The recent financial crisis has redefined the broad contours of regulation of financial sector, globally. The G-20 Working Group 1 on Enhancing Sound Regulation and Strengthening Transparency, constituted in 2008, submitted its recommendations, which formed the blueprint for the global regulatory reform agenda. Procyclicality has been among the identified underlying causes for the recent crisis. Various measures have accordingly been proposed by international standard setters to address the problem of procyclicality. One such measure put forward by the Basel Committee on Banking Supervision (BCBS) was countercyclical capital buffers. In December 2010, BCBS issued Guidance for national authorities operating the countercyclical capital buffer recommending, inter alia, a buffer guide for the consideration of national authorities. In this context, relevant questions are: is the suggested buffer guide suitable for Emerging Market Economies (EMEs), especially for those which largely rely on retail model of banking business akin to India’s? If not, is there any alternative buffer guide more suitable for them? If so, what is it? This paper endeavors to address these questions of relevance in detail and is organised into six sections. Section 1 outlines the BCBS set of proposals with regard to the guidance to the national authorities for operating the countercyclical buffer. Section 2 examines its suitability to EMEs. Section 3 outlines a literature review enumerating studies on the subject. This section attempts to position the paper in the evolution of the literature. While Section 4 analyses methodology for constructing an alternative buffer guide customised to macro-financial environments in which banks in EMEs operate, Section 5 presents the historical performance of the proposed alternative buffer guide in the Indian context. Finally, Section 6 concludes.

