Description
The report highlights the prevention of financial crisis which plays a central role in maintaining financial stability.
Early Warning System for Financial Crises
Nicholas Cheang
Research and Statistics Department, Monetary Authority of Macao
Abstract The prevention of financial crisis plays a central role in maintaining financial stability. One of the preventive actions is to apply an Early Warning System (EWS) to provide signals that reflect the likelihood that an economy would face financial crises over a given time horizon. Based on a review of existing in-sample and out-of-sample studies applied to some Asian economies, it is observed that the signaling-based EWS model could provide useful indication for financial crises. This paper also shows that in terms of categorical variety and series length, statistical information in Macao is considered largely adequate for constructing a signaling-based EWS model.
- 61 -
Early Warning System for Financial Crises
1. Introduction The Asian Financial Crisis in 1997-98 has raised concerns for worldwide financial contagion. Financial liberalisation in the absence of a well-developed regulatory framework had located some emerging-market economies in a precarious position. The weakness of their financial systems in the context of unexpected shock had completely exposed their vulnerability with consequences of slumping currencies, devalued asset prices and precipitous rise in private debts. Financial crisis is certainly not exclusive for developing economies. different degrees of severity. In the past three
decades, more than 20 countries have experienced banking or currency crises in The up-to-the-minute financial tsunami originated from the United States could be the most stringent one since the Great Depression. It has brought about a global recession and an impairment of the functions of the international financial system. Indeed, financial crisis is no stranger on the global stage while many still claim that crises are unforecastable. Some of the major international organisations, such as the International Monetary Fund (IMF), have encouraged individual economies to strengthen monetary and financial stability, regional cooperation, information exchange, and to improve statistical transparency and timeliness of data in order to avoid if possible or at least alert the formation of financial turmoil. Under this context, a number of projects have been initiated to construct Early Warning System (EWS) models, which apply statistical methods to predict the likelihood that an economy would face financial crisis over a given time horizon. The framework of those models is primarily composed of some selected economic and financial indicators that are likely to provide an indication of a vulnerable position at the macro or aggregate level. The original targets for EWS are emerging-market economies, but the ongoing financial tsunami, stemmed from the advanced economies, is likely to extend the scope for usage of these models.
- 62 -
Early Warning System for Financial Crises
This paper attempts to introduce the EWS with the ultimate aim to study the feasibility for Macao to construct a EWS for financial crisis, primarily in the data aspect. The next section outlines a representative EWS model and its construction. The effectiveness of the model is supported by various in-sample and out-of-sample test results conducted by the IMF. The availability of relevant statistical data is the pre-requisite for the construction of EWS. The third section examines the data selection and availability of Macao for constructing a EWS in accordance with the representative model presented in Section 2 and the characteristics of its financial sector. The final section concludes.
2. Construction of EWS models In constructing EWS models, two primary approaches are mostly applied, i.e. probit/logit approach and signaling approach. The former is usually applied on a multivariate model, which allows testing of statistical significance of explanatory variables. This type of models requires large samples and can only accommodate a limited number of explanatory variables to avoid multicolinearity. On the other hand, the signaling approach is frequently applied in univariate models, which involve monitoring a set of high-frequency leading indicators. It is noted that those selected indicators would behave differently prior to a financial crisis until they reach their individual threshold values which are historically associated with the onset of a crisis. Univariate models work better with small samples and impose no restriction on the number of explanatory variables. The remainder of this paper will mainly focus on the signaling-based EWS models. When building a univariate EWS model, the first crucial step is to identify historical crisis episodes and then examine the symptoms of those crises. For instance, the earlier models of balance of payments (BOP) problems, inspired by Latin America’s currency crises in the late 1970s, imply that fiscal deficit leads to a persistent loss of international reserves and ignites a currency crash (Krugman 1979). As for banking crises, Calomiris and Gorton (1991) point out that recessions precede crises, which
- 63 -
Early Warning System for Financial Crises
are most likely to occur when the recession follows a period of drastic credit growth. In this vein, depositors might try to reassess the risk of bank debts when recession unfolds, resulting in depositors withdrawing large amounts from banking institutions. On the front of liquidity crises, Montiel and Reinhart (1999) argue that sudden reversals in capital flows play a significant role. For instance, it is observed that during the Asian Financial Crisis, sudden halt of capital inflows, partly prompted by fluctuations in interest rates in industrialised countries, was rather unexpected when those capital inflows were only short-term in nature. Similarly, McKinnon and Pill (1996) examine the role of capital flows in an economy under the context of an unregulated banking sector with problems in deposit insurance and moral hazard. They suggest that capital inflows under such circumstances could lead to overlending cycles resulting in over-consumption and exaggerate current account deficits. In addition, booms in the stock and real estate markets are considered as some of the symptoms, resulting in pervasive over-exposure of financial institutions to the unstable asset markets.
The next step in model construction is selecting leading indicators with reference to the crisis symptoms highlighted in the first step. Based on the empirical regularities observed in a sample of 20 countries from 1970 to 1995, Kaminsky (2000) construct a EWS model,1 or a model for economic distress indicators, to study the onset of 76 currency crises and 26 banking crises within the sample period. Some symptoms are grouped and considered signs of frailty. economic vulnerability. sign of frailty (Table 1). Kaminsky argues that it is necessary to compute the number of those negative signs of frailty in order to quantify the state of A set of indicators, which could be used to reflect the possibility of currency and banking crisis eruptions, is singled out according to each
1
The EWS model developed by Kaminsky in 2000 is a modified version, originated from a model designed by Kaminsky, Lizondo and Reinhart (1997). - 64 -
Early Warning System for Financial Crises
Table 1: Symptoms and Leading Indicators Symptoms
Overborrowing cycles Bank runs Monetary policy Current account problems
Leading Indicators
M2 multiplier Domestic credit-to-GDP ratio Domestic and external liberalisation. Bank deposits “Excess” M1 Balance Exports Imports Terms of trade Real exchange rate Reserves M2-to-reserves ratio Real interest rate differential World real interest rate Foreign Debt Capital Flight Short-term Foreign Debt Output Domestic real interest rate Lending-to-deposit interest rate ratio Stock prices financial
Capital account problems
Growth slowdown
Source: Kaminsky (2000).
