1. Monetary and financial policies for "de- euroization" - a case study of recent Croatian experience Tomislav Galac & Evan Kraft Croatian national bank

2. 2 Motivation  Episodes of significant currency depreciations during the 2008-2009 global crisis -> risks of F/C lending in CEE countries with high levels of credit euroization exposed: debt burden and debt-servicing burden up steeply and suddenly (EBRD, 2010) -> NPL ratios soared (Rainer and Haiss, 2010) -> monetary authorities support domestic currency to prevent balance-sheet risks from materializing, BUT these measures in themselves are inherently pro-cyclical -> uncertainty about the ultimate net-effect on NPLs

3. 3 Motivation  Croatia – one of the most euroized countries in Europe  Policy advice for reducing financial euroization (Zettelmeyer et al, 2010; Rainer and Haiss, 2010; and Kokenyne et al, 2010): 1) building a more flexible ER regime in a low inflation environment; 2) developing markets for local currency debt instruments; 3) increasing the price of F/C borrowing and decreasing the scope for legal uses of F/C 4) building F/C reserve buffers for periods of ER pressures + arranging contingency lines with the ECB and IMF  Monetary and financial policies of 2003-2007 similar to (3)-(4), for slowing credit growth and capital inflows, seem to have reduced euroization level prior to crisis -> natural experiment

4. 4 Recent literature  Causes and consequences of the so called unofficial financial euroization (Ize and Levy-Yeyati, 2003, and Levy-Yeyati, 2006): inflation, ER, FS related  If banks extend D/C loans when majority of deposits are F/C, the banking system becomes open to ER risk of nominal depreciation of D/C  Solutions: 1) fixing the nominal ER -> reduces the freedom of monetary policy; 2) limiting banks' net-open F/C positions -> credit euroization or underbanking if borrowers cannot hedge against F/C risk  Problem with credit euroization (where FIs’ net-open F/C positions are limited by regulation): foreign currency-induced credit risk (FCICR) - a systematic risk of default by those who borrow in F/C while at the same time having no F/C income/savings

5. 5 Recent literature  Euroization in Croatia in earlier periods has been studied by Šonje and Vujčić (1999), Feige (2002), Kraft (2003), and Kraft and Šošić (2006) among others  Authors argued that the prolonged period of high deposit euroization after the successful disinflation program in 1994 and the end of the war in the late 1995 was due to the deep trauma of these experiences, resulting in the so called "persistent" deposit euroization  Authors further argued that the only way to develop post-war credit markets, while maintaining sound F/C risk management at financial institutions was to allow F/C lending & limiting net-open F/C position -> high credit euroization -> FCICR

6. 6 Our Goal & Method  We look at euroization in Croatia mostly during 2000-2010  We use three datasets for descriptive analysis, but only one for econometric analysis: monthly dataset, which runs from Jan. 2004, has a total of 81 observations, and thus covers most of the interesting period in the context of central bank behavior, although it provides the currency breakdown only at the very aggregated level  We try to determine how much the CNB policies and measures in 2004 -2010 have contributed to partial de-euroization in 2002-2007, and re- euroization in 2008-2010, along with looking at other “standard” determinants of financial euroization in Croatia

7. 7 Euroization in Croatia – Stylized Facts

8. 8

9. Monetary and financial policies in Croatia  Main goal (2003-2007): slowing rapid credit growth fueled by abundant and inexpensive foreign capital and curtailing negative side-effects of the two (CGR, MRR, CAR measures), and in 2008- 2010 discouraging disorderly capital outflows (removal of MRR, increase of KHR, maintenance of CGR)  Some policies/measures aimed directly at reducing the level of financial euroization in the country, or at least at building buffers for dealing with potential consequences: MRR, FCLR, KHR, FCICR risk-weight add-ons  Others may have contributed to “status quo”: NOFCP limits

10. M/F policies - FCLR

11. M/F policies – FCICR add-ons

12. M/F policies – MRR

13. M/F policies – CGR

14. M/F policies – KHR

15. 15 Modeling euroization in Croatia Model 1:

16. 16 Modeling euroization in Croatia  Model 1 / main results:  interest rate spreads do not appear to be important determinants of the dynamics of financial euroization  spreads do appear to react to central bank signals and general economic environment in a more or less expected manner  inflation, MRR dummy, crisis dummy, and FCLR rate have expected and statistically significant impact on the dep. spread  crisis dummy and inflation have a significant and expected impact on the loan spread