2. The BCBS Proposals for Operating Countercyclical Capital Buffer 2.1

BCBS recommends that deviation of credit-to-GDP ratio from its long-term trend could be a useful starting reference point for assessing the build-up of systemwide risk in the financial system. Before recommending this buffer guide, BCBS considered a broad range of indicator variables. The variables assessed can be divided into three groups. The first includes aggregate macroeconomic variables: GDP growth, (real) credit growth and deviations of the credit to GDP ratio from a long term trend and deviations of real equity prices as well as real property prices from their respective long-term trend. The second includes measures of banking sector performance: profits (earnings) and proxies for (gross) losses. The final group includes proxies for the cost of funding, in the form of credit spreads. BCBS felt that the credit-to-GDP gap was the best performing of the range of variables considered. It is smoother and normalised by the size of the economy, and is therefore not influenced by the normal cyclical patterns of credit growth: Furthermore by being based on credit, it has the significant advantage over many of the other variables by appealing directly to the objective of the countercyclical capital buffer, which is to achieve the broader macroprudential goal of protecting the banking sector from periods of excess credit growth. How does the credit-to-GDP ratio behave in EMEs? An attempt is made to assess and verify how the credit-to-GDP gap performs in the Indian context. Figure 1 presents time series of bank credit- to- GDP ratio and its long-term trend since 1950-51. The trend line was based on the Hodrick-Prescott (HP) filter with λ = 400,000 (as suggested by the BCBS guidance). Some of the observations of the figure 1 are as follows. Firstly, in the 1950s, 1960s, 1970s, and 1980s, credit-toGDP ratio was almost close to the trend. Secondly, in the 1990s the credit-toGDP ratio began slightly deviating from the trend but basically remained below but close to the trend up until 2002-03. Lastly, since 2003-04 the credit-to-GDP ratio remained above the trend and positive deviation widened thereafter. In the light of the above observations, the usefulness of credit-to-GDP ratio as a policy guide in the Indian context is debatable for the fundamental reason that the credit-to-GDP gap did not show any worthwhile variability up until 2002-03 with the standard deviation of the gap during 1950-51 to 2002-03 which was estimated at just 3.0 (this finding was corroborated by the Financial Stability Report (June 2011) of Reserve Bank of India) though there emerged a significant positive gap since 2003-04 as the standard deviation almost doubled during this five year period. In other words, from the historical perspective, the credit-to-GDP gap was too small to be of any value in the Indian context for policy purposes. How to rationalise and explain the inapplicability of the BCBS guidance as posited above? The observed lack of variability in the credit-to-GDP gap in the Indian context until recently is not hard to explain. Informal credit market has to Emerging Market Economies: An Exploration: 1-30 dominated the Indian scene over the decades, though the formal credit market has expanded its reach in the meantime. Illustratively, analysis of share of rural household debt by source as revealed by All India Debt and Investment Surveys, is of relevance in this regard. The share of non-institutional agencies, consisting of money lenders, traders, landlords and friends and relatives, in the rural household debt remained more or less stagnant at over 30 per cent during the last 30 years. Approximately 40 per cent of the population across India has bank accounts. Furthermore, the credit market in India largely remained underdeveloped due to the so-called financial repression reflecting fiscal dominance. The preemption of banking sector resources to fund persistently large fiscal deficits was once as high as 63.5 per cent of net demand and time liabilities of all the banks, though this has waned because of financial sector reforms in the recent decade. Thus, in the historical sense, the credit market in India remained dormant, reflecting the absence of structural drivers of credit, thereby resulting in credit-to-GDP gap being too small to be of any use from the policy perspective. Thus in the ex post sense, BCBS guidance is not applicable to India. Besides, it is further argued below that it is not applicable going forward into the future either, in the ex ante sense. Essentially, the BCBS buffer guide implicitly assumes that the long-term trend of the credit-to-GDP ratio is a reliable proxy for the optimal/equilibrium credit required for an economy and any positive/ negative deviation denotes excess/deficit credit growth. This is presumably valid in the case of the advanced economies operating generally at equilibrium with full employment (potential growth), and mature and integrated financial markets. Points on the long-term trend line typically signify equilibrium credit market compatible with full employment/potential employment. However, the BCBS capital buffer guide of deviation of credit-to-GDP ratio from its long-term trend is not suitable for EMEs. Fundamentally, rise in credit-to-GDP ratio may be unrelated to any signs of over-leverage in the credit market. The long-term past trend does not represent optimal/equilibrium credit requirements of these economies. Factors which predominantly determine the credit-to-GDP ratio in the EMEs in the future include various structural drivers viz., structural shift from services to manufacturing (Subbarao, 2011), financial deepening from a low base, rising efficiency of goods markets, rising efficiency of credit markets, and policy initiatives to improve flow of credit to sectors like the agriculture, small scale units and infrastructure (Mohan, 20061). Most of the EMEs are refocusing on positioning manufacturing sector as drivers of growth, going forward. Illustratively, India’s New Manufacturing Policy aims to grow manufacturing about 3 per cent faster than GDP so that its contribution to GDP can increase from 16 per cent to 25 per cent in the next 15 years. Typically, credit intensity of manufacturing is higher per unit of GDP. Further, in EMEs, various segments of the real sector continue to be outside the purview of the formal credit market, funded