definition of agreement (i.e. including also “somewhat agree”). The main results survive, but the detected effects are partly smaller and statistically less significant.

Fourth, as a robustness check with respect to model selection, we apply the least absolute shrinkage and selection operator (LASSO) method, i.e. a machine-­driven selection of explanatory variables from a larger set of candidate variables. The LASSO results largely confirm the preferred probit baseline specification, whereby a few variables that lacked robustness or were not highly significant in the probit estimations were selected by LASSO (results are shown in the online supplement).

Finally, concerning the estimation technique, the significance of results turns out robust if standard errors are clustered at the regional level or if we use Huber-­White robust standard errors.

4 Conclusion

This paper has sought to contribute accordingly, by analyzing people’s attitudes toward public debt incurrence in ten CESEE countries, using unique survey data collected in the fall of 2018 through the OeNB Euro Survey. A stylized fact from our analysis is that CESEE residents have, in general, negative views about public debt, with a big majority believing that higher public debt levels compromise the opportunities of future generations and that there is a link between higher debt incurrence and the perspective of higher taxes and lower social benefits. The belief that higher public debt levels allow for higher investments today is also widespread but less common. There are cross-country differences, with individuals in some CESEE countries being more skeptical than in others, but the overall conclusion refers to all countries.

Turning to the profile of individuals that expressed certain attitudes toward public debt, our econometric analysis stresses the importance of several groups of factors. First, a more skeptical view of public debt is strongly affected by personal debt-related attitudes, in particular aversion to incurring personal debt. Second, economic well-being matters. Individuals who belong to more disadvantaged societal groups (particularly individuals who have experienced economic hardship) tend to be most concerned about public debt. Conversely, this does not uniformly imply that better-off individuals are more debt tolerant. For instance, we provided some evidence that wealthier individuals also express comparatively more skepticism about higher public debt levels, pointing to the existence of Ricardian considerations in our sample. Third, people who are more financially literate tend to be more skeptical about the future impact of higher public debt, but also agree that higher public debt levels allow for higher investments today. Fourth, we have also learned that individuals’ general attitudes about politics underpin their public debt attitudes: A more confident view of public debt is associated with satisfaction with public services, preference for income distribution by the state, the perception that the state is managing taxpayer money diligently, and trust in government.

The finding that worse-off societal groups are more skeptical toward higher levels of public debt contrasts with several existing studies for advanced economies that found that it is primarily the better off who tend to be more skeptical and prefer deleveraging. The skepticism of disadvantaged (or disappointed) societal groups in CESEE may be explained by the comparatively lower level of social spending and the predominance of regressive tax systems in the region, which could make worse-off individuals believe that the burden of higher debt must eventually be shouldered by them. Thus, to gain people’s support for temporarily higher public debt levels – as e.g. needed during the current crisis situation – our results would call for channeling increased government spending to the more disadvantaged groups of society (Eller and Scheiber, 2020) and strengthening the progressivity of tax systems (as recently emphasized by Gaspar et al., 2021).

Our descriptive analysis pointed to notable cross-country differences in individuals’ perception of public debt. As parts of future research efforts, the reasons for these differences across countries could be explored – e.g. by adding country-­specific regressors to the model (instead of country fixed effects) and moving ­toward multilevel modeling to account for the hierarchical structure of the data. In line with our reasoning above, it would be interesting to examine whether a different perception of public debt across countries is indeed related to differences in public spending intensity and composition (e.g. the share of social expenditure) or the degree of progressivity of the taxation system. In addition, besides obvious macroeconomic factors, it would be interesting to investigate several institutional factors that could explain local variation in individuals’ debt attitudes and fiscal knowledge, such as the way fiscal topics (e.g. fiscal council recommendations) are covered by public media, a country’s past track record in complying with fiscal rules or meeting fiscal targets, or more generally the quality of public sector governance – to mention only a few.

Once the dust of the COVID-19 pandemic settles, it will be of interest to ­examine whether people’s attitudes toward public debt have changed significantly compared to the pre-crisis survey data presented in this paper. Therefore, in future waves of the OeNB Euro Survey, we will attempt to repeat parts of the survey presented here and realign the answers in order to uncover post-crisis debt consolidation preferences of individuals in CESEE (e.g. analogously to Stix, 2013; or Hayo and Neumeier, 2017) and to investigate how these preferences change once consolidation measures are in fact taken (e.g. building on Kalbhenn and Stracca, 2020).

References

Abbas, A., O. Basdevant, S. Eble, G. Everaert, J. Gottschalk, F. Hasanov, J. Park, C. Sancak, R. Velloso and M. Villafuerte. 2010. Strategies for Fiscal Consolidation in the Post-Crisis World. IMF Fiscal Affairs Department.

Alesina, A., N. Roubini and G. D. Cohen. 1997. Political Cycles and the Macroeconomy, MIT Press.

Ardanaz, M., M. Hallerberg and C. Scartascini. 2020. Fiscal consolidations and electoral outcomes in emerging economies: Does the policy mix matter? Macro and micro level evidence from Latin America. In: European Journal of Political Economy 64. 101918.

Arias, E. and D. Stasavage. 2019. How large are the political costs of fiscal austerity? In: The Journal of Politics 81(4). 1517–1522.

Bansak, K., M. M. Bechtel and Y. Margalit. 2021. Why austerity? The mass politics of a contested policy. In: American Political Science Review 115(2). 486–505.

Beckmann, E. and H. Stix. 2015. Foreign Currency Borrowing and Knowledge About ­Exchange Rate Risk. In: Journal of Economic Behavior & Organization 112. 1–16.

Beyer, R. C. M., E. Chabhra, V. Galdo and M. Rama. 2018. Measuring districts’ monthly economic activity from outer space. World Bank Policy Research Working Paper 8523.

Blesse, S., F. Heinemann, E. Janeba and J. Nover. 2021. Die Zukunft der Schuldenbremse nach der Pandemie: Einsichten aus einer Befragung aller 16 Landesparlamente. In: Zeitschrift für Wirtschaftspolitik 70(1). 1–16.

D’Acunto, F., D. Hoang and M. Weber. 2018. Unconventional fiscal policy. In: American Economic Review – Papers & Proceedings 108(5). 1–5.

Denisova, I., M. Eller and E. Zhuravskaya. 2010. What do Russians think about transition?
In: Economics of Transition 18(2). 249–280.

EBRD. 2006. Life in Transition I – A survey of people’s experiences and attitudes.

Eller, M. and T. Scheiber. 2020. A CESEE conundrum: low trust in government but high hopes for government-led job creation. In: Focus on European Economic Integration Q3/20. OeNB. 81–97.