17. 17 Modeling euroization in Croatia  Model 1 / modeling issues: a number of significant unexpected coefficients, difficult to interpret absence from the final model of the variables representing some presumably very important central bank measures difficult to explain: 1) FCICR risk-weight add-ons not important in the loan interest rate spread equation, 2) inclusion of F/C indexed deposits into the FCLR base not important in the deposit interest rate spread equation (anecdotal evidence on the former would suggest that banks simply began offering long term HRK denominated loans, and on the latter that they stopped collecting indexed deposits)  The robustness of Model 1 results is reexamined by estimating Model 2, arrived at by dropping the apparently statistically irrelevant interest rate spreads from the system

18. 18 Modeling euroization in Croatia Model 2:

19. 19 Modeling euroization in Croatia  Model 2 / main results:  Inflation and ER movements again display some expected and some unexpected effects  ER movements are not important for deposit euroization, while lagged inflation changes have negative coefficients. This could be a direct consequence of modeling nominal rather than inflation-adjusted quantities  ER movements appear very important for credit euroization and with the expected positive sign, while inflation is mildly insignificant

20. 20 Modeling euroization in Croatia  Existence of MRR and FCICR weight add-ons (before their increase) appear to be much more important for credit euroization than deposit euroization, with the expected negative impact  Expected negative impact is found for the period after the indexed deposits were added to the FCLR base, both for the deposit and credit euroization  For both euroization measures, expected positive correlation is found between their changes and lower rates of FCLR in the crisis period  CGR displays positive correlation with changes in both the deposit and credit euroization. This should probably be explained by the coincidence of CGR and the crisis, but possibly also by the primary effect of the CGR - a reduction of the bank credit growth which could in the process alter the composition of the stock of bank credit

21. 21 Modeling euroization in Croatia  Unlike in Model 1, the crisis dummy is not present in Model 2, indicating that its effect may be accounted for by other variables in the specification, in the credit equation possibly by the exchange rate movement, in the deposit equation by three successive reductions in the FCLR ratio, and by CGR in both equations  The feedback between deposit and credit euroization appears rather weakly in Model 2. It is however possible that this relationship is heavily confounded in the model by the dummies for those central bank measures which had an impact on both the deposit and loan side of the model. This proposition is tested by estimating Model 3...

22. 22 Modeling euroization in Croatia Model 3:

23. 23 Modeling euroization in Croatia  Model 3 / main results:  On the credit side, CGR loses its significance, while initial introduction of FCICR weight add-ons remains highly significant with the expected negative sign. The inflation coefficient remains negative on the deposit side, while the exchange rate becomes insignificant in both equations, as well as inflation in the credit equation  The estimated feedback coefficients assume the expected signs and become significant in this model. The most important central bank measure on the deposit side is now change of the FCLR rate, which had contributed to re-euroization in the crisis period. The introduction of indexed liabilities into the FCLR base is only significant in the first two months after the fact, and not for the entire period after its implementation

24. 24 Conclusion  CNB measures could likely be credited with partial deeuroization in 2004-2007, but may have also contributed to re-euroization later in the crisis period  Caveat: it is difficult to attribute general effects to particular measures, and this effort appears largely model-specific, but introduction of FCICR weight add- ons, and of F/C indexed dep. into the FCLR base may be the key to partial de-euroization in the pre-crisis period

25. 25 Conclusion  Inflation does not seem to play the role in the period analyzed, or if it does it is masked by its unexpectedly negative effect on both measures which should most likely be attributed to using nominal and not real stocks of loans and deposits in the analysis  The ER changes do not appear important, except in one model specification where they appear with the expected positive sign

26. 26 Conclusion  The question of feedback between the deposit and credit euroization appears to have a model-specific answer. Either, there is a strong and expectedly positive relationship between the two, but there is no room for nominal exchange rate movements and less room for central bank measures in the explanation of financial euroization in Croatia, OR  CNB measures should be interpreted as having influenced directly both the credit and deposit euroization in the past

27. 27 Final thoughts  Our findings could be interpreted as indicating that even within a context of a rigid exchange rate regime, a combination of monetary and prudential measures could lead to a reduction in financial euroization during normal times  Caveat: policy had some effect in Croatia, but far from creating full de-euroization in the sense of the literature: “the euroization level, measured as foreign currency deposits in broad money, reduced by 20 percentage points and below 20% without major macroeconomic costs.”