Although there could be a number of possible indicators showing deterioration signs, it is not necessary that one or more of these worsening indicators would stir the economy into a crisis. Hence, the technical question remained was how to define the critical level at which a fluctuation in an indicator would make a crisis almost unavoidable. To set the thresholds for those leading indicators, the economist argues that it is important to understand the distribution of each indicator. A threshold generally divides its distribution into a region considered “normal” and another region considered “abnormal”. The indicator is said to be giving a warning signal when it falls into “abnormal” territory. In addition, the allowed interval of time between the signal and the crisis is set at 24 months. In this vein, a warning signal could be true if a crisis follows within 24 months or false if no crisis emerges within the designated timeframe. Proposing to set the optimal threshold for an abnormal region, Kaminsky (2000) minimises the noise-to-signal ratio (NSR), which is defined as the ratio of the probability of an
- 65 -
Early Warning System for Financial Crises
indicator signaling in tranquil times, to the probability of the indicator signaling in crisis times. A composite leading index is then constructed on the assumption that the greater the number of leading indicators signaling a financial crisis, the higher the probability that such a crisis is likely to occur. The composite index is a weighted average of the signals by each of the selected indicators with larger weights given to those better performing indicators, i.e. those with lower “minimum NSRs”. Finally, the composite index would issue a warning signal when the observed outcome in a particular period exceeds the critical level. Results indicate that around 70% of the indicators signaled deterioration in macroeconomic fundamentals in the two years prior the Asian Financial Crisis for the cases of Thailand and the Philippines. The probabilities of currency crises for these two economies increased from 20% in 1995 to at least 80% in 1997, signaling increased financial fragilities. Meanwhile, Malaysia, which had a number of common features with Thailand, also had 70% of the indicators showing signs of distress at the onset of the crisis. Similarly, Zhuang and Dowling (2002) apply a univariate EWS model to study the economic conditions of the most-affected Asian countries prior to the Asian Financial Crisis. However, their objective is slightly different, by providing empirical evidence on economic and financial fragility in those economies with a view to discriminating between two hypotheses: “weak fundamentals” and “investors’ panic”. They apply a modified EWS model using monthly data from 1970 to 1995 for in-sample2 testing and 1996-1997 for out-of-sample3 predictions, targeting six economies: Indonesia, South Korea, Malaysia, the Philippines, Thailand and Singapore. The first five were at the centre of the crisis, i.e. those worst-hit by the crisis whereas Singapore was the exception. The model is then applied to test whether there were warning signals in each of these economies.
2
In-sample testing measures how well a model fits the crisis in a particular sample. When the allowed interval of time between the signal and the crisis is set at 24 months, data applied for the in-sample testing would be 24 months dated back to the onset of the Asian Financial Crisis. 3 In out-of-sample testing, the predictions of an existing model are compared to a new set of observations, not belonging to the estimation sample. - 66 -
Early Warning System for Financial Crises
Zhuang and Dowling (2002)’s methodology generally follow that of Kaminsky (2000), except for including some more variables, such as the ratio of foreign liabilities to foreign assets of the banking sector and the real exchange rate of the Japanese yen against the US dollar. They show that the overall composite leading index did offer persistent warning signals prior to the Asian Financial crisis in all of the five most-affected economies. The results also lend support to the hypothesis that weaknesses in economic and financial fundamentals in those economies largely triggered the crisis. On the contrary, Singapore had no signals of crisis as shown in the overall composite leading index, which largely mirrored the reality. Berg et al. (2004) systematically track various EWS models and look into detail about the performance of those models in practice. As a useful tool in monitoring vulnerabilities, a EWS must hold up in “real time” after the model has been formulated. Therefore, they attempt to emphasise the distinction between in-sample and out-of-sample predictions and focus on the actual forecasts made since 1999. The goodness-of-fit of the popular Kaminsky model 4 is tested over the entire in-sample as well as out-of-sample periods in order to obtain a more systematic assessment. It is shown that the goodness-of-fit for in-sample is reasonably stable, while the model also presents satisfactory goodness-of-fit for the out-of-sample prediction. The indication of the Kaminsky model is therefore viewed as highly informative. However, it is worth noting that those results should be interpreted cautiously as the number of crises actually observed are limited, translating into a small effective sample size. Minor changes in sample could induce a relatively large difference in the goodness-of-fit indicators.
3. Feasibility of Constructing a EWS for Macao So far formal study of EWS for Macao has not been published, though the half-yearly Monetary and Financial Stability Review published by the Monetary Authority of
4
They actually study the original model by Kaminsky, Lizondo and Reinhart (1997). - 67 -
Early Warning System for Financial Crises
Macao (AMCM) since 2005 has, in practice, played a role in monitoring systemic risk. The ongoing financial tsunami has highlighted the impact of contagion and the necessity of policy foresight. For a highly open financial system as Macao,5 it is of great importance to strengthen the capacity of surveillance over the system by means of a proper EWS in order to alert policy makers and financial institutions about the probability of financial crisis. In this section, we attempt to study the data availability for constructing a EWS for Macao based on the Kaminsky model. In view of the features of Macao’s financial sector, some bank-specific indicators based on the existing framework of the IMF’s Financial Soundness Indicators (FSIs), which are currently adopted by the AMCM’s Monetary and Financial Stability Review, will also be introduced for a thorough analysis. 3.1 Data availability Based on the EWS framework by Kaminsky (2000), the first procedure of selecting useful indicators applied in EWS is to identify economic symptoms which usually come to surface prior to financial crisis. Past experiences in some of the crisis-hit economies show that both banking and currency crises are linked to overborrowing cycles. In some cases, the substantial credit growth could be fueled by financial liberalisation and elimination of capital and financial account restrictions, which, however, are not quantifiable. The mirroring indicators include M2 multiplier and ratio of domestic credit-to-GDP. In the case of Macao, the AMCM has compiled monetary and financial statistics (MFS) since 1984. Some of the statistical time series can be measured up to two decades plus. For the broad measure of money supply, M2 in monthly frequency is available up to June 1984. Monetary base (i.e. denominator of M2 multiplier) has
5
At present, banks remain dominant in Macao’s financial system, accounting for about 95% of total assets in the local financial sector. The total number of banks in Macao stands at 28; of which 16 are branches of overseas banks and 12, including a postal savings bank, are locally incorporated. Most locally incorporated banks are subsidiaries of foreign enterprises. At the end of 2008, the share of international assets by Bank for International Settlements definition, i.e. external assets plus local assets in foreign currencies, in total banking assets in Macao was over 80%. - 68 -
Early Warning System for Financial Crises
been compiled in monthly frequency since July 1989, while data of domestic credit are available in monthly frequency up to June 1984. Meanwhile, annual data of Macao’s GDP complied by Macao Statistics and Census Service (DSEC) have been available for more than two decades whereas the quarterly data have only been available since 1998. Banking and currency crises can be preceded by bank runs. As depositors withdraw massively their deposits, the likelihood of bank default increases. The phenomenon has a destabilising effect, and the mirroring indicator is bank deposits, 6 which correspondingly exhibit dramatic negative movements during bank panic. statistics of bank deposits are part of the MFS. different types of deposit under the resident and non-resident sectors. In Macao, The data in The said statistics are classified into
monthly frequency have been available since June 1984. To deflate deposits, the composite consumer price index (CPI) compiled by the DSEC could be applied. The CPI in monthly frequency has been available for more than two decades. Inappropriate monetary policy stance would pose potential harm to a financial system. Krugman (1979) argues that loose monetary policy could fuel a currency crisis while devaluation could worsen the health of a banking sector and trigger a banking crisis. The mirroring indicator is “excess” M1 balance.7 Regarding Macao, data of narrow money supply, M1, have been compiled in monthly frequency since June 1984. Besides, local inflation rate could be generated from Macao’s composite CPI. Current account problems are considered as one of the symptoms for financial crisis. Those problems could be reflected in the performances of external trade, terms of trade and real exchange rate. Real exchange rate overvaluation and a weak external sector are potential factors for currency crisis. A loss of competitiveness and weak external markets could lead to recession, business failure, and deterioration in loan quality.
