Estimation of bank default probability
Investigation of the influence of various microeconomic financial and macroeconomic factors on the probability of a bank default. Analysis of the main reasons for revoking a banking license by the Central Bank of Russia using two binary choice models.
| Рубрика | Банковское, биржевое дело и страхование |
| Вид | дипломная работа |
| Язык | английский |
| Дата добавления | 01.12.2019 |
| Размер файла | 513,8 K |
Отправить свою хорошую работу в базу знаний просто. Используйте форму, расположенную ниже
Студенты, аспиранты, молодые ученые, использующие базу знаний в своей учебе и работе, будут вам очень благодарны.
For small and medium-sized businesses banks also offer settlement accounts and money transfer services. However, they are complemented by services that directly address the needs of the business. We will consider the key ones. Firstly, these are cash and settlement services (the provision of services for the implementation of settlements with customers and suppliers through the connection to the IT-systems of the bank). Secondly, acquiring services, or in other words - services of reception of payments via bank cards. Thirdly, it is collection (the services for the transportation of cash from the cash registers). Here it should be noted a unique feature of the Russian market of collection. Commercial banks are actively represented there, although in developed country markets, most collection services are provided by private specialized corporations. For corporate clients banks provide the same services as for small and medium-sized businesses, but they are complemented by more complex services for liquidity management and information services, which are particularly relevant for large corporations organized in the form of holdings. For example, the bank can provide services to optimize the use of funds within the holding through the consolidation of accounts of subsidiaries. This service is called «cash pooling». An example of information services is to assist the holding company in the preparation of management reporting on the basis of knowledge about all carried out payments within the holding company.
Transactional operations can bring to the bank not only commission income, but also provide two types of business benefits. Firstly, transactional operations are a source of unique customer information that can be used in many applications, such as credit risk assessment. In fact, if the bank sees all transactions of the client, the bank may know about the client even more than the client knows about himself. Secondly, transactional operations are often the starting point for the development of business relations with the client. They allow further cross-selling of more marginal products, such as loans.
Transactional operations also have a number of important financial characteristics. Firstly, transactional operations do not carry financial risks by themselves, thus the development of transactional business increases the financial stability of the bank. It also means that in order to increase the volume of transaction operations, it is not necessary to increase the capital beforehand. In addition, transactional operations can appear considerably less margin compared to loans. However, if we subtract all costs associated with covering risks from the margin on loans, the total marginality of transaction products and loans would be equal. In emerging markets with high inflation the interest margin on loans is usually so large that transactional products are less attractive than loans even after risk has been taken into account. In developed markets, interest margins are significantly lower. Secondly, if the bank opens current accounts for clients within the framework of transactional services, access to cheap funding from these accounts must be taken into the consideration when assessing the economic efficiency of these services. In some cases, access to funding is the main source of profit from the sale of services. Salary projects are a good illustration of it: agreements between the bank and any legal entity on the transfer of salaries of its employees to accounts and payment cards of the bank. Salary projects always involve significant costs for banks. This can be the opening of the accounts, the issuance of cards, as well as the installation of ATMs. These projects are not necessarily paid-for by the client or require cross-purchase of other services of the bank, such as cash and settlement services. Actually many large banks make salary projects free. At the same time, they achieve their self-sufficiency through access to cheap funding and cross-selling.
1.8 The overview of the bank risks
Classification by types or categories of risks is necessary for the further modelling and choosing predictors.
The biggest risk of classical banking business is credit risk, the risk of non-repayment of funds by the debtor in accordance with the terms and conditions of the loan agreement. An effective credit risk management system affects the overall performance of the Bank. This is even more evident in times of crisis.
In second place after losses from credit risk in crisis periods it is the volume of losses of Russian banks due to the implementation of interest rate risk. Interest rate risk is the risk of losses due to changes in interest rates. Most banks attract short and low-cost liabilities and place them in long assets at higher rates, thus earning net interest income, which is the main source of profit of any universal commercial bank. However, when interest rates in the market are rising, banks are forced to urgently raise rates on attracting customers, as short liabilities are quickly overestimated. At the same time, rates on long assets remain unchanged for a long time at low pre-crisis levels, which leads to a decrease in the net interest income of banks.