Ferragina, E. and A. Zola. 2020. An Obituary for Austerity Narratives? An Experimental Analysis of Public Opinion Shifts and Class Dynamics during the Covid-19 Crisis. MaxPo Discussion Paper 20/5.

Gaspar, V., M. Keen, A. Klemm and P. Mauro. 2021. A COVID-19 Recovery Contribution. In: IMF Blog. April 16. https://blogs.imf.org/2021/04/16/a-covid-19-recovery-contribution/?utm_medium=email&utm_source=govdelivery

Hayo, B. and F. Neumeier. 2017. Public Attitudes toward Fiscal Consolidation: Evidence from a Representative German Population Survey. In: Kyklos 70. 42–69.

Hayo, B. and F. Neumeier. 2019. Public Preferences for Government Spending Priorities: ­Survey Evidence from Germany. In: German Economic Review 20(4). 20 (4), e1–e37.

Heinemann, F. and T. Henninghausen. 2012. Understanding Public Debt Preferences.
In: FinanzArchiv/Public Finance Analysis 68(4). 406–430.

Heinemann, F., E. Janeba and M. Todtenhaupt. 2020. Incumbency and expectations of fiscal rule compliance: Evidence from surveys of German policy makers. ZEW Discussion Papers 20-057.

Henderson, J. V., A. Storeygard and D. N. Weil. 2012. Measuring economic growth from outer space. In: The American Economic Review 102(2). 994–1028.

Horvath, R. 2013. Does trust promote growth? In: Journal of Comparative Economics 41(3). 777–788.

Hübscher, E., T. Sattler and M. Wagner. 2020. Voter responses to fiscal austerity. In: British Journal of Political Science. 1–10. doi:10.1017/S0007123420000320.

Johnson, D. S., J. A. Parker and N. S. Souleles. 2006. Household Expenditure and the ­Income Tax Rebates of 2001. In: American Economic Review 96(5). 1589–1610.

Lusardi, A. and P. Tufano. 2015. Debt literacy, financial experiences, and overindebtedness. In: Journal of Pension Economics & Finance 14(4). 332–368.

Kalbhenn, A. and L. Stracca. 2020. Mad about Austerity? The Effect of Fiscal Consolidation on Public Opinion. In: Journal of Money, Credit and Banking 52(2-3). 531–548.

Mueller, E. 1963. Public Attitudes Toward Fiscal Programs. In: The Quarterly Journal of Economics 77(2). 210–235.

Persson, T. and G. Tabellini. 2000. Political Economics. MIT Press.

Pitlik, H., G. Schwarz, B. Bechter and B. Brandl. 2011. Near Is My Shirt but Nearer Is My Skin: Ideology or Self-Interest as Determinants of Public Opinion on Fiscal Policy Issues. In: Kyklos 64(2). 271–290.

Roth, C., S. Settele and J. Wohlfart. 2020. Beliefs about Public Debt and the Demand for Government Spending. CESifo Working Paper Series 8087.

Sangnier, M. 2013. Does trust favor macroeconomic stability? In: Journal of Comparative Economics 41(3). 653–668.

Shapiro, M. D. and J. B. Slemrod. 2009. Did the 2008 tax rebates stimulate spending? NBER Working Paper 14753. February.

Stix, H. 2013. Does the Broad Public Want to Consolidate Public Debt? – The Role of Fairness and of Policy Credibility. In: Kyklos 66(1). 102–129.

Sturzenegger, F. and J. Zettelmeyer. 2006. Debt Defaults and Lessons from a Decade of Crises. MIT Press.

Zak, P. J. and S. Knack. 2001. Trust and growth. In: The Economic Journal 111(470). 295–321.

Annex

Table A1: Definitions of variables  

Label	Description
Respondents’ attitudes toward public debt (dependent variables in the regressions)
Higher public debt compromises ­ opportunities of future generations	Dummy equals 1 if respondents agreed/strongly agreed with the statement “higher public debt levels diminish the chances of future generations”; 0 if they somewhat agreed/somewhat disagreed/ disagreed/strongly disagreed (6-point Likert scale). Excluding respondents who answered “don’t know” or did not provide an answer.
Higher public debt allows for higher ­investments today	Dummy equals 1 if respondents agreed/strongly agreed with the statement “higher public debt levels make it possible to conduct necessary investments today (e.g. into public infrastructure like schools and streets)”; 0 if they somewhat agreed/somewhat disagreed/disagreed/strongly disagreed (6-point Likert scale). Excluding respondents who answered “don’t know” or did not provide an answer.
Higher public debt implies higher ­future taxes	Dummy equals 1 if respondents agreed/strongly agreed with the statement “higher public debt levels imply that I will have to pay more taxes in the future”; 0 if they somewhat agreed/somewhat disagreed/disagreed/strongly disagreed with the statement (6-point Likert scale). Excluding respondents who answered “don’t know” or did not provide an answer.
Higher public debt implies lower future pensions and benefits	Dummy equals 1 if respondents agreed/strongly agreed with the statement “higher public debt levels imply that I will receive lower state pensions and/or lower welfare benefits in the future”; 0 if they somewhat agreed/somewhat ­disagreed/disagreed/strongly disagreed (6-point Likert scale). Excluding respondents who answered “don’t know” or did not provide an answer.
Respondents’ sociodemographic and socioeconomic (household) characteristics
Age	Dummy variables for three age groups: 19 to 34 years, 45 to 54 years (base category) and 55 and more years.
Female	Dummy variable that takes the value 1 if respondents were female, 0 otherwise (base category).
Condition of respondents’ home	Dummy variables: interviewer assessment whether respondents’ home is “excellent and well-maintained”; “good, needs some minor repairs” (base category); or “poor, needs major work/very poor, some walls or ceilings need replacement.”
Savings	Dummy variables: respondents have savings (themselves or together with their partners), refused to answer or were not able to provide an answer. Base category: have no savings.
Household size	Dummy variables for the number of persons living permanently in a given household (two individuals; three and more individuals). Base category: 1-person household.
Respondents’ general attitudes
Present bias (index)	Index is constructed as the average score of agreement on a 6-point Likert scale using the following two statements: “I tend to live for today and let tomorrow take care of itself” and “I am impulsive and tend to buy things even when I cannot really afford them.” Excluding respondents who answered “don’t know” or did not provide an answer.
Discomfort of owing money	Dummy equals 1 if respondents rather/strongly agreed with the statement “Owing money to somebody, even if it is just a small amount, makes me feel uncomfortable”; 0 otherwise (6-point-Likert scale). Excluding respondents who ­answered “don’t know” or did not provide an answer.
General trust in institutions (index)	Index constructed as the average trust score across five institutions: police, domestically owned banks, foreign-owned banks, the EU, the national central bank. Respondents were asked about how much they trust these five institutions on a 5-point Likert scale. Respondents answering “don’t know” were re-assigned to category 3 “I neither trust nor ­distrust.” Excluding respondents who did not provide an answer.
Source: Authors’ compilation.