6 7
Bank deposits are deflated by consumer prices. M1 deflated by consumer prices less an estimated demand for money. The demand for real balances is determined by real GDP, domestic consumer price inflation, and a time trend. Domestic inflation is used in lieu of nominal interest rates; the time trend, which can enter log-linearly, linearly, or exponentially, is motivated by its role as a proxy for financial innovation and/or currency substitution. - 69 -
Early Warning System for Financial Crises
In Macao, statistics of external trade and terms of trade are compiled by the DSEC. Time series of more than two decades for monthly export and import statistics are available. Terms of trade are only compiled in quarterly frequency. Meanwhile, the real exchange rate index could be derived from the nominal exchange rate index, which has been available in monthly frequency since June 1984, adjusted for consumer prices. Capital account problems become more severe in the context of enlarging foreign debt and increasing capital flight, which raise concern for debt unsustainability. Vulnerability of a country to external shocks is more likely to increase if foreign debt is dominantly concentrated in short maturities. The selected indicators of this area include foreign exchange reserves, ratio of M2 to foreign exchange reserves, real interest rate differential, 8 world real interest rate, 9 foreign debt, 10 short-term foreign debt ratio11 and capital flight.12 In Macao, the AMCM has compiled the statistics of M2, foreign exchange reserves and selected interest rates since June 1984. The data series are available in monthly frequency. Regarding the possible statistics applied in foreign debt, short-term foreign debt and capital flight, it is suggested to adopt the statistics of liabilities/deposits of Macao residents to/with reporting banks of the Bank for International Settlements (BIS). The said statistics are published in BIS Quarterly Review – International Banking and Financial Market Developments. Severe slowdown in economic growth or recession as well as the burst of asset price bubbles could precede financial crises. Kaminsky (2000) argues that high real interest rates could be a sign of liquidity crunch, which leads to an economic slowdown and banking fragility. The mirroring indicators included output,13 real domestic interest rate,14 lending/deposit interest rate ratio, and stock prices.15
8
Interest rates in the domestic economy are compared with interest rates in the United States if the domestic economy pegs its currency to the US dollar. Real rates are deposit rates deflated by consumer prices. 9 World real interest rate refers to US deposit rate deflated by consumer prices. 10 Foreign debt refers to liabilities of residents to BIS reporting banks. 11 Short-term foreign debt refers to liabilities of residents to BIS reporting banks with maturities up to one year divided by total liabilities of residents to BIS reporting banks. 12 Deposits of residents with BIS reporting banks divided by foreign exchange reserves. 13 Measure of output used for most countries is industrial production index or an index of output of primary commodities is used if industrial production index is not available. 14 Real domestic interest rate is deflated deposit rate using consumer prices. 15 Stock prices are used to indicate economic growth. It is expected that stock prices exhibit negative sign with symptom of growth slowdown. - 70 -
Early Warning System for Financial Crises
Table 2: Data Availability of Selected Indicators for EWS in Macao
Indicator M2 multiplier Ratio of domestic credit to GDP Bank deposit Data Category ? M2 ? Monetary base ? Domestic credit ? GDP ? Deposits ? Composite CPI (including rent) ? M1 ? Composite CPI (including rent) ? Exports of goods ? Imports of goods ? Terms of trade ? Real effective exchange rate index ? Foreign exchange reserves ? M2 ? foreign exchange reserves ? Domestic interest rate for deposit ? Composite CPI (including rent) ? US Interest rates ? US deposit rate ? Liabilities of residents to BIS reporting banks ? Deposits of residents with BIS reporting banks ? Foreign exchange reserves ? Industrial production index ? Domestic interest rate for deposit ? Composite CPI (including rent) ? Domestic interest rate for deposit 17 ? Prime lending rate ? Hong Kong Hang Seng Index
Data Source
AMCM AMCM AMCM DSEC AMCM DSEC AMCM DSEC DSEC DSEC DSEC AMCM AMCM AMCM AMCM DSEC IMF IMF BIS BIS AMCM DSEC AMCM DSEC AMCM HKMA HKEX
Data Periodicity
Monthly Monthly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Quarterly Quarterly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly
Data Length
Over 20 years Since July 1989 Over 20 years Since 1998 Over 20 years Since 1998 Over 20 years Since 1998 Over 20 years Over 20 years Since 1993 Since 1997 Over 20 years Over 20 years Over 20 years Since 1998 Over 20 years Over 20 years Since 199516 Since 199516 Over 20 years Since 2005 Over 20 years Since 1998 Over 20 years Over 20 years Over 20 years
“Excess” M1 balance Exports Imports Terms of trade Real exchange rate Foreign exchange reserves Ratio of M2 to foreign exchange reserves
Real interest rate differential World real interest rate Foreign debt
Capital flight
Output Real domestic interest rate Ratio of lending rate to deposit rate Stock prices
16
BIS official website provides time series for external loans and deposits of reporting banks vis-à-vis individual countries. 17 In view of the inadequate data length of the prime lending rate in Macao, Hong Kong’s prime lending rate published by the HKMA could be applied as a proxy for the Macao rate. Traditionally, the lending rate quoted by the local banking sector would not derivate much from its Hong Kong counterpart. - 71 -
Early Warning System for Financial Crises
In the case of Macao, the industrial production index compiled by the DSEC has been available in quarterly frequency since the third quarter of 2005 while annual indices could be dated back to the year 2000. Various interest rates have been available in monthly frequency since June 1984. Since Macao does not have a physical stock market, Hong Kong’s Hang Seng Index could serve as an alternative in view of the popularity of Macao residents investing in securities listed on the Hong Kong Stock Exchange (HKEX).18 3.2 Bank-specific indicators Just passing the 10th anniversary of the Asian Financial Crisis, the international financial system has been facing another severe crisis. Although the centre of the current crisis is different from that of the Asian Financial Crisis, some argue that the similarities between the two financial episodes are investor panic in the face of uncertainty over the security and valuation of assets as well as liquidity positions of the banking system. Meanwhile, Macao’s simplistic financial sector is dominated by banks, which hold the key to the stability of the entire system. Primarily based on the FSIs of the IMF, we single out some indicators in relation to asset quality and liquidity positions of banks, which are not applied in the Kaminsky model but currently monitored in the half-yearly Monetary and Financial Stability Review of the AMCM. In reflecting the asset quality of the banking sector, the ratio of non-performing loans (NPL) to total loans19 and ratio of NPL net of provisions to capital20 are on the indicators list. The former is intended to identify problems with asset quality in the loan portfolio while the latter is an indicator of the capacity of bank capital to
Macao residents’ portfolio investment is mainly placed in securities issued by Hong Kong entities and Mainland entities listed on the Hong Kong Stock Exchange according to the AMCM’s Coordinated Portfolio Investment Survey (CPIS). 19 Definition adopted from the IMF’s Compilation Guide for Financial Soundness Indicators (2006). The ratio is calculated by taking the value of NPLs as the numerator and the total value of the loan portfolio (including NPLs, and before the deduction of specific loan loss provisions) as the denominator. 20 The ratio is calculated by taking the value of NPLs less the value of specific loan provisions as the numerator, and capital, which is measured as capital and reserves, as the denominator. - 72 18
Early Warning System for Financial Crises
withstand loan losses.