basically through informal sources of credit, as mentioned earlier. As the process of financial deepening gathers pace and manifests in the form of these segments seeking formal credit, there is a switch in sources of credit from informal to formal. Typically, EMEs are supply-constrained economies. As the supply constrains ease over time, goods markets tend to become more and more efficient. Rising factor mobility leading to enhanced allocative efficiency is one such manifestation. These improvements in goods markets generally get reflected in structural shifts in supply elasticities, resulting in increasing demand for credit to finance rising production. Over time, credit markets in EMEs become more efficient in intermediating funds between the users and providers of credit, facilitating easier fund mobility at lower transaction costs. In other words, as the cost of intermediation drops, credit off-take rises. Illustratively, the intermediation cost (defined as the spread between cost of deposits and return on loan assets) for the scheduled commercial banks in India consistently fell from 6.24 per cent in 1991-92 to 3.59 per cent in 1999-2000 and further to 3.31 per cent in 2009-10. Moreover, there have been conscious policy initiatives in EMEs to augment credit flow to certain identified sectors including agriculture, small scale units and infrastructure. Illustratively, the Government of India, as part of its strategy to boost agriculture production, announced a package to double the flow of institutional credit to agriculture within three years starting 2004-05. The agricultural credit in fact doubled in 2-year period, as against the stipulated period of three years. Furthermore, for EMEs including India, credit demand is also expected to go up due to investment needs of infrastructure and the demand for upscaling financial inclusion. For example, in India credit requirements for the next five years for infrastructure development are estimated at US $ 1 trillion. Additionally, subtle behavioural changes underway, typical of a fast developing economy, push up credit and these include, inter alia, rising consumption financed through debt as consumers become wealthy, i.e., behavioural changes arising out of wealth effect. Reflecting these structural determinants, the long-term trend itself would be bodily shifting upwards over time, thereby rendering it less and less useful as a secular benchmark. Consequently, any deviation (positive/negative) from the frequently shifting yardstick loses its theoretical underpinning. Positive deviation per se does not necessarily signify over-leverage nor does negative deviation per se necessarily denote under leverage. Therefore, credit (including bank and non-bank) growth in EMEs, thus, fundamentally will embody both structural and cyclical components. While it is necessary to address cyclical components through countercyclical capital buffer, structural components, on the other hand, should not be impacted by such buffer. In practice, it is almost impossible to identify and differentiate structural and cyclical components of credit growth. It could, however, be argued that corroborative evidence from other variables could be sought to decipher two components of credit growth. However, this exercise is fraught with the potential risk of adversely impacting the structural component of credit growth. Thus, interpretation of secular movement of creditto Emerging Market Economies: An Exploration: 1-30 to-GDP ratio vis-à-vis its long-term trend will be ambiguous in the context of EMEs. To cap this discussion, it would be apt to quote the following extracts: The BCBS framework uses the metric “Credit to GDP ratio” and its upward deviation from the long term trend to signal the need to build up countercyclical capital buffer. This metric is not suitable for Indian economy and other EMEs, as was also pointed out in the Financial Stability Report (FSR) of June 2011, due to structural changes taking place in the economy on account of high growth rate and financial inclusion etc. (Sinha, 2011). In a structurally transforming economy with rapid upward mobility, credit demand will expand faster than GDP for several reasons. First, India will shift increasingly from services to manufactures whose credit intensity is higher per unit of GDP. Second, we need to at least double our investment in infrastructure which will place enormous demands on credit. Finally, financial inclusion, which both the Government and the Reserve Bank are driving, will bring millions of low income households into the formal financial system with almost all of them needing credit. What all this means is that we are going to have to impose higher capital requirements on banks as per Basel III at a time when credit demand is going to expand rapidly. The concern is that this will raise the cost of credit and hence militate against growth (Subbarao, 2011). Hence, there is a need for an alternative buffer guide, which can unambiguously mirror the macro-financial environment, especially the leverage conditions, in which banks in EMEs operate. In fact, BCBS proposal (Principle 2) acknowledges the fact that the credit-to-GDP ratio does not need to play a dominant role in the information used by authorities to make and explain buffer decisions. Furthermore, supervisors in each jurisdiction are free to emphasise any other variables and qualitative information that make sense to them for purposes of assessing the sustainability of credit growth and the level of system-wide risk, as well as in making and explaining buffer decisions. Then, what could be the alternative buffer guide? What does the literature offer in this regard? The next section reviews the relevant literature. 