The impact of expected losses provisioning on credit growth: the case of Mexico Closing Conference of the BIS CCA CGDFS Working Group June 13 rd

Size: px

Start display at page:

Download "The impact of expected losses provisioning on credit growth: the case of Mexico Closing Conference of the BIS CCA CGDFS Working Group June 13 rd"

Sharon Cummings
5 years ago
Views:

1 The impact of expected losses provisioning on credit growth: the case of Mexico Closing Conference of the BIS CCA CGDFS Working Group June 13 rd Mexico City Banco de México, DGEF

2 The impact of expected losses provisioning on credit growth: the case of Mexico 1 Gabriel Levin Konigsberg Banco de México Calixto López Castañón Banco de México Fabrizio López-Gallo Dey Banco de México 1 The views expressed on this presentation are those of the authors and do not necessarily reflect the opinion of Banco de México or its governing board 2 / 31

3 Introduction Ever since the last financial crisis, macro-prudential policies have gained importance in terms of financial stability Yet, there are not enough studies on the effectiveness of this policies, and their general impact One of the most relevant effects that should be studied is the effect of macro-prudential policies on credit Here we will study loan loss provisioning, a common micro-prudential policy that in Mexico was given a macro-prudential perspective 3 / 31

4 A quick literature review Flores et al (21) study loan losses provisioning in Mexico as a macroprudential tool Jiménez et al (212) find that dynamic provisioning in Spain help smooth credit supply cycles Claessens (213) et al find macroprudential policies to be effective in different dimensions Cerutti (215) et al find that these policies help reduce credit growth Levin Konigsberg (215) finds loan loss provisions in Mexico to be wellfare improving 4 / 31

5 Motivation In Mexico, credit provisioning rules considerably misestimated expected losses To solve this problem, the Mexican regulator changed the provisioning rules The new provisioning rules intend not only to accurately estimate expected losses but also to deal with systemic risk Given this target the new provisioning rules can be considered a macroprudential policy The new rules substantially increase the amount that should be provisioned for each loan This in turn makes giving credit more expensive for banks 5 / 31

6 The old rules The old system calculated provisions in an extremely simple fashion Every loan was to be provisioned in a proportion that depended only on how much time its payment was past due As a result provisions were extremely backward-looking and mimicked non-performing loans Past due Provisioning periods proportion / 31

7 Non-performing loans and provisions Consumption loans Non-performing loans Non-performing loans Non-performing loans m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 1 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 2 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 3 Non-performing loans Non-performing loans Non-performing loans m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Source: cnbv Bank 4 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 5 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 6 7 / 31

8 Non-performing loans and provisions Commercial loans Non-performing loans Non-performing loans Non-performing loans m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 1 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 2 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 3 Non-performing loans Non-performing loans Non-performing loans m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Source: cnbv Bank 4 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 5 22m1 24m1 26m1 28m1 21m1 212m1 214m1 216m1 Bank 6 8 / 31

9 The new rules Basel III framework contemplates the use of expected losses provisioning In Mexico, the Banking Commission (cnbv) modified the methodology used to determine provisions This methodology was introduced in a gradual fashion: Credit cards (October 29) Non-revolving consumption loans and mortgages (March 211) Credits granted to state and local governments (October 211) Commercial loans (June 213) 9 / 31

10 The new rules use econometric techniques to calculate PD, EAD, and LGD on a loan-by-loan basis, to obtain the expected losses of each credit as the product of these three factors PD is estimated for each credit as a function of the credit current characteristics (eg arrears, value of the loan, payment periodicity) Although it is not clear that the new provisions reflect the expected loss, they have changed the cost of supplying a loan 1 / 31

11 Losses and provisions Consumption loans Losses Losses Losses Before After Before After Before After Bank 1 Bank 2 Bank 3 Losses Losses Losses Before After Before After Before After Source: cnbv Bank 4 Bank 5 Bank 6 11 / 31