In addition, the delinquency ratios of property-related loans
and credit card receivables can reflect the quality of major assets in Macao banks’ loan portfolios. The ratio of property-related loans to total private sector credit can also indicate the degree of bank exposure to a particular asset market. Reflecting the external positions of the banking sector, ratio of external assets to total assets and ratio of foreign currency assets to foreign currency liabilities could be applied in a EWS. The former is an indicator reflecting the external positions of banks in terms of assets while the latter highlights the risk of currency mismatch in view of international exposure. Table 3: Availability of Bank-specific Indicators for EWS in Macao
Indicator Ratio of NPL to total loans ? NPL ? Total loans ? NPL ? NPL provisions ? Capital ? Reserves ? Credit card receivables overdue for more than three months ? Total credit card receivables ? Property-related loan repayments overdue for more than three months ? Property-related loans outstanding ? Credit extended to individuals for house purchase ? Total private sector credit ? Residential mortgage and commercial real estate loans ? Foreign assets ? Total assets ? Assets and liabilities denominated in foreign currencies ? Liquid assets ? Total assets ? Loans ? Deposits Data Category Data Source AMCM AMCM AMCM AMCM Data Periodicity Monthly Monthly Monthly Monthly Data Length Since 1990 Over 20 years Since 1990 Over 20 years
Ratio of NPL net of provisions to capital
Delinquency ratio of credit card receivables
AMCM
Quarterly
Since 2006
Delinquency ratio of property-related loans
AMCM
Quarterly
Since 2009
Ratio of mortgage loans to total private sector credit
AMCM
Quarterly
Over 20 years
AMCM
Quarterly
Since 2009
Ratio of external assets to total assets Ratio of foreign currency assets to foreign currency liabilities Ratio of liquid assets to total assets Ratio of loans to deposits
AMCM
Monthly
Over 20 years
AMCM
Monthly
Over 20 years
AMCM AMCM
Monthly Monthly
Over 20 years Over 20 years
- 73 -
Early Warning System for Financial Crises
When considering liquidity position of the banking sector, the ratio of liquid assets to total assets21 which indicates the liquidity available to meet expected and unexpected demand for cash is widely used. Concurrently, the ratio of loans to deposits,22 which depicts the percentage of deposit funding that is tired-up in loan portfolios, is also an important indicator measuring the vulnerability of the banking system. Thanks to Macao’s participation in the General Data Dissemination System, the development of Macao’s statistical system has been satisfactorily by international standards in terms of data quality and variety.23 Still, it is likely that data periodicity and time series length for some of the aforementioned indicators in Macao are inadequate to produce statistically stable results when constructing a EWS. Those indicators may include output on the selected list as well as delinquency ratios of credit card receivables and property-related loans on the bank-specific list.
4. Conclusion By applying statistical methods to predict the likelihood of an economy facing financial crises over a given time horizon, the EWS could serve as an important instrument to alert policy makers and financial institutions about potential turmoil. This paper primarily studies the leading indicators advocated in a representative signaling-based EWS model. Statistical results show that the model produce satisfactorily in-sample forecast for some Asian economies in the context of the Asian Financial Crisis, while various tracking studies support that the model remains informative in producing out-of-sample forecast. This paper also studies the feasibility of constructing a similar EWS model for Macao that is obviously conducive to maintaining its monetary and financial stability. The initial yet crucial step is to investigate whether Macao’s data availability is adequate
21
The ratio is calculated by using the core measure or broad measure of liquid assets as the numerator and total assets as the denominator. 22 For a ratio over one, it implies that banks depend on borrowing to finance lending. 23 Details are available in Leong (2008). - 74 -
Early Warning System for Financial Crises
to be applied in a similar EWS model. It is observed that Macao can meet the requirements for most categories of indicators applied in the Kaminsky model and bank-specific indicators that can be included in Macao’s EWS model in view of the characteristics of its financial sector, as inspired by the IMF’s FSIs. Moreover, it is believed that the majority of these statistics have data series lengthy enough to produce statistically stable results. Future studies, therefore, may proceed to the next level involving using the existing data to conduct econometric exercises in the construction of a signaling-based EWS model for Macao.
- 75 -
References Berg, A. and C. Pattillo (2000), “The Challenges of Predicting Economic Crises,” IMF Economic Issues, No. 22. Berg, A., E. Borensztein and C. Pattillo (2004), “Assessing Early Warning Systems: How Have They Worked in Practice?,” IMF Working Paper, WP/04/52. Berg, A., E. Borensztein, G. M. Milesi-Ferretti and C. Pattillo (1999), “Anticipating Balance of Payments Crises: The Role of Early Warning Systems,” IMF Occasional Paper, No. 186. Calomiris, C. W. and G. Gorton (1991), “The Origins of Banking Panics: Models, Facts, and Bank Regulation,” in G. Hubbard (ed.), Financial Markets and Financial Crises, University of Chicago Press, 109-173. Corsetti, G., P. Pesenti and N. Roubini (1998), “What Caused the Asian Currency and Financial Crisis?,” NBER Working Paper Series, No. 6833. Ee, K. H. and K. R. Xiong (2008), “Asia: A Perspective on the Subprime Crisis,” IMF Finance and Development, Vol. 45, No. 2. International Monetary Fund (2006), Financial Soundness Indicators Compilation Guide. Kaminsky, G.., S. Lizondo and G. M. Reinhart (1997), “Leading Indicators of Currency Crises,” IMF Working Paper, WP/97/79. Kaminsky, G. (2000), Currency and Banking Crises: The Early Warnings of Distress, George Washington University. (http://www.gwu.edu/~clai/working_papers/Kaminsky_Graciela_07-00.pdf)
- 76 -
Krugman, P. (1979), “A Model of Balance-of-Payments Crises,” Journal of Money, Credit, and Banking, Vol. 11, 311-325. Leong, M. N. (2008), “International Data Standards and Their Application to Macao,” Macao Monetary Research Bulletin, Issue No. 7, April, 63-79. McKinnon, R. I. and H. Pill (1996), “Credible Liberalizations and International Capital Flows: the Overborrowing Syndrome,” Financial Deregulation and Integration in East Asia, NBER-EASE, Vol. 5, 7-50. Montiel, P. and C. Reinhart (1999), The Dynamics of Capital Movements to Emerging Economies during the 1990s, University of Maryland. (http://www.puaf.umd.edu/faculty/reinhart/txt0308.pdf) Zhuang, J. and J. M. Dowling (2002), “Causes of the 1997 Asian Financial Crisis: What Can an Early Warning System Model Tell Us?,” ERD Policy Brief Series, No. 7.
- 77 -
doc_863937541.pdf
The report highlights the prevention of financial crisis which plays a central role in maintaining financial stability.