Most banks make transactions in various currencies, including foreign ones. If the bank's assets in foreign currency are not equal to liabilities in this currency, there is a currency risk - the risk of losses due to adverse changes in foreign exchange rates to the ruble. For example, if the bank has to return to counterparties more dollars than it now has, in case of growth of the dollar against the ruble, it will incur losses for the purchase of more expensive missing dollars.
Another important one for a bank operating in financial markets is a market risk. Market risk of the trading book is the risk of decrease in the value of assets of the trading portfolio due to changes in market factors, such as oil prices or gold.
Ensuring an appropriate level of liquidity is one of the most important tasks of any bank's management. An insufficient level is often the first sign of a bank's serious financial difficulties. The liquidity crisis of banks in 2008 clearly demonstrated how important it is to manage liquidity risk, which can be defined as the risk of losses and even the inability to continue its activities due to the mismatch in maturity of assets and liabilities. Liquidity problems, among others, led to the bankruptcy of Lehman Brothers, formerly one of the world's leading investment banks. The Bank was financed mainly by short-term funds. These funds were invested in long-term low-liquidity instruments. In 2008, at the height of the crisis, access to the borrowing market was closed and the bank was unable to meet its obligations.
And of course any activity (not only banking) inherent operational risk. This is the risk of losses arising from four reasons. The first reason is shortcomings in internal processes. The second is errors or failures in information systems. The third is illegal actions or mistakes of employees. Finally the fourth is external events, such as force majeure. Operational risk is inherent in all banking products, activities, processes and systems. Effective management of operational risk is always one of the main elements of the bank's risk management system.
It is worth noting that the set of risks inherent in banking is not permanent. The world is changing and there are new risks. Recent studies show the increasing importance of cybersecurity risk due to the rapid growth of Information Technologies and cybercrime. Also one of the most significant risks is regulatory risk, which was forced forward by the tightening of regulatory requirements after financial crises.
1.9 Banks business models
From the point of view of the product focus offer, there are four models of banking: universal (or corporate), regional model with a focus on servicing small and medium businesses, retail and investment banking business models.
The first business model is the business model of a universal bank. It is the most important group in the Russian market. Corporate lending plays a more important role for the banks of this group than retail lending, and that is evident from the structure of the assets. On the other hand, the opposite situation is observed on the side of liabilities, where retail deposits are a more important source of funding. This balance sheet structure reflects the history of banking in Russia. In the USSR, banking has traditionally been perceived as primarily work to attract deposits of the population and credit the industry through these deposits. The banking business of post-communist Russia has largely adopted this approach. From the point of view of finance, it is possible to distinguish a number of features of universal banks. Firstly, working primarily with corporate borrowers, they carry less credit risk than other business models banks. Secondly, they operate with a large maturity gap and have developed competencies for managing balance sheet risks due to the presence of long-term corporate loans in the portfolio. Thirdly, they are more focused on mortgage lending from the point of view of retail lending. Moreover, they are the leaders of the mortgage lending market.
The second business model is small regional banks that specialize in servicing small and medium-sized businesses in their region. This is a very important business model in the banking systems of almost all developed countries. For example, there are the so-called community banks in the US and banking cooperatives in Germany and Spain. The main idea of this type of banks is to provide quality loans to small and medium-sized businesses in their region of presence due to the fact that the bank's management has information about what and who of the local entrepreneurs actually does, what risks and needs are relevant. In lending to SMEs (small and medium-sized enterprises) it is often difficult to rely on anything other than personal knowledge of the local economy by the bank's employees, as the SME usually does not have audited financial statements. At the same time, the speed of decision-making is very important for the SMB (small and medium-sized business) enterprise, as unlike corporations they have much less financial reserves. For this reason, large banks, with their inevitably more complex decision - making processes, may be inferior in efficiency to small banks. In Russia, the model of regional SMB banks does not function very successfully. Regional SMB banks most often became either targets of acquisitions by large players, or they turned themselves into large universal banks if they survived on their own. There are two reasons for this situation with SMB banks in Russia. Firstly, even small banks in Russia must comply with a full set of numerous regulatory requirements. The regulatory burden creates fixed costs that are much easier to bear for larger banks due to the economy of scale. Secondly, the system of regional banks in developed countries has grown organically for decades or even centuries. In many cases, these banks are family businesses. In Russia, the banking system grew on the ruins of the centralized financial infrastructure of the USSR, which initially predetermined the focus on building banks at least at the level of the largest cities and subjects of the Federation.