4. Literature Review Ever since the BCBS has come out with the countercyclical buffer guidance in December 2010, there have been studies empirically verifying its usefulness for a variety of jurisdictions. Repullo and Saurina (2011), based on the select advanced economies, argued that credit-to-GDP ratio as the common reference point for application of the buffer would tend to reduce capital requirements when GDP growth is high and increase them when GDP growth is low, so it may end up exacerbating the inherent pro-cyclicality of risk-sensitive bank capital regulation. Instead, they recommended deviation of credit growth with respect to the long-term average as a common reference point for capital buffer operations. This author feels that while the drawbacks of the BCBS guidance for capital buffers acerbating procyclicality are properly diagnosed, the alternative proposed, namely credit growth gap is not suitable for EMEs in the sense that historical long-term trend for credit growth (backward looking) is not an equilibrium credit growth as explained earlier in Section 2. Hence, it does not reflect the realities of the EMEs. Geršl and Seidler (2012) criticised the BCBS guidance for the way the excess credit growth is defined. positive credit-to-GDP gap from the longterm trend derived by the HP filter. They contended that the backward looking long-term trend is not an appropriate benchmark for assessing the excess credit growth, reason being that HP-filter based long-term trend does not reflect macroeconomic fundamentals of converging economies and hence does not account for the needed catch-up in and convergence of credit growth of EMEs vis-àvis credit growth of advanced economies. Instead, they proposed estimating equilibrium credit elasticities of the advanced EU countries (involving a suitable econometric model) and applying these credit elasticities to Central and East European (CEE) countries to arrive at credible (excess) credit growth and equilibrium credit level. This paper recognises the fact that credit growth in EMEs (what is called convergence countries) needs to catch up with that of the advanced economies and any methodology for assessing excessive credit growth of these economies should need to factor in this imperative. However, this author feels that the paper’s criticism of the HP filter that it does reflect the economic fundamentals is not fair as the problem is not with the filter per se but with the indicator on which the filter is applied. The HP filter is one method which statistically extracts the long-term trend for a time-series. If the equilibrium credit levels of the advanced economies were to be the benchmark for assessing excessive credit growth of the EMEs, the straight forward option is to compare the actual credit-to-GDP growth of EMEs against the long-term trend of creditto-GDP growth of advanced economies. More importantly, the paper ignores the significance of funding sources for financing identified equilibrium credit levels. It is that attaining estimated equilibrium credit levels financed through unstable source of funding is not sustainable and hence has a built in systemic vulnerability for the banks. Mathias Drehmann et al. (2011) analysed the behaviour of a wide range of possible indicator variables around episodes of systemic banking crises, drawing on the empirical evidence from more than 40 crises in 36 countries for setting the level of the countercyclical regulatory capital buffer requirements for banks. The authors found evidence that the gap between the ratio of credit to GDP and its long-term backward-looking trend performs best as an indicator for the accumulation of capital, because this variable captures the build-up of system-wide vulnerabilities that typically leads to banking crises. Other to Emerging Market Economies: An Exploration: 1-30 indicators, such as credit spreads, are better at indicating the release phase, as they are contemporaneous signals of banking sector distress that can precede a credit crunch. Against the above-mentioned backdrop, the alternative common reference point proposed in this paper does not rely on any estimated equilibrium credit levels. It accommodates expansion in the asst-side of the balance sheet of the banks and thereby the catch up in credit growth of EMEs, to the extent that it is financed by stable sources from the liability-side of the balance sheet. Hence, it ensures sustainable catch up in credit growth. Furthermore, the alternative capital buffer guide is not procyclical and moves in synchronisation with the business cycles as measured by real GDP growth. The next section attempts to present the methodology underlying the alternative common reference point. However, before the methodology is presented, it needs to be noted that this paper does not claim to have found the single indicator capable of guiding buffer decisions (both build up and release) in EMEs. All indicators provide false signals. Thus, no fully rule-based mechanism is perfect. Some degree of judgment, both for the build-up and particularly for the release phase, would be inevitable when setting countercyclical capital buffers in practice. 5. Methodology for an Alternative Capital Buffer Guide for EMEs2 As mentioned at the outset, the countercyclical capital buffer should dampen liquidity cycles. The optimal candidate for buffer guide should, therefore, reflect the evolving macro-financial environment associated with credit growth. In particular, the buffer guide should be able to capture system-wide vulnerabilities and risks associated with what is perceived to be excessive credit growth. Banking business model in EMEs, including India, is basically retail in nature meaning that the principal source of funding for the banking business is retail deposit base. This business model has endured over the years, inter alia, because for banks in the EMEs, including India, dependence on credit risk sensitive purchased sources of funding is limited. This historical dependence of these banks on deposits as a source of funding has inherently imparted an element of built-in stability to the banking sector in these jurisdictions. Any departure from the reliance on deposits to fund credit growth, on a sustained basis, would signal build-up of system-wide risk. Theoretically it could, however, be argued that the process of financial development in the EMEs may involve a trend increase in securitisation, which would bias the proposed measure of excessive credit growth. In other words, in the context of pick up in securitisation, departure from the reliance on deposits may not necessarily signify over-leverage and thereby rising systemic risk, but may instead underscore financial development in the form of diversification of funding sources for banks involving investors, like pension funds, insurance companies, etc. In this context, it is essential to note the following. The securitisation markets in EMEs, at present, are either small in size and nascent in stage, or non-existent. In fact, the regulatory framework in EMEs and India in particular underlying securitisation strives