Early Warning System for Financial Crises
Nicholas Cheang
Research and Statistics Department, Monetary Authority of Macao
Abstract The prevention of financial crisis plays a central role in maintaining financial stability. One of the preventive actions is to apply an Early Warning System (EWS) to provide signals that reflect the likelihood that an economy would face financial crises over a given time horizon. Based on a review of existing in-sample and out-of-sample studies applied to some Asian economies, it is observed that the signaling-based EWS model could provide useful indication for financial crises. This paper also shows that in terms of categorical variety and series length, statistical information in Macao is considered largely adequate for constructing a signaling-based EWS model.
- 61 -
Early Warning System for Financial Crises
1. Introduction The Asian Financial Crisis in 1997-98 has raised concerns for worldwide financial contagion. Financial liberalisation in the absence of a well-developed regulatory framework had located some emerging-market economies in a precarious position. The weakness of their financial systems in the context of unexpected shock had completely exposed their vulnerability with consequences of slumping currencies, devalued asset prices and precipitous rise in private debts. Financial crisis is certainly not exclusive for developing economies. different degrees of severity. In the past three
decades, more than 20 countries have experienced banking or currency crises in The up-to-the-minute financial tsunami originated from the United States could be the most stringent one since the Great Depression. It has brought about a global recession and an impairment of the functions of the international financial system. Indeed, financial crisis is no stranger on the global stage while many still claim that crises are unforecastable. Some of the major international organisations, such as the International Monetary Fund (IMF), have encouraged individual economies to strengthen monetary and financial stability, regional cooperation, information exchange, and to improve statistical transparency and timeliness of data in order to avoid if possible or at least alert the formation of financial turmoil. Under this context, a number of projects have been initiated to construct Early Warning System (EWS) models, which apply statistical methods to predict the likelihood that an economy would face financial crisis over a given time horizon. The framework of those models is primarily composed of some selected economic and financial indicators that are likely to provide an indication of a vulnerable position at the macro or aggregate level. The original targets for EWS are emerging-market economies, but the ongoing financial tsunami, stemmed from the advanced economies, is likely to extend the scope for usage of these models.
- 62 -
Early Warning System for Financial Crises
This paper attempts to introduce the EWS with the ultimate aim to study the feasibility for Macao to construct a EWS for financial crisis, primarily in the data aspect. The next section outlines a representative EWS model and its construction. The effectiveness of the model is supported by various in-sample and out-of-sample test results conducted by the IMF. The availability of relevant statistical data is the pre-requisite for the construction of EWS. The third section examines the data selection and availability of Macao for constructing a EWS in accordance with the representative model presented in Section 2 and the characteristics of its financial sector. The final section concludes.
2. Construction of EWS models In constructing EWS models, two primary approaches are mostly applied, i.e. probit/logit approach and signaling approach. The former is usually applied on a multivariate model, which allows testing of statistical significance of explanatory variables. This type of models requires large samples and can only accommodate a limited number of explanatory variables to avoid multicolinearity. On the other hand, the signaling approach is frequently applied in univariate models, which involve monitoring a set of high-frequency leading indicators. It is noted that those selected indicators would behave differently prior to a financial crisis until they reach their individual threshold values which are historically associated with the onset of a crisis. Univariate models work better with small samples and impose no restriction on the number of explanatory variables. The remainder of this paper will mainly focus on the signaling-based EWS models. When building a univariate EWS model, the first crucial step is to identify historical crisis episodes and then examine the symptoms of those crises. For instance, the earlier models of balance of payments (BOP) problems, inspired by Latin America’s currency crises in the late 1970s, imply that fiscal deficit leads to a persistent loss of international reserves and ignites a currency crash (Krugman 1979). As for banking crises, Calomiris and Gorton (1991) point out that recessions precede crises, which
- 63 -
Early Warning System for Financial Crises
are most likely to occur when the recession follows a period of drastic credit growth. In this vein, depositors might try to reassess the risk of bank debts when recession unfolds, resulting in depositors withdrawing large amounts from banking institutions. On the front of liquidity crises, Montiel and Reinhart (1999) argue that sudden reversals in capital flows play a significant role. For instance, it is observed that during the Asian Financial Crisis, sudden halt of capital inflows, partly prompted by fluctuations in interest rates in industrialised countries, was rather unexpected when those capital inflows were only short-term in nature. Similarly, McKinnon and Pill (1996) examine the role of capital flows in an economy under the context of an unregulated banking sector with problems in deposit insurance and moral hazard. They suggest that capital inflows under such circumstances could lead to overlending cycles resulting in over-consumption and exaggerate current account deficits. In addition, booms in the stock and real estate markets are considered as some of the symptoms, resulting in pervasive over-exposure of financial institutions to the unstable asset markets.
The next step in model construction is selecting leading indicators with reference to the crisis symptoms highlighted in the first step. Based on the empirical regularities observed in a sample of 20 countries from 1970 to 1995, Kaminsky (2000) construct a EWS model,1 or a model for economic distress indicators, to study the onset of 76 currency crises and 26 banking crises within the sample period. Some symptoms are grouped and considered signs of frailty. economic vulnerability. sign of frailty (Table 1). Kaminsky argues that it is necessary to compute the number of those negative signs of frailty in order to quantify the state of A set of indicators, which could be used to reflect the possibility of currency and banking crisis eruptions, is singled out according to each
1
The EWS model developed by Kaminsky in 2000 is a modified version, originated from a model designed by Kaminsky, Lizondo and Reinhart (1997). - 64 -
Early Warning System for Financial Crises
Table 1: Symptoms and Leading Indicators Symptoms
Overborrowing cycles Bank runs Monetary policy Current account problems
Leading Indicators
M2 multiplier Domestic credit-to-GDP ratio Domestic and external liberalisation. Bank deposits “Excess” M1 Balance Exports Imports Terms of trade Real exchange rate Reserves M2-to-reserves ratio Real interest rate differential World real interest rate Foreign Debt Capital Flight Short-term Foreign Debt Output Domestic real interest rate Lending-to-deposit interest rate ratio Stock prices financial
Capital account problems
Growth slowdown
Source: Kaminsky (2000).