The third business model is the business model of specialized retail banks. These banks focus on consumer lending in the form of conventional loans and credit cards, although formally most of them are universal and have a certain share of business with legal entities. From the financial point of view, these banks operate with a higher level of credit risk and thereby a higher interest margin compared to universal banks. In addition, consumer loans are relatively short. They are on average in the area of one year, so these banks are not characterized by a high gap of urgency, thus they earn primarily on the transformation of credit risk, and not on the production of liquidity. An important feature of the Russian market in comparison with the banking markets of other countries is the lack of specialized mortgage banks. As we mentioned earlier, Russian retail banks are virtually absent from the mortgage market. The leaders here are universal banks. It happens due to the fact that the issuance of mortgage loans means taking high liquidity risks and interest rate risk of the bank portfolio because of their long terms. Small banks can not bear these risks, so the model of a specialized mortgage bank is possible only in those markets where there are many tools to hedge these risks. For example, there is a developed market of mortgage-backed securities in the US, as well as quite common products with floating interest rates. Thus, this business model of specialized mortgage banks works there.
The fourth business model is investment banks. Those banks earn the main profit on two specialized types of operations. Firstly, these are trading operations in the financial markets. These operations carry a high level of risk and are subject to strong fluctuations during the economic cycle. Despite the fact that traders can theoretically make money on falling financial markets, on average, they suffer losses in falling markets. The second type of specialized operations is investment banking commission services. Firstly, it is the services of placement of shares in financial markets (ECM - Equity Capital Markets), secondly, it is the services of placement of bonds in financial markets (DCM - Debt Capital Markets), thirdly, it is consulting services on mergers and acquisitions (M&A), and, fourthly, it is services to provide complex structured financing (primarily syndicated loans). The golden era of investment banking in Russia fell on the pre-crisis 2007 and 2008 years, as well as on the peak of post-crisis recovery in 2011 year. The activity in the markets of investment banking services is sharply reduced in crisis years, and this is another risk factor for this business model. The key players in the market of investment banking services in Russia are investment units of state banks, as well as international investment banks.
Many banks represented on the Russian market are members of various formal and informal financial and industrial groups. When the owners of the bank have a non-bank business, commensurate or exceeding the size of the bank business, it gives an opportunity to analyze the relationship between the bank and the parent group through the established business model. There are three types of significant business model relationships between the group and the subsidiary bank.
The first type of relationship is the group's support of the bank's business in various forms. This may be, for example, the provision of customer relations to the bank. It can also be the provision of retail network offices for the sale of banking services. In general, this kind of interaction does not distort the product business model of the bank.
The second type of relationship is the bank's support of the group's business through the provision of so-called loans to related parties. If the share of such loans in the bank's portfolio is large, the bank ceases to be a bank from an economic point of view and becomes a part of the corporate treasury for the group. Such banks are called captive banks.
The third type of relationship is the bank's support for the group's business by financing sales of the group's main product. Such banks can be called complimentary banks. In the second and in the third cases, the banks' business is actually the secondary one to the group's business.
Captive banks, also called pocket or wallet banks, are generally a negative phenomenon for the banking sector. If the bank is mainly engaged in lending to related parties, there is a number of problems. Firstly, such loans are often issued on non-market terms and without adequate risk control. This case undermines the financial stability of the bank. Secondly, the level of diversification of credit risk in the portfolio of such banks is low, so there is no effect of transformation of the credit risk level. Thirdly, the use of such banks allows a group of companies to attract unfairly cheap funds, deceiving investors. Investors, such as retail depositors, think that they lend to a bank that has a low level of risk by diversifying the portfolio, and therefore they agree to a lower rate than in case of lending to a group of companies directly, for example through the purchase of bonds. This effect causes the prevalence of captive banks in the market. Fourthly, lending to related parties allows funds to be withdrawn from the bank in the form of non-refundable loans. For all these reasons, the Russian Central Bank like banking regulators in all countries is struggling with captive banking. The main regulatory instrument is currently the standard for the maximum amount of loans to related parties. Nevertheless, the Central Bank does not always manage to prevent the growth of captive banks to a dangerous size for the economy. Many high-profile defaults of Russian banks in recent years are associated with the captive business model.