to promote orderly development of the securitisation market. Moreover, EMEs do have a huge untapped potential rural retail deposit base to harness going forward. It will, therefore, be a long time before securitisation markets acquire critical mass in EMEs. Secondly, any large deviation from the reliance on deposits as a primary source of funding on a sustained basis-be it due to securitization - does highlight potential build-up of systemic risk as market liquidity conditions tend to acquire greater influence on the stability of the banking sector. The episode of the failure of Northern Rock in the UK illustrates the case in point. Furthermore, as would be evident later in this section, the methodology for developing an alternative capital buffer guide has built-in cushion to tolerate and accommodate prudent credit growth funded by non-deposit sources. Viewed from the above stand point, it is not the credit growth per se but the pattern of funding of the credit growth that needs to be the criteria for the conduct of capital buffer operations on the theoretical premise that expansion of the asset-side of the balance sheet of banks (credit growth) supported by increasingly unstable growth of liability-side signifies worsening system-wide risk. Thus, capital buffers need to be built to protect banks from vulnerabilities arising out of excessive credit growth, not vis-à-vis GDP but vis-à-vis retail deposits. Against this back drop, rising Credit Deposit Ratio (CDR) over time could denote increasing system-wide leverage and hence deterioration in macrofinancial environment in which banks in EMEs operate. Ideally, the alternative buffer should capture the combined movement of absolute and incremental CD ratios.3 Absolute CD ratio measures leverage on stock basis, while incremental CD ratio measures leverage on flow basis. However, given the fact that from a purely arithmetic stand point, one is derived from the other, reflecting thereby a strong correlation between the two, use of both absolute CD ratio and the incremental CD ratio deserves a detailed justification. Incremental CD ratio, in isolation, provides only a partial view of the extent of leverage by the banks. Nor does the absolute CD ratio alone give a complete picture on banks’ leverage. The following illustration would underscore this point. Let absolute CD ratio at t1 be 45 per cent (45/100) and at t2 57 per cent (68/120). The incremental CD ratio during t1and t2 works out to 115 per cent. However, ICD of 115 per cent does not necessarily signify over-leverage, as it is on the back of a lower absolute CD ratio at t1. Higher credit growth during t1and t2 might be supported by an overhang of relatively large deposits at t1. To illustrate further, let absolute CD ratio at t1 be 75 per cent (75/100) and at t2 be 77.5 per cent (93/120). The incremental CD ratio during t1and t2 works out to be 90 per cent. However, ICD of 95 per cent does not necessarily signify under-leverage, as it is on the back of a relatively larger overhang of credit manifest in higher absolute CD ratio at t1. Thus, the holistic view of banks’ leverage is provided to Emerging Market Economies: An Exploration: 1-30 only when the incremental CD ratio is seen in conjunction with the absolute CD ratio. In fact, historically, credit aggregates - both absolute and incremental - have been amongst the host of variables, forming an integral part of macroeconomic and prudential policy formulation in India. In this context in particular, the following extract from a speech by Smt Usha Thorat (2010) assumes relevance: Absolute and incremental credit aggregates (including credit deposit ratio) are amongst the host of variables, forming an integral part of macro-economic and prudential policy formulation. In the Indian context, an incremental credit-deposit ratio of more than 100 per cent, when the system itself has a high overall absolute credit deposit ratio (say beyond 70 per cent) is taken as a sign of over-leverage. A prudential focus on credit deposit ratio encourages the banks in India to raise deposits for funding credit Absolute and credit aggregates (including credit deposit ratio) Absolute and incremental incremental credit aggregates (including credit deposit ratio) are Absolute and incremental credit aggregates (including credit deposit ratio) are are Absolute and incremental credit aggregates (including credit deposit ratio) are amongst the host ofof variables, forming an integral ofof macro-economic amongst theof host variables, forming anfunds. integral part macro-economic and amongst thehost host of variables, forming integral partofpart of macro-economic and and the variables, forming ananintegral part macro-economic and flow amongst and minimises the use of purchased the prudential policy formulation. InIn the context, an credit-deposit prudential policy formulation. theIndian Indian context, anincremental incremental credit-deposit prudential policy formulation. the Indian context, incremental credit-deposit prudential policy formulation. InInthe Indian context, ananincremental credit-deposit ratio ofofthan more than per cent, the itself aaoverall high overall absolute ratio more than perwhen cent, when thesystem system itself has high overall absolute ratioof ofmore more than100 per cent, whenwhen thesystem system itselfhas hasa ahas high overall absolute ratio per cent, the itself high absolute credit deposit ratio (say beyond per cent) ismethodology as ofoffor over-leverage. credit deposit ratio (say beyond peris cent) istaken taken asaof asign sign over-leverage. credit deposit ratio(say (saybeyond beyond per cent) istaken taken asa asign sign of over-leverage. credit deposit ratio 7070per cent) as over-leverage. AA AA Against this theoretical underpinning, the constructing prudential on deposit ratio encourages the inin India toto raise prudential focus on credit credit deposit ratio encourages the banks banks India raise prudential focusfocus oncredit credit deposit ratioencourages encourages thebanks banks India toraise raise prudential focus on deposit ratio the ininIndia to alternative buffer guide iscredit enumerated deposits for credit flow and minimises the use ofofpurchased funds. deposits forfunding funding credit flow andbelow: minimises the use purchased funds. deposits funding flowand andminimises minimises theuse use purchased funds. deposits forforfunding credit flow the ofof purchased funds.