12 Losses and provisions Commercial loans Losses Losses Losses Before After Before After Before After Bank 1 Bank 2 Bank 3 Losses Losses Losses Before After Before After Before After Source: cnbv Bank 4 Bank 5 Bank 6 12 / 31

13 Performing loan-value real growth Commercial loans Real growth rate (de-meaned) Bank Before After Large Bank Source: cnbv 13 / 31

14 Data For loan information we use regulatory report R4C which consists of the universe of commercial loans given by banks operating in Mexico We use quarterly data at the account level, ie our observation unit is the value of the loans outstanding of each creditor on each bank Our sample spans from 29q3 to 215q1 and includes around 63, different accounts given by 4 different banks We drop one percent observations on each tail Bank balance-sheet data is taken from information published by cnbv, the Mexican supervisor Macroeconomic data is taken from information published by Banco de México 14 / 31

15 Summary statistics Variable Description Mean SD policy rate Change in monetary policy rate log gdp Log-change in GDP fx USD-MXN exchange rate current acc Current account -586e+7 397e+7 log assets Log of assets roa Return on assets 4353% % liquidity Cash and securities to total assets ratio funding Deposit to liabilities ratio c ratio Tier 1 Capital ratio Median 15 / 31

16 Summary statistics by currency and firm size (loan growth) Observations Large Medium/Small Local 144, ,194 Foreign 21,727 31,358 Mean Large Medium/Small Local Foreign S D Large Medium/Small Local Foreign Foreign/Large Local/Large Foreign/Small Local/Large 16 / 31

17 Loan growth distribution for some banks Credit growth excludes outside values Before After 17 / 31

18 Impact of the new provisioning rule on credit growth To evaluate the impact of the new provisioning rules on credit growth, we estimate the following equation: 4 log credit ibt = α ib + δ bt + δbt m + δn bt + X tβ 1 + Y bt β 2 + q jt + ε fbt Where: credit ibt is the total value of the outstanding loans given to firm f by bank b at time t δ bt is a dummy variable that is worth for banks using the new methodology from the moment it was implemented j=1 δ n bt is a dummy variable that is worth from the moment bank n was authorized to use its own methodology δ m bt is a dummy variable that is worth from the moment bank m was authorized to use its own methodology X t is a vector of controls that vary across time, macroeconomic variables Y bt is a vector of controls that vary across bank and time, bank-specific characteristics 18 / 31

19 Variables of interest According to the Mexican regulation, banks may either use the methodology proposed in the regulation or request authorization to use their internal methodology We therefore identify different effects One captured through δ b, that encompasses the effect of the methodology as established in the regulation And the effect of the use of internal methodologies One for each bank that is authorized to use internal methodologies (δ n b and δ m b ) δ b Bank 1 Old New Bank 2 Old New Bank n Old New δ n b Bank 1 Old New Bank 2 Old New Bank n Old New / 31

20 Regression results for local currency loans All Large Medium Sectoral Growth δbt (646) (698) (668) (599) δbt m (712) (487) (734) (579) δbt n (936) (592) (975) (895) log gdpt (169) (142) (175) (966) policy ratet (17) (17) (11) (151) Current Acct 488e-1 377e-1 492e-1 566e-1 (19e-1) (731e-11) (111e-1) (123e-1) FXt (455) (249) (467) (65) VIXt (529) (45) (546) (583) log assetst (322) (254) (337) (349) c ratiot (13) (17) (18) (115) fundingt (158) (17) (159) (129) liquidityt (158) (117) (159) (12) roat (251) (299) (262) (338) Constant (452) (324) (473) (468) Creditor-Banf Fe Yes Yes Yes Yes Seasonal Dummies Yes Yes Yes Yes Clustered Standard Errors By bank By bank By bank By bank Observations R Standard errors in parentheses p < 1, p < 5, p < 1 2 / 31