Complimentary banking business model is presented on the Russian market by so-called auto banks. These banks are engaged in lending to the sales of the respective car brands. The business idea is as follows. Firstly, the offer of lending services to buyers is beneficial for automakers, as it allows to increase car sales. If the lending is carried out by a subsidiary bank, it is possible to subsidize the rates, covering the lost profit at the expense of increased sales of the group. Secondly, choosing between a partnership with an external bank and the use of its own bank, the argument in favor of using its own bank is to preserve the profits of the group's banks and various synergies. Firstly, it is easy to integrate IT of complimentary bank with IT of dealerships. Secondly, it is the ability to use a single database of customers. In fact, the idea of complementary banking is the vertical integration of sales of any goods to individuals on credit. Wherever there is a place for lending at the point of sale (the so-called post-lending), there is potentially a place for complementary banking. At the same time, it is important to understand that the development of its own subsidiary bank to support the sales of the main product can pay off only if there is a scale effect. Therefore, in most cases retailers and manufacturers prefer to enter into partnerships with existing banks to lend to customers.
1.10 The literature review
Bank Default is associated with Russian Central Bank recalling the banking license from a commercial bank. With the fact of recalling the license at the first place Central Bank give a reason for this movement. Most commonly, those reasons are: «money laundering» (criminal activities, associated with cashing illegally derived assets or money), «insolvency» (breaching the determined limits of financial indicators), «violation of the law» (for an example, falsification of financial and bookkeeping accountability) and the scarce «voluntary reasons».
Our work is based on the previous attempts to predict bank default, which can be partly illustrated by the follow papers:
It needs to be mentioned here, that some works base their modelling on the discriminant analysis. Altman E. I. combined discriminant analysis and financial ratio analysis in order to evaluate the performance of the business (Altman, 1968). Eventually it developed into the famous Altman Z-score formula for predicting the bankruptcy within two years. It was such a successful and significant move in this sphere, that even now many companies use a modified Altman Z-score formula. Later Altman E. I., Haldeman R. and Narayanan P. upgraded the original Altman Z-score formula, expanding the prognostic period from two to five years (Altman, 1977). Izan H. Y. applied the similar methodology to the Australian Enterprises (Izan, 1984). However, we suppose that discriminant analysis require an unrealistic assumption, which very often is ignored: the normal distribution of the repressors. Ignoring this assumption, we get inaccurate estimates of the parameters, which is the case with our dataset. In addition, using binary choice models allow us to estimate the significance of the parameters in our model. While examining discriminant models Scott J. stated that the bankruptcy detection could be best explained by the ordinary cashflows (Scott, 1981). To tell more, Lennox C. confirmed, that binary choice logit and probit models demonstrate better results than the discriminant analysis (Lennox, 1999).
Among the works, that uses probit and logit binary models without pre-modelling clusterization we can mention the follow ones. Martin D. first applied econometric model of binary choice (logit model) to predict defaults of US banks for the period from 1975 until 1976 (Martin, 1977). Ohlson J. A. implemented the logit model in the forecasting of default by using accounting ratios as predictors (Ohlson, 1980). Wiginton J. C. compared the logit and discriminant models of consumer credit behavior in order to find the most appropriate quantitative model for credit-granting decisions (Wiginton, 1980). Bovenzi J. F., Marino J. A. and McFadden F. E. continued to analyze the commercial bank failure with the help of prediction models (Bovenzi et al., 1983). Cole R. A. and Gunther J. W. (Cole, 1995; Cole, 1998) estimated the probability of bank default through the use of financial ratios. Moreover, they found, that fact of assigning the new ratio influence on the bankruptcy probability with two or three quarters time lags (Cole et al., 1995; Cole et al., 1998). Estrella A., Park S. and Peristiani S. examined the usefulness of several capital ratios - the leverage, the gross revenues, and the risk-weighted assets - in predicting bank default over various time lags (Estrella et al., 2000). Westgaards S. and Wijst van der N. showed a method of assessing retail bank portfolio failure probabilities basing the analysis on a logistic regression model (Westgaards et al., 2001). Kolari J., Glennon D., Shin H. and Caputo M. continued to apply the logit regression analysis to the large USA commercial banks (Kolari et al., 2002). Godlewski C. J. once again implemented the financial ratings into the bank default probability as the Pillar 3 of the Basel II reform has come at the first place (Godlewski, 2007). Raffaella Calabrese and Paolo Giudici found out that, with regard to the bank defaults in Italy, while capital ratios are significant to explain «ordinary» failures, the macroeconomic ones are relevant only in case of mergers and acquisitions (Raffaella et al., 2015). Kavussanos M. G. and Tsouknidis D. A. combine different debt coverage ratios and liquidity ratios, getting thereby the best ratios in term of applying these indicators to the binary choice models (Kavussanos et al., 2016).