Step Against 1: Calculate time-series data the on both absolute CD ratio the (cd )theand Against this theoretical underpinning, the for Against this theoretical underpinning, the methodology methodology for constructing constructing Against thistheoretical theoretical underpinning, themethodology methodology constructing the jthe this underpinning, forforconstructing incremental CD (icdj) ratios alternative buffer guide isisenumerated below: alternative buffer guide enumerated below: alternative buffer guide isenumerated enumerated below: alternative buffer guide ismoving below: Step 2: Compute maxima of both cdj and icdj ratios. Smoothening (moving maxima) of the ratios is suggested to account for the possible non-linear Step 1:1:Calculate time-series data on both absolute CD ratio andincremental incremental Step Calculate time-series data onabsolute both absolute CD(cd ratio (cd j)incremental j)and )(cd and incremental Step1: 1:Calculate Calculate time-series dataon on both absolute CDratio ratio (cd Step time-series data both CD j) jand impactCD of the ratios on the conduct of capital buffer operations. Moving maxima CD (icd CD ratios j)j)ratios CD(icd (icd )(icd ratios j) jratios is recommended for smoothening to reflect the imperative of conservatism in Step 2:2:Compute moving maxima ofofcd both cd Smoothening (moving Step Compute moving maxima both cd and icdj jratios. ratios. Smoothening (moving j jand Step2: 2:Compute Compute moving maxima both cd icd ratios. Smoothening (moving Step maxima ofofboth icd Smoothening (moving j and j icd j and j ratios. regulation. Empirically, itmoving isisisfound that three-year offered better fit maxima) ofof the suggested toto account for the possible non-linear impact ofof the maxima) the ratios ratios suggested account for thewindow possible non-linear impact the in maxima) the ratios issuggested suggested account the possible non-linear impact the maxima) ofofthe ratios is totoaccount forforthe possible non-linear impact ofofthe on the ofofthe capital buffer operations. Moving maxima isis recommended ratios on the conduct conduct capital buffer operations. Moving maxima recommended for terms of theratios compatibility of alternative guide with other relevant indicators ratios onthe the conduct capital buffer operations. Moving maxima recommended ratios on conduct ofofcapital buffer operations. Moving maxima is isrecommended forfor for smoothening totoreflect the ofofconservatism ininregulation. Empirically, it smoothening reflect theimperative imperative conservatism regulation. Empirically, itisisfound found smoothening reflect theimperative imperative conservatism regulation. Empirically, found totoreflect the ofofconservatism ininregulation. Empirically, it itis isfound of real smoothening sector and asset prices (for details see below). that window offered better ininterms ofof the ofof the thatthree-year three-year window offered better fitterms terms the compatibility compatibility the alternative alternative thatthree-year three-year window offered better fitininfit ofthe the compatibility the alternative that window offered better fit terms of compatibility ofofthe alternative guide with other relevant indicators ofof real and asset prices (for details see guide withrelevant other relevant indicators realsector sector andprices asset prices (for details seebelow). below). Step 3: Construct aindicators moving of Composite CD Ratio (MAXCCDR) guide with other relevant indicators ofreal real sector and asset prices (for details seebelow). below). guide with other ofMaxima sector and asset (for details see by combining moving maxima of both cd and icd ratios with weights. Step 3:3: Construct aa moving Maxima ofof Composite CD (MAXCCDR) Step Construct moving Maxima Composite CD Ratio Ratio (MAXCCDR) by j CD Step3: 3:Construct Construct moving Maxima Composite CDRatio Ratio (MAXCCDR) Step a amoving Maxima ofj ofComposite (MAXCCDR) byby by icd with combining moving maxima ofofcd both cd and icdj jratios ratios withweights. weights. combining moving maxima both cd j jand icd withweights. weights. combining moving maxima both cd icd with combining moving maxima ofofboth j and j ratios j and j ratios