21 Regression results for foreign currency loans All Large Medium Sectoral Growth δbt (847) (173) (126) (98) δbt m (115) (189) (116) (122) δbtn (953) (123) (123) (15) log gdpt (436) (579) (479) (1) policy ratet (299) (283) (467) (449) Current Acct 344e-1 199e-1 477e-1 364e-1 (185e-1) (154e-1) (21e-1) (166e-1) FXt (711) (94) (822) (76) VIXt (615) (132) (778) (776) log assetst (354) (577) (388) (351) c ratiot (335) (859) (335) (337) fundingt (16) (193) (142) (114) liquidityt (19) (157) (145) (112) roat (427) (527) (743) (464) Constant (477) (736) (543) (473) Creditor-Bank FE Yes Yes Yes Yes Seasonal Dummies Yes Yes Yes Yes Clustered Standard Errors By bank By bank By bank By bank Observations R Standard errors in parentheses p < 1, p < 5, p < 1 21 / 31

22 In order to analyze the interaction of the provisioning policy with economic growth and monetry policy, we estimate the following equations: log credit ibt = α ib + δ bt + δbt m + δbt n + log gdp tδ bt + log gdp tδbt m 4 + log gdp tδbt n + X t β 1 + Y btβ 2 + q jt + ε fbt log credit ibt = α ib + δ bt + δbt m + δbt n + pol rate tδ bt + pol rate tδbt m 4 + pol rateδbt n + X t β 1 + Y btβ 2 + q jt + ε fbt j=1 j=1 22 / 31

23 Regression results for foreign currency loans Local currency Local currency Foreign currency Foreign currency δbt (124) (649) (16) (788) δbt m (113) (777) (17) (125) δbt n (118) (982) (116) (78) log gdpt (146) (184) (471) (672) log gdpt δbt (211) (476) log gdpt δbt m (228) (343) log gdpt δbt n (246) (313) policy ratet (179) (37) policy ratet δbt (343) (369) policy ratet δbt m (276) (485) policy ratet δbt n (271) (446) Current Acct 533e-1 561e-1 362e-1 474e-1 (122e-1) (123e-1) (196e-1) (222e-1) FXt (54) (53) (778) (466) VIXt (576) (568) (665) (644) log activot (317) (351) (351) (355) c ratiot (14) (116) (352) (344) fundingt (16) (163) (11) (117) liquidityt (161) (161) (18) (18) roat (263) (35) (43) (55) Constant (443) (476) (474) (46) Creditor-Banf Fe Yes Yes Yes Yes Seasonal Dummies Yes Yes Yes Yes Clustered Standard Errors By bank By bank By bank By bank Observations R Standard errors in parentheses p < 1, p < 5, p < 1 23 / 31

24 Conclusions The new provisioning rules in Mexico had a negative impact on credit growth When taking in consideration the size of the firm, this negative effect is smaller for large firms than for medium and small-sized firms This may be due to their relative less-riskiness The effect appears to be smaller for foreign currency-denominated loans In any case, we do not find a significant effect for large firms in foreign currency-denominated loans 24 / 31

25 Banks that use internal models to calculate their provisions appear to, in some cases, have increased credit This suggests that internal models allow banks to manage their costs more efficiently There appears to be no interaction between the new provisioning policy and monetary policy An important caveat to this is that in our period of study, there is only loosening of monetary policy As of the interaction of the policy with economic growth, we find a negative interaction between the internal models and economic growth 25 / 31

27 Appendices 27 / 31

28 Mexican banking system Lorentz curve Source: cnbv 28 / 31

29 Commercial credit to gdp ratio / 31

30 Credit to GDP ratio Source: Banco de México 3 / 31

31 Mexican banking system capital adequacy ratio Source: cnbv 31 / 31

International Banking and Cross-Border Effects of Regulation: Lessons from Mexico

International Banking and Cross-Border Effects of Regulation: Lessons from Mexico Gabriel Levin-Konigsberg, a,b Calixto López, a Fabrizio López-Gallo, a and Serafín Martínez-Jaramillo a a Banco de México