Table A1 continued: Definitions of variables  

Label	Description
Respondents’ economic hardship experiences
Currently unemployed	Dummy equals 1 if respondents are currently unemployed; 0 otherwise (base category). Excluding respondents who did not provide an answer.
Cut back consumption	Dummy equals 1 if respondents had to cut back on basic consumption in the period from 2008 to 2018; 0 otherwise. Excluding respondents who answered “don’t know” or did not provide an answer.
Household net income	Dummy variables for the level of total household income after taxes (1st quartile, 2nd and 3rd quartile, 4th quartile, don’t know/no answer). Base category: 2nd and 3rd quartile.
Capital city resident	Dummy equals 1 if respondents’ main residence happened to be in the capital city; 0 otherwise (base category).
Local economic activity	Natural logarithm of visible infrared imaging radiometer suite (VIIRS) nightlight data for a radius of 10km around the primary sampling unit (PSU) serving as a proxy for local economic activity. The nightlight data refer to 2018 (annual ­averages) and were cleaned by a contractor according to Beyer et al. (2018).
Access to emergency ­borrowing (index)	Index constructed as the average of reported probabilities. The index is based on respondents’ self-assessment of ­being able to borrow a given amount (representing four times the average monthly wage in their country) in an emergency from five different sources: banks; their employer; payday lenders, pawnshops or internet loan providers; family, relatives or friends; or other private lenders. The responses were coded as ordinal categories, and the following numerical probabilities were assigned to the set of verbal categories, as identified: very likely = 0.9, likely = 0.65, don’t know = 0.5, unlikely = 0.35 and very unlikely = 0.1.
Self-reported health status (index)	Based on the answer to the question “All in all, how would you describe your current state of health? Would you say it is very good/good/fair/poor.” Index ranges from 1 (poor) to 4 (very good). Excluding respondents who answered “don’t know” or did not provide an answer.
Respondents’ financial literacy level and debt knowledge
Very interested in politics	Dummy equals 1 if respondents rather/strongly agreed with the statement “I am very interested in politics”; 0 otherwise (6-point Likert scale). Excluding respondents who did not provide an answer.
Concerned about the development of public debt	Dummy equals 1 if respondents rather/strongly agreed with the statement “The development of public debt [in my country] over the past 10 years is worrisome”; 0 otherwise (6-point Likert scale). Excluding respondents who did not provide an answer.
Financial literacy	Dummy equals 1 if respondents scored 3 or 4 on the financial literacy index; 0 otherwise (base category). The financial literacy index is based on answers to four questions, relating to real interest rates, exchange rates, inflation, and risk ­diversification. The index varies between 0 (= item non-response) and 4 (= all questions regarding financial literacy were answered correctly).
Interaction: financially literate AND concerned about ­development of public debt	Interaction term: dummy equals 1 if the financial literacy dummy equals 1 AND respondents indicated to rather/strongly agree with the statement “The development of public debt [in my country] over the past 10 years is ­worrisome”; 0 otherwise. Excluding respondents who did not provide an answer.
Correctly identified size of public debt	Dummy variables: correctly identified size of public debt or answered “don’t know.” Base category: failed to identify size of public debt, including item non-response. The original variable captures respondents’ actual answers: they were given six brackets (ranging from “below 20% of GDP” up to “above 100% of GDP”) and asked to indicate the appropriate bracket for their country. The debt ratios for Albania, Czechia, Hungary and Poland broadly corresponded to bracket midpoints, so that the respective brackets constituted the only appropriate answer. The ratios for all other countries broadly corresponded to two out of the six brackets, so that both brackets were counted as appropriate. The respective tolerance margin turns out to be +/–3 percentage points.
Interaction: correctly identified size of public debt AND concerned about its development	Interaction term: dummy equals 1 if respondents correctly identifed the size of public debt AND rather/strongly agreed with the statement “The development of public debt [in my country] over the past 10 years is worrisome”; 0 otherwise. Excluding respondents who did not provide an answer.
Respondents’ political attitudes
Satisfaction with public ­services (index)	Index constructed as the average degree of satisfaction with the delivery of public services as reported for six areas: social security; public infrastructure; education; health; defense and public safety; and economic development. Answers range from “very satisfied (4)” to “very dissatisfied (1).” Excluding respondents who answered “don’t know” or did not provide an answer.
State is wasting taxpayer money	Dummy equals 1 if respondents rather/strongly agreed with the statement “the state is wasting taxpayer money”; 0 if they rather/strongly agreed with the statement “the state manages tax revenues conscientiously,” or if they indicated no preference, or considered both statements to be somewhat true, or answered “don’t know.” Excluding respondents who did not provide an answer.
Preference for income ­redistribution	Dummy equals 1 if respondents rather/strongly agreed with the statement “the state should reduce income differences between those with higher incomes and those with lower incomes, e.g. via taxation or public benefits”; 0 if they rather/strongly agreed with the statement “the state should not reduce income differences,” or if they indicated no preference, or considered both statements to be somewhat true, or answered “don’t know.” Excluding respondents who did not provide an answer.
Trust in central government	Dummy equals 1 if respondents reported to somewhat or completely trust the central government/cabinet of ­ministers; 0 if they indicated complete/some distrust or picked neither trust nor distrust (5-point Likert scale). ­Excluding respondents who answered “don’t know” or did not provide an answer.
Source: Authors’ compilation.