The most solid work for the Russian bank dataset is probably the work of Peresetsky A.A. (2007). He uses the probit and logit models for predicting the possible bank default. In this work he made a pre-modelling clusterization of banks in order to illustrate different contribution of various factors to the default probability based on the bank differentiating procedure. Actually, we suppose, these method can not be called a potentially universal one, because clusterization (based on the capital ownership) depends on variety of banks in the market instead of default reasoning diversity. Probably with new banks leaving the sphere a new clusterization is required, which to some extent makes predicting procedure over-complicated (Пересецкий, 2007).
With regard to the Russian market the most significant works are Peresetsky A. A., Karminsky A. M. and Golovan S. V. (Peresetsky et al., 2004; Peresetsky et al., 2011), A. M. Karminsky and A. V. Kostrov (Karminsky et al., 2014), Ivanov V. V. and Fedorova Y. I (Ivanov et al., 2015), Bondarenko M. V. and Semenova M. V. (Bondarenko et al., 2018). Actually, despite varying different factors for predicting the default we consider those works as a platform to combine various indicators in order to improve predictive performance capabilities of logit binary choice model as well as all of them are based on the binary choice analysis (Peresetsky et al., 2004; 2004; Peresetsky et al., 2011; Karminsky et al., 2014; Ivanov et al., 2015; Bondarenko et al., 2018).
In the next work of Peresetsky A. A., he uses multinomial logit regression model to predict different types of default, but in his work he also makes several probit and logit models, and they demonstrate equal performance capabilities to the multinomial logit regression model. We tend to consider several binary choice models better than the only one multinomial logit model, because in the case of another license recalling reason separation the researcher is forced to build just one more regression instead of remaking the only one constructed before. Another advantage of several binary choice models over the multinomial logit models is the possibility to estimate the marginal effects. It makes the interpretation process much easier (Пересецкий, 2013). Similar findings are demonstrated in the work of Jagtiani J., Kolari J., Lemieux C. and Shin H. They stated the advantage of logit models in bank default prediction over the Early Warning Systems (EWS) and Trait Recognition Algorithm (TRA) as a research finding (Jagtiani et al., 2003).
2. Stating the research question
We state research question for our work as follow: look for the regressors, that are best suitable for estimating bank default probability by the use of several binary choice models. For this purpose, we may gather publicly available information, which is a clear demonstration, that even in this sphere it is possible to construct a model without possession of insider information.
We state the following hypotheses in our work:
H1: The increase in the amount of bank net assets accounts negatively to the probability of default for both «money laundering» and «economic reasons» default probabilities. The default of large banks often initiates financial health problems for a lot of enterprises and banks related to this bank. Moreover, the fall in the market value of big bank shares starts panic in the stock market. Thereafter, the investments companies, pension funds and hedge funds are affected. This problem is called «too big to fail». That is the reason, why the Central Bank and other large banks usually undertake some actions to prevent problematical bank from failure. That is why, the size of the bank itself may be a predictor of default.
H2: The increase in the amount of the interbank loans given accounts negatively to the probability of default for both «money laundering» and «economic reasons» default probabilities.
Usually only big banks credit other banks on the interbank capital market. Those banks has higher investment rating, better key performance indicators and more solid financial health. The complex of all this facts lead us to the fact, that their default is much less probable to occur.