Although there could be a number of possible indicators showing deterioration signs, it is not necessary that one or more of these worsening indicators would stir the economy into a crisis. Hence, the technical question remained was how to define the critical level at which a fluctuation in an indicator would make a crisis almost unavoidable. To set the thresholds for those leading indicators, the economist argues that it is important to understand the distribution of each indicator. A threshold generally divides its distribution into a region considered “normal” and another region considered “abnormal”. The indicator is said to be giving a warning signal when it falls into “abnormal” territory. In addition, the allowed interval of time between the signal and the crisis is set at 24 months. In this vein, a warning signal could be true if a crisis follows within 24 months or false if no crisis emerges within the designated timeframe. Proposing to set the optimal threshold for an abnormal region, Kaminsky (2000) minimises the noise-to-signal ratio (NSR), which is defined as the ratio of the probability of an
- 65 -
Early Warning System for Financial Crises
indicator signaling in tranquil times, to the probability of the indicator signaling in crisis times. A composite leading index is then constructed on the assumption that the greater the number of leading indicators signaling a financial crisis, the higher the probability that such a crisis is likely to occur. The composite index is a weighted average of the signals by each of the selected indicators with larger weights given to those better performing indicators, i.e. those with lower “minimum NSRs”. Finally, the composite index would issue a warning signal when the observed outcome in a particular period exceeds the critical level. Results indicate that around 70% of the indicators signaled deterioration in macroeconomic fundamentals in the two years prior the Asian Financial Crisis for the cases of Thailand and the Philippines. The probabilities of currency crises for these two economies increased from 20% in 1995 to at least 80% in 1997, signaling increased financial fragilities. Meanwhile, Malaysia, which had a number of common features with Thailand, also had 70% of the indicators showing signs of distress at the onset of the crisis. Similarly, Zhuang and Dowling (2002) apply a univariate EWS model to study the economic conditions of the most-affected Asian countries prior to the Asian Financial Crisis. However, their objective is slightly different, by providing empirical evidence on economic and financial fragility in those economies with a view to discriminating between two hypotheses: “weak fundamentals” and “investors’ panic”. They apply a modified EWS model using monthly data from 1970 to 1995 for in-sample2 testing and 1996-1997 for out-of-sample3 predictions, targeting six economies: Indonesia, South Korea, Malaysia, the Philippines, Thailand and Singapore. The first five were at the centre of the crisis, i.e. those worst-hit by the crisis whereas Singapore was the exception. The model is then applied to test whether there were warning signals in each of these economies.
2
In-sample testing measures how well a model fits the crisis in a particular sample. When the allowed interval of time between the signal and the crisis is set at 24 months, data applied for the in-sample testing would be 24 months dated back to the onset of the Asian Financial Crisis. 3 In out-of-sample testing, the predictions of an existing model are compared to a new set of observations, not belonging to the estimation sample. - 66 -
Early Warning System for Financial Crises
Zhuang and Dowling (2002)’s methodology generally follow that of Kaminsky (2000), except for including some more variables, such as the ratio of foreign liabilities to foreign assets of the banking sector and the real exchange rate of the Japanese yen against the US dollar. They show that the overall composite leading index did offer persistent warning signals prior to the Asian Financial crisis in all of the five most-affected economies. The results also lend support to the hypothesis that weaknesses in economic and financial fundamentals in those economies largely triggered the crisis. On the contrary, Singapore had no signals of crisis as shown in the overall composite leading index, which largely mirrored the reality. Berg et al. (2004) systematically track various EWS models and look into detail about the performance of those models in practice. As a useful tool in monitoring vulnerabilities, a EWS must hold up in “real time” after the model has been formulated. Therefore, they attempt to emphasise the distinction between in-sample and out-of-sample predictions and focus on the actual forecasts made since 1999. The goodness-of-fit of the popular Kaminsky model 4 is tested over the entire in-sample as well as out-of-sample periods in order to obtain a more systematic assessment. It is shown that the goodness-of-fit for in-sample is reasonably stable, while the model also presents satisfactory goodness-of-fit for the out-of-sample prediction. The indication of the Kaminsky model is therefore viewed as highly informative. However, it is worth noting that those results should be interpreted cautiously as the number of crises actually observed are limited, translating into a small effective sample size. Minor changes in sample could induce a relatively large difference in the goodness-of-fit indicators.
3. Feasibility of Constructing a EWS for Macao So far formal study of EWS for Macao has not been published, though the half-yearly Monetary and Financial Stability Review published by the Monetary Authority of
4
They actually study the original model by Kaminsky, Lizondo and Reinhart (1997). - 67 -
Early Warning System for Financial Crises
Macao (AMCM) since 2005 has, in practice, played a role in monitoring systemic risk. The ongoing financial tsunami has highlighted the impact of contagion and the necessity of policy foresight. For a highly open financial system as Macao,5 it is of great importance to strengthen the capacity of surveillance over the system by means of a proper EWS in order to alert policy makers and financial institutions about the probability of financial crisis. In this section, we attempt to study the data availability for constructing a EWS for Macao based on the Kaminsky model. In view of the features of Macao’s financial sector, some bank-specific indicators based on the existing framework of the IMF’s Financial Soundness Indicators (FSIs), which are currently adopted by the AMCM’s Monetary and Financial Stability Review, will also be introduced for a thorough analysis. 3.1 Data availability Based on the EWS framework by Kaminsky (2000), the first procedure of selecting useful indicators applied in EWS is to identify economic symptoms which usually come to surface prior to financial crisis. Past experiences in some of the crisis-hit economies show that both banking and currency crises are linked to overborrowing cycles. In some cases, the substantial credit growth could be fueled by financial liberalisation and elimination of capital and financial account restrictions, which, however, are not quantifiable. The mirroring indicators include M2 multiplier and ratio of domestic credit-to-GDP. In the case of Macao, the AMCM has compiled monetary and financial statistics (MFS) since 1984. Some of the statistical time series can be measured up to two decades plus. For the broad measure of money supply, M2 in monthly frequency is available up to June 1984. Monetary base (i.e. denominator of M2 multiplier) has
5
At present, banks remain dominant in Macao’s financial system, accounting for about 95% of total assets in the local financial sector. The total number of banks in Macao stands at 28; of which 16 are branches of overseas banks and 12, including a postal savings bank, are locally incorporated. Most locally incorporated banks are subsidiaries of foreign enterprises. At the end of 2008, the share of international assets by Bank for International Settlements definition, i.e. external assets plus local assets in foreign currencies, in total banking assets in Macao was over 80%. - 68 -
Early Warning System for Financial Crises
been compiled in monthly frequency since July 1989, while data of domestic credit are available in monthly frequency up to June 1984. Meanwhile, annual data of Macao’s GDP complied by Macao Statistics and Census Service (DSEC) have been available for more than two decades whereas the quarterly data have only been available since 1998. Banking and currency crises can be preceded by bank runs. As depositors withdraw massively their deposits, the likelihood of bank default increases. The phenomenon has a destabilising effect, and the mirroring indicator is bank deposits, 6 which correspondingly exhibit dramatic negative movements during bank panic. statistics of bank deposits are part of the MFS. different types of deposit under the resident and non-resident sectors. In Macao, The data in The said statistics are classified into
monthly frequency have been available since June 1984. To deflate deposits, the composite consumer price index (CPI) compiled by the DSEC could be applied. The CPI in monthly frequency has been available for more than two decades. Inappropriate monetary policy stance would pose potential harm to a financial system. Krugman (1979) argues that loose monetary policy could fuel a currency crisis while devaluation could worsen the health of a banking sector and trigger a banking crisis. The mirroring indicator is “excess” M1 balance.7 Regarding Macao, data of narrow money supply, M1, have been compiled in monthly frequency since June 1984. Besides, local inflation rate could be generated from Macao’s composite CPI. Current account problems are considered as one of the symptoms for financial crisis. Those problems could be reflected in the performances of external trade, terms of trade and real exchange rate. Real exchange rate overvaluation and a weak external sector are potential factors for currency crisis. A loss of competitiveness and weak external markets could lead to recession, business failure, and deterioration in loan quality.