Notationally, Notationally, Notationally, Notationally, Notationally,

cdforj nj n=for let moving maxima cd be nn=n=3,6,9,---let3-year 3-year moving maxima cd ratio beMax for 3,6,9,---cd jbe jratio 3-year moving maxima ofcdcd ratio beMax =3,6,9,---3,6,9,---cd3cdjj nj for letlet3-year moving maxima ofof ratio jof j of Max let 3-year moving maxima cd ratio be for = 3,6,9,---Max j n 3 j n n 3nj3njnn

j for be nnn===3,6,9,--let moving maxima icd for 3,6,9,--ratio be Max let3-year 3-year moving maxima of icd icd jbe for jratio beMax for n=for =3,6,9,--3,6,9,--3-year moving maxima oficd icd icdicdj jicd ratio letlet3-year moving maxima ofof jn j ratio jof let 3-year moving maxima icd ratio be 3,6,9,--Max Max j n jjn n n33 n 3nj3nj n then

j0<w<0 +cd ))++(1-w) )],)],0<w<0 then ( (Max ( (Max thenMAXCCDR MAXCCDR 0<w<0 icd j j== cd icd [(w) ([(w) thenMAXCCDR MAXCCDR ) (1-w) +(1-w) ( Max cdcdj j )+ icdicdj j)], [(w) ( [(w) then ((1-w) MAXCCDR =j =[(w) ],)],0<w<0 j j0<w<0 j (1-w) j= Max Max Max Max Max j jjn n n33 n 3nj3njnn jjn n n33 n 3nj3njnn where w isisweights the (for ofofweights please see where wweights theweights weights (fordetermination determination weights please seeendnote endnote where wis isthe the (fordetermination determination weights please seeendnote endnote where w (for ofofweights please see 4)4) 4)4)

where w is the weights (for determination of weights please see endnote The theoretical justification for the use ofofmoving maxima and the The4) theoretical justification for the use moving maxima and thedetermination determination Thetheoretical theoretical justification the use moving maxima andthe the determination The justification forforthe use ofof moving maxima and determination ofof ofof weights asas follows. InIn the on timeseries data analysis, moving weights follows. the literature literature on financial financial timeseries data analysis, moving weights areasare asare follows. the literature financial timeseries dataanalysis, analysis, moving weights are follows. InInthe literature ononfinancial timeseries data moving

The theoretical justification for the use of moving maxima and the maxima isisemployed especially in the Multivariate Extreme Theory, which maxima employed especially inarea the area Multivariate Extreme Value Theory, which determination of weights are as follows. Inofofthe literature on Value financial timeseries maxima employed especially the of Multivariate Extreme Value Theory, which is isis maxima is is employed especially ininthe area ofarea Multivariate Extreme Value Theory, which is concerned the distribution ofof extremes ofof multiple random variables. concerned with the joint joint distribution extremes multiple random variables. concerned withwith the joint joint distribution extremes multiple random variables. concerned with the distribution ofof extremes ofof multiple random variables.

data analysis, moving maxima is employed especially in the area of Multivariate Extreme Value Theory, which is concerned with the joint distribution of extremes of multiple random variables. Multivariate Extreme Value Theory has applications in banking and finance also wherein extreme events dependent across different assets occur in clusters. Estimation of such joint distributions generally involves modeling extreme multivariate events based on Moving Maxima (MM) process and a multivariate extension known as Multivariate Maxima of Moving Maxima (M4) process (Stuart et al 1991; Chamu Morales, 2005). Furthermore, literature also supported the use of maxima for calibration of macroprudential policy. Davis et al. (2010), inter alia, estimated the impact of capital adequacy and liquidity on probability of financial crisis. In particular, they generated the required maxima for capital adjustment and liquidity adjustment and both together for protecting against banking crisis anywhere in the world. Drawing from the work of Davis et al. use of maxima is recommended in this paper for calibrating countercyclical capital buffers. Regarding determination of weights, assigning equal weights to various components of a composite indicator is generally an accepted practice in the literature relating to financial/banking regulation. For instance, BCBS (2011) assigned equal weights to 5 indicators for identification of Globally Systemically Important Banks (G-SIBs). Drawing from this standard practice, cdj and icdj were assigned equal weights while computing MAXCCDR4. Thus computed MAXCCRD could be the alternative buffer guide in the EMEs context. Statistical details underlying the computation of MAXCCDR is provided in Appendix 1. The theoretical interpretation of MAXCCDR is unambiguous, unlike the credit-deposit ratio. Deviations from the long-term trend of the MAXCCDR (TMAXCCDR) do reflect underlying changes in the macro-financial environment i.e. leveraged funding conditions in which banks in EMEs operate. The long-term trend is computed using Hodrick Prescott (HP) filter with lambda (λ) = 400000 with the help of E-Views, as suggested by the BCBS. In the literature, there were various detrending methods and Rochelle and Ralf (2011) provided an excellent summary of these methods and the implications thereof. Any practical attempt to detrend a series would need to involve some consideration of issues such as the deterministic or stochastic nature of the trend and the most appropriate filter to use. The MAXCCDR, based on an augmented Dickey-Fuller test, has a unit root and thus has a stochastic trend. This implies that the HP filter, which is able to remove a unit root, is more appropriate to apply to MAXCCDR since deterministic detrending methods will generate spurious cycles. Thus, contrary to the stand taken by Geršl and Seidler (2012), this paper supports HP filter methodology for detrending. Actual MAXCCDR being higher than the computed trend MAXCCDR (positive gap) depicts the situation wherein the three-year window moving maxima is higher than the trend MAXCCDR signifying thereby worsening to Emerging Market Economies: An Exploration: 1-30 macro-financial environment of over-leverage5. On the other hand, negative gap denotes under-leverage. Next section attempts to present the historical performance of MAXCCDR in the context of Indian banks using India relevant data.