H3: Investments in shares indicator will be statistically significant both «money laundering» and «economic reasons» default probabilities at least at 10 percent of statistical significance. The investments in shares makes banks portfolio much more volatile. Thus the instability of financial performance may be a significant source of banks bankruptcy risk.
3. Methodology of the research
This work has two important preconditions:
Firstly, we divide the bank default into two categories: «money laundering» and «economic reasons».
Secondly, we use quarterly accountings. We suppose, that varying lags in terms of quarterly time score is best suitable for the bank default probability estimation.
The need to involve time lags into the model lies in the field of endogeneity. There are indicators that are unobservable for the external individual, but observable for the bank officer. That is partly the reason, why the bank may start taking assets out. Taking the time lag helps to get rid of endogeneity.
We are interested in varying time lags within one year (four quarters) before the accident, because longer time lags are almost impossible to interpret.
The final data results in panel dataset of 798 banks and the quarterly recorded indicators from 2014 to 2018. All data manipulations and econometric modelling were done with the help of econometric statistical package Eviews 7.
Before introducing the set of chosen regressors it needs to be stated here, that our analysis rests mainly upon well-known CAMELS system of indicators. It is one of the most developed and internationally recognized approaches to the analysis of financial stability of banks. In the CAMELS system, each letter refers to one aspect of the financial stability of bank: capital adequacy, asset quality, management quality, profitability, liquidity and funding, sensitivity to market risk. In this paper we ignore management and sensitivity parts of bank analysis, thus focusing attention on the rest four parts.
Capital adequacy is the most important indicator of financial stability, as it determines the ability of the bank to absorb losses from the implementation of all risks including business risk (the risk of reducing profits from business operations). Moreover in practice banks often cease to exist due to violations of the regulator's minimum capital requirements. Capital adequacy is calculated as the ratio of regulatory capital of different types to the assessment of losses on credit, market and operational risks. This estimate is expressed in the amount of risk-weighted assets, which are calculated according to the rules established by the regulator, but are generally correlated with the value of operating assets. Thus the amount of capital sets limits on business growth. It is possible to violate the capital adequacy not only due to the increase in the level of risk, which affects the financial results of capital through the creation of reserves, but also on the basis of the increase in business volumes at consistent average level of risk. Consequently capital adequacy should be assessed on the basis of the safety margin of the estimated capital adequacy metrics compared to the minimum regulatory requirements. It should be noted that capital adequacy cannot be calculated on the basis of the bank's balance sheet and income statement. Regulatory capital differs from accounting capital as it is calculated on the assumption of liquidation of the bank. Consequently, it does not include certain assets that have zero value in liquidation, such as goodwill. Subordinated debt is also included in the calculation of regulatory capital as it has a privilege on the absorption of losses for senior creditors in the event of financial difficulties with the bank. Similarly the amount of risk-weighted assets is also calculated according to special regulatory rules. It should be noted that high capital adequacy values should also serve as a dangerous signal for the financial analyst, since very high capital adequacy means low profitability for shareholders.
The quality of assets is most often understood as the quality of the loan portfolio and the portfolio of debt securities, in other words the quality of assets bearing credit risk. The quality of assets is assessed using a number of indicators: the share of loan portfolio reservation, the share of loans overdue by more than 90 days, the cost of risk.
The bank's profitability is important for financial stability as the bank's capital is formed and increased from the profit. Rough profitability indicators include the following ones: net interest margin (this is the ratio of net interest income for the period to the average carrying amount of interest-bearing assets before provisions), expense to income ratio (it is calculated as the ratio of operating expenses to operating income of the bank for a certain period), return on equity (ROE is the ratio of the bank's net profit to the average value of its own funds for a certain period), return on assets (ROA is calculated as the ratio of net profit to the average value of assets for a certain period of time). It should be noted that the ROE indicator is universal but when comparing banks with significantly different indicators of capital adequacy it is necessary to pay attention to the ROA indicator.