6 7
Bank deposits are deflated by consumer prices. M1 deflated by consumer prices less an estimated demand for money. The demand for real balances is determined by real GDP, domestic consumer price inflation, and a time trend. Domestic inflation is used in lieu of nominal interest rates; the time trend, which can enter log-linearly, linearly, or exponentially, is motivated by its role as a proxy for financial innovation and/or currency substitution. - 69 -
Early Warning System for Financial Crises
In Macao, statistics of external trade and terms of trade are compiled by the DSEC. Time series of more than two decades for monthly export and import statistics are available. Terms of trade are only compiled in quarterly frequency. Meanwhile, the real exchange rate index could be derived from the nominal exchange rate index, which has been available in monthly frequency since June 1984, adjusted for consumer prices. Capital account problems become more severe in the context of enlarging foreign debt and increasing capital flight, which raise concern for debt unsustainability. Vulnerability of a country to external shocks is more likely to increase if foreign debt is dominantly concentrated in short maturities. The selected indicators of this area include foreign exchange reserves, ratio of M2 to foreign exchange reserves, real interest rate differential, 8 world real interest rate, 9 foreign debt, 10 short-term foreign debt ratio11 and capital flight.12 In Macao, the AMCM has compiled the statistics of M2, foreign exchange reserves and selected interest rates since June 1984. The data series are available in monthly frequency. Regarding the possible statistics applied in foreign debt, short-term foreign debt and capital flight, it is suggested to adopt the statistics of liabilities/deposits of Macao residents to/with reporting banks of the Bank for International Settlements (BIS). The said statistics are published in BIS Quarterly Review – International Banking and Financial Market Developments. Severe slowdown in economic growth or recession as well as the burst of asset price bubbles could precede financial crises. Kaminsky (2000) argues that high real interest rates could be a sign of liquidity crunch, which leads to an economic slowdown and banking fragility. The mirroring indicators included output,13 real domestic interest rate,14 lending/deposit interest rate ratio, and stock prices.15
8
Interest rates in the domestic economy are compared with interest rates in the United States if the domestic economy pegs its currency to the US dollar. Real rates are deposit rates deflated by consumer prices. 9 World real interest rate refers to US deposit rate deflated by consumer prices. 10 Foreign debt refers to liabilities of residents to BIS reporting banks. 11 Short-term foreign debt refers to liabilities of residents to BIS reporting banks with maturities up to one year divided by total liabilities of residents to BIS reporting banks. 12 Deposits of residents with BIS reporting banks divided by foreign exchange reserves. 13 Measure of output used for most countries is industrial production index or an index of output of primary commodities is used if industrial production index is not available. 14 Real domestic interest rate is deflated deposit rate using consumer prices. 15 Stock prices are used to indicate economic growth. It is expected that stock prices exhibit negative sign with symptom of growth slowdown. - 70 -
Early Warning System for Financial Crises
Table 2: Data Availability of Selected Indicators for EWS in Macao
Indicator M2 multiplier Ratio of domestic credit to GDP Bank deposit Data Category ? M2 ? Monetary base ? Domestic credit ? GDP ? Deposits ? Composite CPI (including rent) ? M1 ? Composite CPI (including rent) ? Exports of goods ? Imports of goods ? Terms of trade ? Real effective exchange rate index ? Foreign exchange reserves ? M2 ? foreign exchange reserves ? Domestic interest rate for deposit ? Composite CPI (including rent) ? US Interest rates ? US deposit rate ? Liabilities of residents to BIS reporting banks ? Deposits of residents with BIS reporting banks ? Foreign exchange reserves ? Industrial production index ? Domestic interest rate for deposit ? Composite CPI (including rent) ? Domestic interest rate for deposit 17 ? Prime lending rate ? Hong Kong Hang Seng Index
Data Source
AMCM AMCM AMCM DSEC AMCM DSEC AMCM DSEC DSEC DSEC DSEC AMCM AMCM AMCM AMCM DSEC IMF IMF BIS BIS AMCM DSEC AMCM DSEC AMCM HKMA HKEX
Data Periodicity
Monthly Monthly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Quarterly Quarterly Monthly Quarterly Monthly Monthly Monthly Monthly Monthly
Data Length
Over 20 years Since July 1989 Over 20 years Since 1998 Over 20 years Since 1998 Over 20 years Since 1998 Over 20 years Over 20 years Since 1993 Since 1997 Over 20 years Over 20 years Over 20 years Since 1998 Over 20 years Over 20 years Since 199516 Since 199516 Over 20 years Since 2005 Over 20 years Since 1998 Over 20 years Over 20 years Over 20 years
“Excess” M1 balance Exports Imports Terms of trade Real exchange rate Foreign exchange reserves Ratio of M2 to foreign exchange reserves
Real interest rate differential World real interest rate Foreign debt
Capital flight
Output Real domestic interest rate Ratio of lending rate to deposit rate Stock prices
16
BIS official website provides time series for external loans and deposits of reporting banks vis-à-vis individual countries. 17 In view of the inadequate data length of the prime lending rate in Macao, Hong Kong’s prime lending rate published by the HKMA could be applied as a proxy for the Macao rate. Traditionally, the lending rate quoted by the local banking sector would not derivate much from its Hong Kong counterpart. - 71 -
Early Warning System for Financial Crises
In the case of Macao, the industrial production index compiled by the DSEC has been available in quarterly frequency since the third quarter of 2005 while annual indices could be dated back to the year 2000. Various interest rates have been available in monthly frequency since June 1984. Since Macao does not have a physical stock market, Hong Kong’s Hang Seng Index could serve as an alternative in view of the popularity of Macao residents investing in securities listed on the Hong Kong Stock Exchange (HKEX).18 3.2 Bank-specific indicators Just passing the 10th anniversary of the Asian Financial Crisis, the international financial system has been facing another severe crisis. Although the centre of the current crisis is different from that of the Asian Financial Crisis, some argue that the similarities between the two financial episodes are investor panic in the face of uncertainty over the security and valuation of assets as well as liquidity positions of the banking system. Meanwhile, Macao’s simplistic financial sector is dominated by banks, which hold the key to the stability of the entire system. Primarily based on the FSIs of the IMF, we single out some indicators in relation to asset quality and liquidity positions of banks, which are not applied in the Kaminsky model but currently monitored in the half-yearly Monetary and Financial Stability Review of the AMCM. In reflecting the asset quality of the banking sector, the ratio of non-performing loans (NPL) to total loans19 and ratio of NPL net of provisions to capital20 are on the indicators list. The former is intended to identify problems with asset quality in the loan portfolio while the latter is an indicator of the capacity of bank capital to
Macao residents’ portfolio investment is mainly placed in securities issued by Hong Kong entities and Mainland entities listed on the Hong Kong Stock Exchange according to the AMCM’s Coordinated Portfolio Investment Survey (CPIS). 19 Definition adopted from the IMF’s Compilation Guide for Financial Soundness Indicators (2006). The ratio is calculated by taking the value of NPLs as the numerator and the total value of the loan portfolio (including NPLs, and before the deduction of specific loan loss provisions) as the denominator. 20 The ratio is calculated by taking the value of NPLs less the value of specific loan provisions as the numerator, and capital, which is measured as capital and reserves, as the denominator. - 72 18
Early Warning System for Financial Crises
withstand loan losses.
