Capital adequacy ratio Ratio of liquid assets
65 55 45 35 25 15 5 65 55 45 35 25 15 5 2007 2009 2007 2009 2007 2009
Croatia FYR Macedonia Turkey
45 40 35 30 25 20 15 10 45 40 35 30 25 20 15 10 2007 2009 2007 2009 2007 2009
Croatia FYR Macedonia Turkey
Non-performing loan ratio Return on assets
20 16 12 8 4 0 20 18 16 14 12 10 8 6 4 2 0 2007 2009 2007 2009 2007 2009
Croatia FYR Macedonia Turkey
4 3 2 1 0 -1 -2 -3 4 3 2 1 0 -1 -2 -3 2007 2009 2007 2009 2007 2009
Croatia FYR Macedonia Turkey
Loan-to-deposit ratio 140 120 110 130 100 70 80 90 60 140 120 110 130 100 70 80 90 60 2007 2009 2007 2009 2007 2009 Croatia FYR Macedonia Turkey
Source: National sources.
F E A T U R E S
With regard to the return on assets, the three candidate countries are in very different positions. Although in Croatia the entire return distribution shifted downwards, the indicator remains positive for three-quarters of the banks. In the former Yugoslav Republic of Macedonia, the decrease is much more pronounced, with individual values for almost half of the banks in the negative range. This may be due to increased impairment charges related to the adverse developments in loan portfolio quality (see the respective country section). In stark contrast to these two countries, in Turkey the average return on assets remained high and all banks reported positive returns on account of improved interest and trading incomes (see the respective country section for more details).
Turning to the loan-to-deposit ratio, its average has increased in Croatia and the former Yugoslav Republic of Macedonia, reaching a value very close to 100% in the former. This development can be traced back to the episodes of deposit withdrawals and the incomplete recovery of deposit levels in these two countries. The former Yugoslav Republic of Macedonia stands out, with outliers on the upper side, a third quartile of 129% and a maximum value as high as 260%, suggesting a very high level of credit exposure for some banks.46
4.2 RESULTS OF MACRO STRESS TESTS IN THE CANDIDATE COUNTRIES
The recent global downturn and the possible spillovers into the fi nancial sector have directed attention to macro-prudential analysis methods. This horizontal issue presents the results of a two-step macro stress-testing exercise for Croatia, the former Yugoslav Republic of Macedonia and Turkey. First, the major sources of risk were identifi ed and their potential impact on the credit quality of the banking system was assessed. In the second step, the identifi ed link between macroeconomic factors and credit quality was used to project the likely impact of different macroeconomic scenarios on individual banking institutions and on the system as a whole. This allowed an assessment
of the propensity of the fi nancial system to withstand shocks and an evaluation of the degree of systemic risk. Interbank exposures and hence potential contagion were also taken into account.47 The main fi nding is that under extreme but not implausible scenarios, the banking systems of the three candidate countries appear well-capitalised and suffi ciently endowed with liquidity to withstand shocks, provided that such shocks do not occur simultaneously.
The main sources of risk in the candidate countries appear to be the unfavourable external environment lasting longer than currently expected and a further deterioration in the domestic business cycle, spilling over into the local banking systems through a deterioration in credit quality and constrained access to foreign funding. Moreover, the high share of foreign-denominated lending is an additional source of risk in the event of adverse exchange rate developments.48
The link between economic activity and credit quality is addressed empirically through a simple econometric model using historical data. Changes in credit quality (measured by the ratio of non-performing loans to total credit) are regressed to allow for lags in real GDP growth and changes in the exchange rate. The estimated results for the individual country models are listed in the fi rst three columns of Table 24. There are two important caveats, namely the low number of observations and the simplifi ed model structure, disregarding potential feedback effects from credit quality to the macroeconomic variables of the model. Nevertheless, in line with intuition, a real GDP decline is related to worsening credit quality as a consequence of the increased diffi culty on the part of borrowers to
This high loan-to-deposit ratio for some banks in the former 46
Yugoslav Republic of Macedonia mainly concerns banks that are not very active in attracting deposits.
The risk of possible interbank contagion is taken into account 47
by estimating bilateral exposures using the maximum entropy principle. See, for example, Upper and Worms (2002). For a detailed overview of recent trends in foreign exchange- 48
repay their loans. In the individual regressions, the differing lag structures and coeffi cients suggest that the magnitude and speed of the impact will vary across the countries. For instance, the impact of a 1 percentage point decrease in real GDP appears to be much stronger in the former Yugoslav Republic of Macedonia and Turkey than in Croatia (by considering the cumulative value, obtained by taking the sum of GDP lag coeffi cients). The exchange rate is signifi cant for Croatia and Turkey, meaning that an increase in the exchange rate (depreciation of the domestic currency) might lead to an additional increase in NPLs. These individual results are broadly compatible with those of an extended panel model covering several additional eastern European countries (shown in the fourth column of Table 24). Although the individual regressions provide an insight into the country-specifi c credit quality patterns, the panel arrangement was preferred for compiling the NPL forecast owing to its larger number of observations and better out-of-sample forecasting properties.