The computed MAXCCDR and TMAXCCDR on the basis of the abovemethodology, involving data since 1950-51 is presented in Figure 2. Analysis of Figure 2 reveals the following: During the last 60 years, banking sector in India operated, by and large, below but closer to the long-term trend. This means that leverage position of Indian banks on a secular basis remained in balance. Further, there have been 5 episodes of over leverage (positive gap) during mid-1950s, mid-1960s, mid-1970s, mid-1990s and since 2003-04. Out of these 5 episodes, there were 3 episode of positive gap (during mid-1950s, mid-1960s and since 2003-04) exceeding the long-term trend by a substantial margin. There have been 3 episodes of under-leverage (negative gap) during late 1950s, late 1970s to early 1990s and during 1998-99 to 2003-04. Till 196566, the amplitude of alternate swings of phases of under leverage and overleverage, as measured by standard deviation at 27.4, was indeed high. Since 1965-66 to 1995-96, there has been a discernible moderation in amplitude of credit cycles as standard deviation during this period was estimated at 8. Since 1995-96, alternate phases of pronounced under and over-leverage are again apparent as standard deviation rose to 23. There was a brief period of over-leverage during mid-1990s followed by a prolonged period of underleverage till 2003-04, which was replaced by a phase of over-leverage since then. Meaning, credit cycles have relatively become more pronounced in India since mid-1990s. From the above exposition, it could be inferred that the proposed alternative buffer guide is able to track phases of over/under leverage in the banking sector in India. Can, thus, MAXCCDR be taken as lead indicator for capital buffer operations? The BCBS guidance cautioned about the potential possibility of misleading signals emanating from the buffer guide and hence recommended looking for evidence as to whether the inferences from the buffer guide are consistent with those of other variables such as real GDP growth, asset prices, etc. So, do the phases of over/under leverage in the banking sector, as identified by the MAXCCDR, correspond/coincide with the phases of over/under-leveraging in the Indian real sector? In other words, how does the MAXCCDR map and track the performance of the economic activity, as measured by real GDP vis-à-vis credit to GDP ratio?

Std. Error t-Statistic

R-squared

Mean dependent var

Adjusted R-squared

S.D. dependent var

S.E. of regression

Sum squared resid

Log likelihood

Durbin-Watson stat

Prob. 0.0711 0.0213 5.169231 11.48727

TCGDP400000

TCGDP400000

Figure 7: MAXCCDR Sensex returns. Figure 7: MAXCCDR gapgap andand the the Sensex returns. MAXCCDRGAP

(continued)

to Emerging Market Economies: An Exploration: 1-30 Year

Bank Credit (C)

Aggregate Deposits (D)

CD

ICD

MaxCD (3-year moving window)

MaxICD (3-year moving window)

Weighted MAXCD

Weighted MAXICD

Source: Calculated by the authors based on the data from Handbook of Statistics on Indian Economy 2010-11.

Appendix 2 Statistical Significance of Correlation Coefficients An attempt was made to examine the statistical significance of the correlation coefficients estimated in this paper. Under H0: ρ = 0 and H1: ρ ≠ 0, the statistical significance was tested using the following formula for t-distribution with degrees of freedom = N-2 and N ≥ 6.

1) The results of the statistical significance of the correlation coefficient between MAXCCDR gap and GDP are presented below. Statistical Significance of Correlation Coefficient between MAXCCDR gap and GDP MAXCCDR gap and GDP

Calculated t-value

1990-91 to 2009-10

1990-91 to 2003-04

2004-05 to 2009-10

Degrees of Freedom (df)