European banks in Russia: developments and perspectives from 2017 through the COVID-19 pandemic (2020/2021)

Stephan Barisitz, Philippe Deswel ⁵²

Russia’s recent economic growth has been supported by the country’s banking sector, including the European banks operating in the market. While Russia’s pre-pandemic GDP growth had suffered from a weak investment climate, oil price volatility and sanctions, the strong financial buffers built up in recent years were an asset for the country and its banking system during the pandemic-triggered recession. The European banks that qualify as significant institutions (Raiffeisenbank Russia, Rosbank/Société Générale and UniCredit Bank Russia), while pursuing different strategies, have remained committed to the Russian market. During the crisis, banking and economic activity were supported by temporary regulatory forbearance with respect to asset (loan) valuation and provisioning as well as the central bank’s key rate cuts and targeted government subsidies. European banks in Russia nevertheless keep facing exogenous risks, such as sustained compliance with sanctions regimes in a situation that remains volatile and sensitive to adverse geopolitical developments. Foreign currency fluctuations and the depreciation of the ruble require adequate risk management, and climate risk represents an emerging challenge. There is also strong competition driven by the digital transformation of banking. In March and April 2021, Russia’s central bank raised its key policy rate again amid rising inflationary pressures and signs of incipient economic recovery. Once regulatory lenience and lending subsidies expire, the banking sector would in general appear sufficiently capitalized to cover a potential increase of loan losses and provisioning needs. This goes especially for the European banks in Russia, which tend to have better-than-average asset quality and a sound capital base, although their market environment is expected to remain challenging. Besides, generous reserves remain at the disposal of the authorities should financial problems emerge, a scenario whose implications remain untested in the case of European banks due to their resilient ­performance.

JEL classification: G21, G28, P34

Keywords: banking sector, financial stability, COVID-19, crisis, credit risk, European banks, FDI, nonperforming loans, profitability, recovery, regulatory forbearance, restructuring, Russia, sanctions

1 Pre-pandemic macroeconomic conditions and banking sector developments (2017 to March 2020)

1.1 Macroeconomic developments: modest growth amid sanctions, accompanied by the build-up of sizable financial buffers

In recent years, Russian GDP growth was driven by domestic demand, mostly by private consumption and to a lesser degree by gross fixed capital formation. On the production side of GDP, growth continued to be fueled by resource extraction, manufacturing (notably automobile production), agriculture, and to a lesser degree by retail trade. Russia’s unemployment rate (ILO definition) steadily declined to a post-Soviet historical low of 4.5% in early 2020. Monetary policy (inflation targeting since late 2014) remained cautious, also against the backdrop of high inflation expectations. The key policy rate of the Central Bank of Russia (CBR) – the one-week repo rate – was reduced from 7¾% in late 2017 to 6% in March 2020 – with a temporary upward correction by half a percentage point in late 2018 and early 2019 against the backdrop of intermittent ruble depreciation pressures. CPI inflation fell from a peak of above 5% in early 2019 (due also to housing and communal tariff adjustments and a VAT increase) to 2.5% at end-March 2020 – substantially below the CBR’s 4% target. Greater confidence in the Russian currency contributed to further de-dollarization.

Improved tax administration and pension reform (adjustment of retirement age) contributed to low budget deficits (2017) or budget surpluses (2018 and 2019). The latter held down government debt exposure and contributed to replenishing the National Wealth Fund, which expanded from about 4% of GDP at end-2017 to 11% of GDP at end-March 2020 (table 1). The Finance Ministry’s forex purchase program carried out by the CBR in order to (further) strengthen the country’s ­international reserve position helped keeping the ruble’s exchange rate “competitive.” This gave Russian oil and gas corporations, which operate on a ruble basis and whose revenues are (mostly) earned in US dollars, some advantages in terms of profitability over some of their competitors. Tight macroeconomic policies and the weak ruble caused the current account balance to remain in surplus, notwithstanding oil price volatility. The resulting twin surpluses – fiscal and current ­account, from 2018 to the first quarter of 2020 – as well as Western sanctions contributed to reining in gross external indebtedness, which declined from 31% of GDP in 2017 to 27% of GDP at end-March 2020. Meanwhile, gross international reserves (including gold) grew from 28% of GDP to 33%.

Table 1: Russia: selected macroeconomic indicators  

	2017	2018	2019	End-March 2020	End-June 2020	2020	End-March 20213
	%
Real GDP growth (annual change)	1.8	2.8	2.0	1.4	–3.4	–3.0	–0.7
Inflation (CPI, end of period, year on year)	2.5	4.3	3.0	2.5	3.2	4.9	5.8
Unemployment rate (ILO definition, average)	5.2	4.8	4.6	4.7	5.4	5.8	5.8
Budget balance (general government, ratio to GDP)	–1.5	2.9	1.9	2.5	–1.7	–4.0	2.5
National Wealth Fund (end of period, ratio to GDP)1	4.1	3.9	6.9	11.3	10.7	11.7	11.8
General government gross debt (end of period, ratio to GDP)	12.6	12.0	12.3	13.8	14.0	16.0	16.2
Current account balance (ratio to GDP)	2.3	6.9	3.8	5.5	3.1	2.1	3.7
Net private capital flows (end of period, ratio to GDP)	–2.0	–3.8	–1.6	–4.3	–4.1	–3.1	–2.7
Gross external debt (end of period, ratio to GDP)	31.1	28.1	28.9	26.8	29.8	30.0	29.9
Gross international reserves (incl. gold, end of period, ­ratio to GDP)	27.5	28.5	32.6	33.0	35.1	38.2	37.6
CBR key rate2 (end of period)	7.75	7.75	6.25	6.0	4.5	4.25	4.5
	Period average in USD or RUB
Urals grade crude oil price (USD/barrel)	53.0	70.0	63.6	48.2	39.7	41.7	59.8
RUB per 1 USD	58.33	62.54	64.73	66.10	69.13	71.94	74.34
RUB per 1 EUR	65.87	73.87	72.51	72.90	76.20	82.04	89.70
Source: Rosstat, Central Bank of Russia, Ministry of Finance.
1 The predominant part of this fund is also included in Russia’s gross international reserves.
2 The central bank’s one–week repo rate.
3 Preliminary data.

1.2 Continued banking sector stabilization and strengthening after the 2014/2015 crisis

Benefiting from high post-crisis real interest rates, deposits expanded markedly initially (+10% in 2017 in real terms and exchange rate-adjusted). Subsequently though, deposit expansion lost some steam (+4% at end-March 2020 year on year) in view of slowly declining real interest rates. With some variations due to ­exchange rate volatility, the share of forex deposits in total deposits remained slightly below one-quarter in this period (table 2), and the respective share in retail deposits hovered around one-fifth. In a remarkable development, government agencies stocked up their funds held at commercial banks from 5% to 8% of total deposits. The modest economic revival and the recovery of credit growth contributed to a drop in the ratio of nonperforming loans, from 10% (narrow definition) respectively 18% (broader definition, including doubtful loans) ⁵⁶ at end-2017 to 9% respectively 16% at end-March 2020. Loan loss provisions have continued to be somewhat smaller than the volume of nonperforming loans when defined narrowly and continued to follow the latter’s gentle decline. Thus, the loan loss provisions remain at slightly more than half the level of nonperforming loans when broadly defined.