In the second step of the process, the estimated coeffi cients are used as an input for a macro stress test of the banking system using an approach similar to the IMF stress-testing method.49 It considers the impact of a one-off NPL increase on the bank’s capital adequacy
ratio via additional provisions,50 and does not take profi ts into account as a buffer against potential shocks. This gives the method a static character, but it is a prudent way of assessing the shock impact (Cihak, 2007). For each country, the top ten banks in terms of total assets are considered.51
A baseline and a downside credit shock scenario are considered for all countries, as follows: country-specifi c GDP forecasts are used to obtain plausible forecasts for NPL ratios. At the same time, it is assumed in the baseline scenario that exchange rates remain stable. The downside scenario incorporates the idea of a double-dip recession and has identical assumptions across the countries: a real GDP decline in the fi rst three quarters and a return to the projected path from the baseline scenario thereafter, combined with a 5% domestic currency depreciation in two consecutive quarters. The magnitude and quarter-on-quarter profi le of the GDP decline is broadly in line with those observed in the three candidate countries in the period 2008Q3-2009Q2 during
See Čihák (2007). 49
A uniform provisioning rate of 50% of new NPLs was assumed 50
for all banks.
An important caveat in these calculations is the timeliness of 51
the bank data used as input: data are taken from BankScope and refer to the end of 2008 given the unavailability of more recent and suffi ciently detailed individual bank data.
Table 24 Econometric estimates: credit quality model
(Dependent variable: percentage change in the non-performing loan ratio)
FYR Macedonia Croatia Turkey Panel regression
Constant 0.77 2.27** -0.94 2.29** ΔGDP ΔGDP (-1) -1.72** -1.78*** ΔGDP (-2) -2.36** -2.17*** -1.32** ΔGDP (-3) -2.86** -1.41** ΔGDP (-4) -1.30** ΔER ΔER (-1) 1.98** 0.61** 0.58*** ΔER (-2) ΔER (-3) 2.14** 0.35** 0.69*** Adj. R2 0.34 0.51 0.53 0.38 Sample 2003Q4-2009Q4 1999Q2-2009Q4 2002Q1-2009Q4 2000Q2-2009Q4
Source: ECB staff estimates using data from national sources.
Notes: *, ** and *** denote coeffi cients signifi cant at the 10%, 5% and 1% level respectively. The exchange rate used is the nominal effective exchange rate, expressed as the domestic currency price of a trade-weighted basket of currencies. The unbalanced panel includes the three candidate countries, as well as Bulgaria, Romania, Russia and Ukraine. The panel regression uses a fi xed-effect estimate, and the length of the time series varies country by country.
F E A T U R E S
the recent crisis. Moreover, the resulting NPL change is also in line with assumed NPL increases in stress tests conducted by the
respective central banks.52 In addition to the two credit scenarios, a liquidity scenario is also considered, which assumes the withdrawal of deposits that have to be paid out of the liquid assets available to the bank. It is also designed in two versions: a moderate and a severe bank run, corresponding, respectively, to 4% and 8% deposit withdrawals on each of two consecutive days.53
The following panel of charts compares the results of the two credit shock scenarios for the three candidate countries, presenting the weighted average along with the fi rst and third quartile of the capital adequacy ratios for the top ten banks before and after each of the shocks. For Croatia and Turkey the impact of the baseline and even of the downside scenario is relatively modest, with the top ten banks staying well above the minimum capital adequacy ratio. Although the average decrease in the capital adequacy ratio is more substantial for the former Yugoslav Republic of Macedonia, there are no bank failures 54 or contagion. However, the shock might also result in some institutions not meeting the minimum capital requirements or having insuffi cient Tier 1 capital,55 and thus requiring recapitalisation. The estimated recapitalisation needs of the top ten banks resulting from the two scenarios are modest when measured as a share of GDP (in Turkey, no recapitalisation is needed even under the downside scenario).
As a result of the liquidity scenario (not shown in the charts), the banks appear to have suffi cient liquid assets to withstand the assumed deposit withdrawals: in the three countries, there is
For instance, for Croatia the NPL increase in the downside 52
scenario (55%) is comparable with the 48% increase in the stress-testing of the CNB, see CNB (2009b) as of August 2009. The “moderate” rate of withdrawal roughly corresponds to 53
the withdrawal rate during the run on the Kaupthing bank (Iceland, 2009).
A bank failure is defi ned as equity capital falling below zero. 54
Tier 1 capital consists of equity capital and Tier 2 of hybrid 55
capital and subordinated debt. The requirement that Tier 2 capital makes up no more than 50% of total capital forms part of the Basel I accord.
Chart 6 Changes in the capital adequacy ratio