In addition, the delinquency ratios of property-related loans
and credit card receivables can reflect the quality of major assets in Macao banks’ loan portfolios. The ratio of property-related loans to total private sector credit can also indicate the degree of bank exposure to a particular asset market. Reflecting the external positions of the banking sector, ratio of external assets to total assets and ratio of foreign currency assets to foreign currency liabilities could be applied in a EWS. The former is an indicator reflecting the external positions of banks in terms of assets while the latter highlights the risk of currency mismatch in view of international exposure. Table 3: Availability of Bank-specific Indicators for EWS in Macao
Indicator Ratio of NPL to total loans ? NPL ? Total loans ? NPL ? NPL provisions ? Capital ? Reserves ? Credit card receivables overdue for more than three months ? Total credit card receivables ? Property-related loan repayments overdue for more than three months ? Property-related loans outstanding ? Credit extended to individuals for house purchase ? Total private sector credit ? Residential mortgage and commercial real estate loans ? Foreign assets ? Total assets ? Assets and liabilities denominated in foreign currencies ? Liquid assets ? Total assets ? Loans ? Deposits Data Category Data Source AMCM AMCM AMCM AMCM Data Periodicity Monthly Monthly Monthly Monthly Data Length Since 1990 Over 20 years Since 1990 Over 20 years
Ratio of NPL net of provisions to capital
Delinquency ratio of credit card receivables
AMCM
Quarterly
Since 2006
Delinquency ratio of property-related loans
AMCM
Quarterly
Since 2009
Ratio of mortgage loans to total private sector credit
AMCM
Quarterly
Over 20 years
AMCM
Quarterly
Since 2009
Ratio of external assets to total assets Ratio of foreign currency assets to foreign currency liabilities Ratio of liquid assets to total assets Ratio of loans to deposits
AMCM
Monthly
Over 20 years
AMCM
Monthly
Over 20 years
AMCM AMCM
Monthly Monthly
Over 20 years Over 20 years
- 73 -
Early Warning System for Financial Crises
When considering liquidity position of the banking sector, the ratio of liquid assets to total assets21 which indicates the liquidity available to meet expected and unexpected demand for cash is widely used. Concurrently, the ratio of loans to deposits,22 which depicts the percentage of deposit funding that is tired-up in loan portfolios, is also an important indicator measuring the vulnerability of the banking system. Thanks to Macao’s participation in the General Data Dissemination System, the development of Macao’s statistical system has been satisfactorily by international standards in terms of data quality and variety.23 Still, it is likely that data periodicity and time series length for some of the aforementioned indicators in Macao are inadequate to produce statistically stable results when constructing a EWS. Those indicators may include output on the selected list as well as delinquency ratios of credit card receivables and property-related loans on the bank-specific list.
4. Conclusion By applying statistical methods to predict the likelihood of an economy facing financial crises over a given time horizon, the EWS could serve as an important instrument to alert policy makers and financial institutions about potential turmoil. This paper primarily studies the leading indicators advocated in a representative signaling-based EWS model. Statistical results show that the model produce satisfactorily in-sample forecast for some Asian economies in the context of the Asian Financial Crisis, while various tracking studies support that the model remains informative in producing out-of-sample forecast. This paper also studies the feasibility of constructing a similar EWS model for Macao that is obviously conducive to maintaining its monetary and financial stability. The initial yet crucial step is to investigate whether Macao’s data availability is adequate
21
The ratio is calculated by using the core measure or broad measure of liquid assets as the numerator and total assets as the denominator. 22 For a ratio over one, it implies that banks depend on borrowing to finance lending. 23 Details are available in Leong (2008). - 74 -
Early Warning System for Financial Crises
to be applied in a similar EWS model. It is observed that Macao can meet the requirements for most categories of indicators applied in the Kaminsky model and bank-specific indicators that can be included in Macao’s EWS model in view of the characteristics of its financial sector, as inspired by the IMF’s FSIs. Moreover, it is believed that the majority of these statistics have data series lengthy enough to produce statistically stable results. Future studies, therefore, may proceed to the next level involving using the existing data to conduct econometric exercises in the construction of a signaling-based EWS model for Macao.
- 75 -
References Berg, A. and C. Pattillo (2000), “The Challenges of Predicting Economic Crises,” IMF Economic Issues, No. 22. Berg, A., E. Borensztein and C. Pattillo (2004), “Assessing Early Warning Systems: How Have They Worked in Practice?,” IMF Working Paper, WP/04/52. Berg, A., E. Borensztein, G. M. Milesi-Ferretti and C. Pattillo (1999), “Anticipating Balance of Payments Crises: The Role of Early Warning Systems,” IMF Occasional Paper, No. 186. Calomiris, C. W. and G. Gorton (1991), “The Origins of Banking Panics: Models, Facts, and Bank Regulation,” in G. Hubbard (ed.), Financial Markets and Financial Crises, University of Chicago Press, 109-173. Corsetti, G., P. Pesenti and N. Roubini (1998), “What Caused the Asian Currency and Financial Crisis?,” NBER Working Paper Series, No. 6833. Ee, K. H. and K. R. Xiong (2008), “Asia: A Perspective on the Subprime Crisis,” IMF Finance and Development, Vol. 45, No. 2. International Monetary Fund (2006), Financial Soundness Indicators Compilation Guide. Kaminsky, G.., S. Lizondo and G. M. Reinhart (1997), “Leading Indicators of Currency Crises,” IMF Working Paper, WP/97/79. Kaminsky, G. (2000), Currency and Banking Crises: The Early Warnings of Distress, George Washington University. (http://www.gwu.edu/~clai/working_papers/Kaminsky_Graciela_07-00.pdf)
- 76 -
Krugman, P. (1979), “A Model of Balance-of-Payments Crises,” Journal of Money, Credit, and Banking, Vol. 11, 311-325. Leong, M. N. (2008), “International Data Standards and Their Application to Macao,” Macao Monetary Research Bulletin, Issue No. 7, April, 63-79. McKinnon, R. I. and H. Pill (1996), “Credible Liberalizations and International Capital Flows: the Overborrowing Syndrome,” Financial Deregulation and Integration in East Asia, NBER-EASE, Vol. 5, 7-50. Montiel, P. and C. Reinhart (1999), The Dynamics of Capital Movements to Emerging Economies during the 1990s, University of Maryland. (http://www.puaf.umd.edu/faculty/reinhart/txt0308.pdf) Zhuang, J. and J. M. Dowling (2002), “Causes of the 1997 Asian Financial Crisis: What Can an Early Warning System Model Tell Us?,” ERD Policy Brief Series, No. 7.
- 77 -
doc_863937541.pdf