Other areas also displayed stability or improvements over the pre-pandemic years. The banking sector’s liquidity remained stable in 2017 and up to early 2020 (highly liquid assets to total assets at around 11%). The ratio of large exposures to total sector assets remained on a declining trend and shrank from about 25% to 20%. Not least due to sanctions and continuous deleveraging, banks’ net external assets further rose to a record of 7.5% of total assets at end-March 2020 (table 2). After some banking turbulences, which had put pressure on profitability in 2017 (Barisitz, 2018, p. 61–64), banks’ profits recovered again and grew dynamically in the following years, mainly driven by the largest banks, and buoyed by credit ­expansion and rising confidence. Thus, return on assets (ROA) as well as return on equity (ROE) approximately doubled from 2017 to 2019 (to 2.0%, respectively 19.5%). Meanwhile, banks’ aggregate capital adequacy ratio remained above 12%. ⁵⁷ At the same time, the modest share of (majority) foreign-owned credit institutions in banks’ total assets slightly declined further (from 12% to 10%) in the above period.

1.3 Regulatory and other central bank measures to bolster banking sector recovery

Although household debt in Russia remains relatively low compared to other countries when measured as a ratio of GDP (about 16% in mid-2019), vigorous retail lending, notably mortgage loans and (up to mid-2019) unsecured consumer loans, substantially lifted households’ debt burden in proportion to disposable ­income, which prompted the CBR to raise risk weights several times in 2018–2019. The CBR furthermore introduced surcharges to the respective risk coefficients from October 2019. This appears to have (temporarily) reined in the swift expansion of lending activity in the mortgage and consumer loan segments. Meanwhile, the CBR continued its cleanup of the banking sector (removing mostly smaller players featuring nonviable business models or fraudulent practices). Accordingly, the total number of operating credit institutions further declined from 561 at end-2017 to 434 at end-March 2020 (table 2). Moreover, in 2018, 149 banks with less than RUB 1 billion in terms of capital were relicensed with a more restrictive license under which they are prohibited from most foreign operations but subject to simplified regulatory requirements. Finally, the CBR phased in Basel III macroprudential buffers, gradually raising the capital surcharge for systemically important banks between 2017 and end-2019 (from 0.35% to 1%) and the capital conservation buffer (from 1.25% to 2.5%). ⁵⁹

Table 2: Russia: selected banking sector stability indicators  

	End-2017	End-2018	End-2019	End-March 2020	End-June 2020	End-Sept. 2020	End-2020
	%
Credit risk
Total loans (annual real growth, exchange rate-­adjusted)1	+1.8	+6.2	+6.3	+6.7	+4.8	+4.7	+3.8
Loans to households (share in total loans)	31.1	33.6	36.9	36.4	36.6	36.9	37.7
Nonperforming loans (share of total loans, narrow definition)2,7	10.0	10.1	9.3	9.4	9.5	9.3	9.0
Nonperforming loans (share of total loans, broader definition)3,7	17.5	16.7	16.0	16.1	16.7	16.3	16.3
Market and exchange rate risk
Foreign currency loans (share in total loans)	14.5	12.7	10.6	12.4	11.1	12.2	11.7
Foreign currency deposits (share in total deposits)	23.5	24.1	21.2	24.7	21.5	23.2	22.7
Liquidity risk
Total deposits (annual real growth, exchange rate-adjusted)4	+10.1	+6.4	+5.3	+3.9	+2.4	+0.9	+1.7
Loan-to-deposit ratio	95.2	93.2	95.2	95.8	96.0	95.7	95.5
Banks’ external assets (share in total assets)5	11.0	12.2	11.2	11.9	11.0	12.2	11.4
Banks’ external liabilities (share in total liabilities)6	5.7	5.6	4.3	4.4	3.9	4.6	4.2
Profitability
Return on assets	1.0	1.2	2.0	2.0	1.7	1.7	1.7
Return on equity	8.3	11.5	19.5	19.0	15.9	16.0	15.7
Shock-absorbing factors
Capital adequacy ratio (capital to risk-weighted assets)7	12.1	12.2	12.3	12.2	12.8	12.7	12.5
Tier 1 capital ratio (Basel III)7	8.5	8.9	9.3	10.1	10.5	10.4	9.7
Loan loss provisions (ratio to total loans)7	9.3	9.1	8.7	8.8	9.0	9.0	8.9
Memorandum items
Total banking sector assets (ratio to GDP)	92.8	90.6	88.4	94.6	96.1	103.5	105.5
Share of majority state-owned banks in total banking assets	61.5	68.0	68.6	.	.	.	.
Share of majority foreign-owned banks in total banking assets	12.3	9.6	9.5	10.2	10.0	9.8	.
Total number of operating credit institutions	561	484	442	434	427	417	406
Source: Central Bank of Russia, Raiffeisen Research, OeNB calculations.
1 Loans and other placements with nonfinancial organizations, government agencies and individuals (resident sectors, excluding interbank transactions). Annual credit growth adjusted for exchange rate effects by applying the same exchange rate to consecutive year-end balances. As the US dollar accounts for the lion’s share of the foreign exchange credit portfolio, we use the USD/RUB exchange for the entire volume to simplify the calculation.
2 Share of problem loans (category IV) and loss loans (category V) in total loans including interbank loans (according to CBR regulation no. 254).
3 Share of doubtful (category III), problem (category IV) and loss loans (category V) in total loans including interbank loans (according to CBR regulation no. 254).
4 Deposits and other funds of nonfinancial organizations, government agencies and individuals (resident sectors, excluding interbank transactions). Annual deposit growth adjusted for ­exchange rate effects by applying the same exchange rate to consecutive year-end balances. As the US dollar accounts for the lion’s share of all foreign exchange deposits, we use the USD/RUB exchange for the entire volume to simplify the calculation.
5 Funds placed with nonresidents, including loans and deposits, correspondent accounts with banks, securities acquired.
6 Funds raised from nonresidents, including loans from foreign banks, deposits of legal entities and individuals.
7 Some data starting in April 2020 may not be fully comparable with previous data due to the application of temporary regulatory forbearance measures aimed at cushioning the economic impact of the COVID-19 pandemic.

2 Pre-pandemic development of European banks in Russia (2017 to 2019)

2.1 Slightly increasing market presence, with differing underlying trajectories

2.2 Dynamic revenue growth, but profitability ratios facing some pressure

2.3 Improving credit quality, compliance with capital requirements ensured

As outlined above, capital requirements were tightened by the Central Bank of Russia in the years ahead of the crisis. Subject to new requirements phased in from January 2016 onward, capital buffers had to be applied additionally to minimum capital ratios for significant institutions, while minimum requirements were kept constant. ⁶⁵ The European banks operating in Russia complied with the capital requirements and kept their capital ratios ⁶⁶ broadly stable from 2017 to 2019.