The implementation of liquidity risk can lead to bankruptcy of even a very well-capitalized bank. Therefore the analysis of liquidity is the second most important after the analysis of capital adequacy. The bank's liquidity is estimated using two groups of metrics. The first group of metrics describes the ratio of the bank's liquidity reserves to potential liquidity outflows due to the instability of funding sources. This group includes the following indicators: instant liquidity ratio, short-term liquidity ratio, liquidity coverage ratio. The second group of metrics describes the ratio of existing funding to assets requiring such funding. This group includes the following indicators: long-term liquidity ratio, net stable funding ratio, the ratio of loans to deposits.
We use two groups of factors.
Microeconomic financial indicators:
1) Natural Logarithm of Net Assets;
2) RONA;
3) Highly Liquid Assets to Net Assets;
4) Loans to Individuals to Net Assets;
5) Overdue Debt to Loans to Individuals;
6) Interbank Loans Given to Net Assets;
7) Loans to Enterprises and Organizations to Net Assets;
8) Overdue Debt to Loans to Enterprises and Organizations;
9) Investments in Shares to Net Assets;
10) Investments in Bonds to Net Assets;
11) Investments in Promissory Notes to Net Assets;
12) Deposits of Individuals to Liabilities;
13) Funds of Enterprises and Organizations to Liabilities;
14) Interbank Loans Gained to Liabilities;
15) Bonds and Promissory Notes Issued to Liabilities;
Macroeconomic indicators:
1) Change in GDP to the previous quarter in current prices (%);
2) Deflator Index (% of the corresponding quarter of the previous year);
3) Key Interest Rate of Central Bank (%);
4) The Exchange Rate of the Ruble to the US Dollar;
«Change in GDP», «Deflator Index», «Key Interest Rate of Central Bank» and «The Exchange Rate of the Ruble to the US Dollar» refer to the macroeconomic environment, the change in which could influence on the banks' business processes.
«Natural Logarithm of Net Assets», «RONA» (Return On Net Assets) and «Highly Liquid Asset to Net Assets» refer to the microeconomic financial indicators of bank's performance. We use the «Natural Logarithm of Net Assets» instead of «Net Assets» in order to smooth the difference between large banks and the small ones. For the same purpose we used the share of «Highly Liquid Asset to Net Assets» instead of «Highly Liquid Assets». The use of the «Highly Liquid Asset to Net Assets» justified by the need to take into account the Liquidity Risk.
«Loans to Individuals to Net Assets», «Interbank Loans Given to Net Assets» and «Loans to Enterprises and Organizations to Net Assets» refer to the allocation of Bank's Assets. The division on the amount of Net Assets was done in order to smooth the difference between large banks and the small ones. The shares «Overdue Debt to Loans to Enterprises and Organizations» and «Overdue Debt to Loans to Individuals» would signify the role of Overdue Debt in the default.
«Investments in Shares to Net Assets», «Investments in Bonds to Net Assets», «Investments in Promissory Notes to Net Assets» refer to the structure of bank investments portfolio. The division on the amount of Net Assets was done in order to smooth the difference between large banks and the small ones.
«Deposits of Individuals to Liabilities», «Funds of Enterprises and Organizations to Liabilities», «Interbank Loans Gaines to Liabilities» and «Bonds and Promissory Notes Issued to Liabilities» refer to the source of funding. The division on the amount of Liabilities was done in order to smooth the difference between large banks and the small ones.
Now, we are going to introduce the underlying econometric theory of this research.
We use logit regression analysis, which is the binary choice model. Binary choice model are the particular case of the limited dependent variable models. In our case the dependent variable may take the value of «0» (survival) or «1» (license revocation). There are two logit regressions for two reasons of default: «economic reasoning» and «money laundering». In case of logit regression, the parameters are estimated by the maximum-likelihood method, because the model is not linear by the parameters:
Where: - the occurrence probability of the event «i»;
- the vector of parameters;
- the vector of occurrence predictors.
In order to compare different models we will use the following metrics:
Firstly, we use McFadden R-squared, which is based on the comparison of the chosen model with the «naive» one (which includes only constant as independent variable). It is calculated in the following way:
Where: - McFadden R-squared;
- natural logarithm of the chosen model likelihood ratio;
- natural logarithm of the «naive» model likelihood ratio.
It is evident from the formula that the higher McFadden R-squared is the better the model is.