3 Macroeconomic and banking sector developments during the COVID-19 crisis

3.1 Rather mild recession on the back of comparatively small service sector, limited restrictive measures and targeted fiscal stimulus

On average, the ruble lost 11% of its value against the US dollar and 13% against the euro in 2020 compared to 2019. This slide, combined with price spikes for some food items, pushed CPI inflation up to 4.9% in December 2020 and 5.5% in April 2021 (year on year), clearly exceeding the CBR’s target. Lending priority to combating the deep crisis, the CBR then cut the key rate further from 6% to 4¼% in the summer of 2020 (a post-Soviet historic low). In March and April 2021, the CBR raised the rate again in two steps to 5%, citing continuing elevated inflation expectations, signs of swifter-than-expected domestic economic recovery in early 2021, and lingering geopolitical uncertainties. After years of fiscal surpluses, the general government budget deficit reached 4% of GDP in 2020. The shortfall was financed with new domestic debt rather than by tapping the National Welfare Fund (NWF). The NWF funds amounted to 12% of GDP at end-March 2021 (see table 1), with liquid assets accounting for two-thirds thereof.

The much lower prices and quantities of oil and gas exports cut into Russia’s current account surplus, which was more than offset by rising private net capital outflows, reflecting stepped-up deleveraging by banks and corporations. While Russia’s gross foreign debt consequently declined to EUR 391 billion at end-March 2021, this corresponded to an increase of 3 percentage points of GDP to 30% (year on year) – given the sharp drop of the ruble; at the same time, the country’s substantially larger international reserves remained more or less stable at EUR 491 billion, reflecting an increase of 5 percentage points of GDP to 38%.

3.2 Banking sector enters regime of regulatory forbearance and remains a pillar of growth

Overall annual credit growth in 2020 eased yet remained at a pace of 4% year on year (in real terms and exchange rate-adjusted). This continued robust growth amid the recession was supported by further falling interest rates and by targeted government subsidies. Loans were also needed as tide-over credits to close (temporary) cash flow gaps during the pandemic, and credit demand was not dampened at all by the forbearance regime. While retail loans continued to grow dynamically on balance, having just lost some momentum, mortgage loan growth re-accelerated to a rate of 16% in 2020 and expanded even faster in the first quarter of 2021 (year on year). Mortgage credit is supported by a preferential state program providing interest rate subsides that cap loan rates at 6.5%. ⁶⁸ Despite the recession, property prices have been rising on the back of low interest rates and growing ­demand for residential property. Given the unrelenting acceleration of mortgage lending, the CBR decided in mid-May 2021 to increase surcharges to the respective risk coefficients from August 2021 (once the preferential state program expires). Loans to some strategically important companies also benefited from state-subsidized interest rates (featuring a rate cap of 5%), and other support schemes shored up lending to SMEs in distressed sectors of the economy. Other factors underpinning the continued growth of banking assets were expanding investments in securities (+28% in 2020, in real terms) notably government bonds, which contributed to financing the 2020 budget shortfall, and the depreciation of the ruble over the year.

In contrast to lending, bank deposit growth slowed down markedly to 2% ­(table 2), and in the case of household deposits to 1%, in 2020 (in real terms and exchange rate-adjusted). In this respect, the impact of forced saving through lockdowns and restrictions has apparently been more than offset by a combination of factors, including reduced incomes and increasing joblessness, dwindling interest rates amid rising inflation, a new savings tax to help fund the government’s response to the pandemic ⁶⁹ and the growing attractiveness of alternative investments in the Russian stock market. ⁷⁰ Forex deposits (mainly denominated in US dollars, and a smaller share in euro) have been losing their attractiveness because of nominal ­interest rates in those currencies approaching zero and because of heightened uncertainty. Meanwhile, government agencies’ share of total deposits slightly declined to 7% (fully ruble-denominated, at end- 2020).

Given the CBR’s temporary forbearance regime from Q2/2020, the transmission of credit losses to banks’ prudential ratios has been limited since then; therefore, while nonperforming loans as well as provisions were largely reported stable through end-2020, these indicators may become less and less reliable the more time passes by. Likewise, as depicted in table 2, reported capital adequacy and the tier 1 capital ratio remained more or less unchanged, and reported ROE and ROA almost upheld their high levels of late 2019. As of late 2020, restructured loans came to about 12% of total loans (IMF, 2021, p. 16). The real extent of problem loans on banks’ balance sheets can be expected to start emerging in mid-2021, when remaining forbearance measures are to be lifted.

4 Focus on main COVID-19 pandemic crisis impacts on European banks in 2020

4.1 COVID crisis stopped European banks’ lending expansion

4.2 Moderate crisis-related profitability contraction driven by provisioning

Expressed in local currency, the impairments of European banks increased by 27% in 2020, compared with an 11% increase to EUR 440 million. As a share of year-end loans, their cost of risk moderately increased from 113 basis points to 161 basis points on average. ⁷² The capacity to generate a lower cost of risks than the market leaders may be supported by the niche player positioning. Additionally, while the impact of IFRS 9 on profit or loss was positive in 2019, the tendency ­reversed in 2020 given the sensitivity of the IFRS 9 framework to crisis environments. For credit quality, no strong impacts can be seen in terms of overdue installments, which remained at a level of between 3% and 4% despite the economic downturn. The capital situation of the European banks improved in 2020, with an increase of 2 percentage points on average, as they all increased their capital levels even amid the crisis. Besides, some signs of additional restrictions on capital outflows from Russia were visible in 2020, ⁷³ the treatment of such flows being a ­sensitive topic for European banks.

Table 3: Russia: selected indicators for European banks  

	End-2017	End-2018	End-2019	End-2020
Total assets (liabilities) of European banks	Billion
Total assets measured in RUB	3,211.6	3,868.2	3,853.0	4,254.2
Total assets measured in EUR (yearly average exchange rate) (memo)	48.6	52.4	52.9	51.2
	%
Annual growth measured in RUB	-	20.4	–0.4	10.4
Annual growth measured in EUR	-	7.8	1.0	–3.3
Share in banking system assets	4.1	4.5	4.4	4.2
Total loans	Billion
Lending measured in RUB	1,987.5	2,388.0	2,417.8	2,399.9
Lending measured in EUR (yearly average exchange rate) (memo)	30.1	32.4	33.2	28.9
	%
Share in total Russian banking sector loans	5.1	5.4	5.1	4.5
Annual growthmeasured in RUB	-	20.2	1.2	–0.7
Annual growth measured in EUR	-	7.5	2.7	–13.1
Corporate loans	Billion
Lending measured in RUB	1,210.3	1,499.1	1,389.0	1,376.3
Lending measured in EUR (yearly average exchange rate) (memo)	18.3	20.3	19.1	16.6
	%
Share in total Russian banking sector loans to enterprises	4.7	5.3	4.7	4.3
Retail loans	Billion
Lending measured in RUB	763.1	873.6	1,007.3	1,004.3
Lending measured in EUR (yearly average exchange rate) (memo)	11.6	11.8	13.8	12.1
	%
Share in total Russian banking sector loans to households	6.3	5.9	5.7	5.0
Foreign currency loans	Billion
Lending measured in RUB billion	639.1	664.8	469.1	566.2
Lending measured in EUR (yearly average exchange rate) (memo)	9.7	9.0	6.4	6.8
	%
Share in total loans of European banks	31.8	26.9	19.6	24.5
Share in total foreign currency loans of Russian banking sector	11.3	11.8	9.3	9.1
Total impairments	Billion
Impairments measured in RUB1	(14.7)	(39.4)	(29.0)	(36.6)
Impairments measured in EUR (yearly average exchange rate) (memo)	(0.2)	(0.5)	(0.4)	(0.4)
Average ratios	%
Loan-to-deposit ratio	94.0	91.1	91.2	79.1
Share of highly liquid assets	22.7	26.4	25.0	31.1
Return on average equity	14.5	14.1	12.8	12.0
Return on average assets	2.1	1.8	1.7	1.6
CET1 equivalent (N1.1) (minimum ratio = 4.5%)	11.1	10.4	11.3	12.9
Tier 1 equivalent (N1.2) (minimum ratio = 6%)	11.3	10.9	11.9	13.7
TCR equivalent (N1.0) (minimum ratio = 8%)	14.8	14.1	14.8	16.6
Source: Banks’ regulatory reporting to the Central Bank of Russia (adapted by Fitch).
Note: Reporting does not follow IFRS/European reporting standards but is subject to local Russian reporting standards, which limits data comparability.
1 Reflecting not only loan loss provisions but also other types of impairments.

5 Assessment of current banking risks and shock-absorbing factors with particular focus on European banks

5.1 Credit risk remaining a central risk driver as crisis-related measures end

Regarding credit risk in the nonretail area, the CBR observed an increasing concentration risk trend for significant Russian banks in the studied period (CBR, 2019), as banks are financing a number of large nonfinancial corporations which are gradually increasing their debt levels. Having conducted industry surveys, the CBR announced that it was considering the introduction of macroprudential tools to limit concentration risk, focusing on highly indebted companies and stressing heterogeneous risk measurement methodologies used by banks.

Overall, according to CBR assessments in mid-February 2021, the peak of loan restructuring has passed, and 20% to 30% of restructured loans might become bad loans. Therefore, additional provisions of 2% to 3% of banks’ total loan portfolio may become necessary in the CBR’s view to strengthen financial buffers against credit losses (Golubkova and Ostroukh, 2021; World Bank Group, 2021, p. 28). Taking total loans at end-2020, the need for additional provisions would thus amount to about USD 15 billion to 25 billion.

5.2 Ruble volatility, a risk to mitigate

5.3 Sanctions risk still one of the major risks in the Russian context

5.4 Climate risk as upcoming challenge

5.5 Bank capital and financial profile among shock-absorbing factors

The overall loan-to-deposit ratio (96% at end-2020) is fairly moderate, and ­depositors remain confident, even if relatively low or negative real interest rates have persuaded some to move funds to stock market investments. Noticeably, the European banks decreased their loan-to-deposit ratio in 2020 to rather low levels (about 80%) compared with local peers. Another factor providing a cushion are credit institutions’ net external assets, which have remained at over 7% of total assets. The fact that state-owned banks account for the majority of Russian banking assets (about two thirds) implies that the authorities are directly responsible for the survival of most of the largest players, which may uphold confidence in uncertain times. If necessary, the CBR is equipped with ample liquidity for recapitalization measures (Banking Sector Consolidation Fund). Russia’s international reserves (including gold) expanded by about 40% from end-2017 to late May 2021 and have attained a new historic record level (EUR 496 billion or 40% of GDP – in absolute terms the fourth-largest reserves of the world), outstripping the country’s gross international debt by almost a third.

While current account and budget data deteriorated markedly under the impact of the COVID-19 crisis and crisis-response measures (as mentioned above), total public liabilities remain very modest – about 18% of GDP at end-2020 – even if pronounced deficit financing in 2020 had pushed up the ratio by 4 to 5 percentage points. Meanwhile, Russia’s comfortable net positive international investment position expanded to 34% of GDP (at end-2020), and the country’s budget balance reverted to surplus (in early 2021, see table 1). Overall, the authorities could mobilize substantial funds should banking problems emerge, a scenario whose implications ­remain untested in the case of European banks due to their resilient performance.

6 Outlook for Russia’s banking sector and European banks in Russia after the crisis

The second half of 2021 may well be a test for the market, as regulatory forbearance for asset (loan) valuation and provisioning is slated to expire. The government support program for housing and residential construction is also planned to end in mid-2021. Therefore, the real extent of problem loans on banks’ balance sheets should start to re-emerge in mid-2021. Nonperforming loans will likely go up by some margin. Given banks’ available financial buffers, the sector should be able to bear that. In any case, against the backdrop of the recent (subsidized) re-­acceleration of mortgage lending growth, the mortgage and housing markets warrant careful monitoring, as confirmed by the CBR’s most recently announced regulatory changes. The same goes for consumer loans. Regarding prospective dis-investment of banks nationalized in 2017–2018, the authorities seem to be sticking to their plans to sell the rescued outfits (notably the largest, Otkrytie), but are currently confronted with uncertain market conditions. Unless highly unexpected events intervene (like a major worsening of the pandemic, a – possibly connected – deep and durable decline of the oil price, or a severe escalation of geopolitical tensions and sanctions), economic recovery should stabilize the expansion of banking activity and have a countervailing impact on deteriorating credit quality, which, in turn, would confirm Russia’s role as an important profit source for European banks. That said, the still weak overall investment climate may rein in the potential extent of the recovery. The prospect of continued oil price and ruble volatility, persisting sanctions risks and moderate recovery trends would still provide a challenging ­environment for European banks active in Russia.

Regarding internal models allowing banks to use their own risk parameters and estimates to calculate capital requirements in the future, the CBR was reported as planning to ease capital buffer requirements for banks using internal ratings-­based (IRB) models, and speed up this transition in the crisis context. In early 2020, IRB models were used only by Raiffeisenbank Russia (for corporates) and Sberbank. All the other significant institutions were encouraged by the CBR to gradually shift to IRB over a five-year horizon, with local lender Alfa bank planning to be one of the first banks ready to switch to IRB. As such, the transition to IRB will have additional impacts for Russian banks and the competitive landscape. This situation illustrates that the European banks in Russia can benefit from the advantages of such regulatory-driven tools like the domestic players. More broadly, the European banks were not found to be in an unfavorable position compared to their local peers regarding the regulatory regime.

Finally, dynamic digital transformation trends in the Russian banking market bring both risks and opportunities, as banks need to modernize in a competitive environment – in a context where market leader Sberbank rebranded itself as “Sber” in 2020, unveiling a new conglomerate-like business model with a broad range of nonbank services (e-commerce, food delivery, mobility, etc.). The pandemic served as a real test for the digital capacity of financial institutions, with the Russian market being perceived as advanced within the CESEE area. Even if branch activities were among the types of services maintained during the lockdown, online banking is on an increasing trend. In this regard, the European banks active in Russia pursue different strategies. Société Générale (SocGen) has reduced its branch infrastructure but nevertheless retains a large physical network spread all over the country. UniCredit’s corporate-driven business model makes it less reliant on branches servicing retail customers. Raiffeisenbank Russia communicated to investors in 2019 a strategy to close around 25% of its branches by 2020 and to put further focus on the Moscow and Saint-Petersburg areas, reducing its geographical footprint in favor of digital networks leveraging on nationwide trends. Summing up, the European banks operating in Russia, with various degrees of profitability, will have to contend with the medium-term perspective of not only sustained sanctions risks and limited economic growth, but also intensified competition. To be prepared for these challenges, they can be expected to go digital with more services, embrace cost discipline with restructuring measures and the like (mergers among SocGen subsidiaries being a case in point) and engage in selective risk-taking to preserve asset quality in the aftermath of the pandemic.

References

Barisitz, S. 2018. The Russian banking sector: between instability and recovery. In: Financial ­Stability Report 35. OeNB. 59–66.

Barisitz, S. 2019. Nonperforming loans in CESEE – a brief update on their definitions and recent developments. In: Focus on European Economic Integration Q2/19. OeNB. 61–74.

BOFIT Weekly (ed). 2021. Moscow stock exchange made gains in second half of 2020. January 22.

CBR. 2019 (Q2–Q3). Financial Stability Review 2 (15).

CBR. 2020 (Q2–Q3). Financial Stability Review 2 (17).

CBR. Banks regulatory forms (collected by the Central Bank of Russia, adapted by Fitch).

Fitzgeorge-Parker, L. 2020. Alfa-Bank chief executive: ‘Russian customers can live with lockdown much better than those in Europe’. In: Euromoney. April 2.

Fitzgeorge-Parker, L. 2021. Russia’s retail revolution. In: Euromoney. April 11.

Foy, H. 2020. Russia: adapting to sanctions leaves economy in robust health. In: Financial Times. January 30.

Golubkova, K. and A. Ostroukh. 2021. Russian banking sector ready to live without COVID-19 support, says central bank governor. In: Reuters News. February 18.

IMF. 2019. Russian Federation – Staff Report for the 2019 Article IV Consultation. July 16.

IMF. 2021. Russian Federation – Staff Report for the 2020 Article IV Consultation. January 19.

King, J. 2020. Top 50 Russian banks ranking: a return to stability. In: The Banker. February 3.

Kuznetsov, I. 2020. Pochti 650 tys. novykh roznichnykh investorov vlozhilis v paevye investitsionnye fondy (PIF) v III kvartale – istorichesky rekord! In: Ekonomika i Zhizn. 27 noyabria. 3.

Pugsley, J. 2020. COVID-19 casts shadow over Russian bank reform push. In: The Banker (Financial Times). July 15.

Raiffeisen Research (ed). 2019. CEE Banking Sector Report 2019. – Close to or with new ­records. July.

Raiffeisen Research (ed). 2020. CEE Banking Sector Report 2020. – Crisis & recovery mode at the same time. November.

Simola, H. 2020. Climate change and the Russian economy, BOFIT (Bank of Finland) Policy Brief 2020 11.

World Bank Group. 2021. Russia’s economic recovery gathers pace. Special focus on cost-effective safety nets. Russian Economic Report 45. May.

Event wrap-ups

87th East Jour Fixe: CESEE’s second transition – challenges on the road to ­low-carbon economies

Andreas Breitenfellner, Mathias Lahnsteiner, Thomas Reininger ⁷⁵

Climate change and policies: where do CESEE economies stand?

In the following discussion, moderated by the OeNB’s chief economist Doris Ritzberger-Grünwald, Ada Ámon (Chief Climate Advisor to the Mayor of Budapest) and Piotr Arak (Director of the Polish Economic Institute) shared their views on ­climate-related issues from both national and European perspectives. The two discussants reflected critically on the climate change awareness of people living in Hungary and Poland and their willingness to take action. They also covered the coal industry in Poland, the use of EU funds for climate-related projects in CESEE, and implications of EU climate policies for Western Balkan economies. Ámon ­argued that – at least in the capital Budapest – there certainly is awareness, but no willingness to take the actions needed. She also sensed that the ruling government may not be interested in meaningfully addressing climate policy issues in Hungary. In her opinion, EU taxpayers should watch closely how EU funds are being used for climate-related projects and whether these projects are being realized in the end. Regarding the Western Balkans, she highlighted the large heterogeneity between countries with respect to their energy mix and their huge potential for renewable energy. Arak explained that people in Poland are aware of climate change, but do not want to pay more for energy. In this respect, he raised the issues of energy poverty and energy cost inflation (in part driven by rising emissions trading scheme (ETS) prices) in his country. Concerning Poland’s energy mix, he drew attention to investments in wind parks and photovoltaic installations as well as to further investment plans to increase the share of these renewable energy sources. He pointed to the political importance of the country’s coal regions, whose population accounts for a large number of votes. The loss of jobs in the coal industry, which dates back to the 1990s, is still subject to political debates in Poland. Finally, Arak argued that it is not realistic for Poland to become climate neutral by 2050.

Investment needs for CESEE’s green transition

Steering the financial sector to address the climate challenge