Secondly, we use the set of metrics, that refer to the number of correctly and incorrectly predicted outcomes. In order to calculate necessary metrics, we need to fill the gaps in the following table:
Table 1 The number of correctly and incorrectly predicted outcomes
|
Forecasted Default |
Forecasted Survival |
||
|
Real Default |
TP |
FN |
|
|
Real Survival |
FP |
TN |
Where: «Forecasted Default» and «Forecasted Survival» stand for the model default and survival respectively.
«Real Default» and «Real Survival» stand for the real-life default and survival respectively.
Therefore, «TP» stands for True Positives outcomes, «TN» stands for True Negative outcomes, «FP» stands for False Positives outcomes (so-called alpha error) and «FN» stands for False Negative outcomes (so-called beta error).
Then it is possible to calculate the following metrics:
Where: - model sensitivity (the proportion of true positive cases);
- True Positives outcomes;
- False Negative outcomes.
Where: - model specificity (the proportion of true negative cases);
- True Negative outcomes;
- False Positives outcomes.
Where: - the proportion of correctly classified outcomes (compared with the "naive" model);
- True Positives outcomes;
- True Negative outcomes;
n - number of sample cases.
It is evident from the formulas that the higher the model sensitivity, the model specificity and countable R-squared are the better the model is.
Thirdly, we use information criteria: Akaike info criterion and Schwarz info criterion.
Where: - Akaike info criterion;
- True Positives outcomes;
- True Negative outcomes;
n - number of sample cases.
Where: - Schwarz info criterion;
- natural logarithm of the chosen model likelihood ratio;
n - number of sample cases;
- number of predictors.
It is evident from the formulas, that the lower the information criteria are, the better the model is.
While estimating the distribution type of our dataset, it is evident from the histograms (Appendix 1-2) and the Jarque-Bera probabilities from the descriptive statistics (Appendix 3-4), that our dataset has a distribution, which differ from the Normal one. As it was said in the Literature review that is an argument to use logit regression analysis against the discriminant analysis which is exactly the case of our paper. bank default license binary
4. Results description
We estimated the two binary choice logit models for two default reasonings:
Table 2 The binary choice logit model for «economic reasons» license revocation before varying time lags
|
Variable |
Estimate |
Standard Error |
|
|
Const |
-5,51 |
3,89 |
|
|
Change in GDP *** |
0,08 |
0,01 |
|
|
Deflator Index |
0,04 |
0,03 |
|
|
Key Interest Rate |
0,02 |
0,05 |
|
|
The Exchange Rate of the Ruble to the US Dollar *** |
0,04 |
0,01 |
|
|
RONA (Return On Net Assets) *** |
-0,04 |
0,01 |
|
|
Natural Logarithm of Net Assets *** |
-0,27 |
0,08 |
|
|
Highly Liquid Assets to Net Assets *** |
-0,05 |
0,02 |
|
|
Loans to Individuals to Net Assets *** |
-0,04 |
0,01 |
|
|
Overdue Debt to Loans to Individuals * |
0,01 |
0,01 |
|
|
Interbank Loans Given to Net Assets *** |
-0,09 |
0,02 |
|
|
Loans to Enterprises and Organizations to Net Assets |
-0,01 |
0,01 |
|
|
Overdue Debt to Loans to Enterprises and Organizations |
-0,00 |
0,01 |
|
|
Investments in Shares to Net Assets |
-0,02 |
0,04 |
|
|
Investments in Bonds to Net Assets ** |
-0,03 |
0,02 |
|
|
Investments in Promissory Notes to Net Assets *** |
0,05 |
0,02 |
|
|
Deposits of Individuals to Liabilities *** |
0,02 |
0,01 |
|
|
Funds of Enterprises and Organizations to Liabilities ** |
-0,03 |
0,01 |
|
|
Interbank Loans Gained to Liabilities |
-0,03 |
0,02 |
|
|
Bonds and Promissory Notes Issued to Liabilities ** |
0,04 |
0,02 |
|
|
McFadden R-squared |
0,24 |
||
|
Akaike info criterion |
0,09 |
||
|
Schwarz info criterion |
0,10 |
Note. *, **, *** mean 10%, 5%, 1% of statistical significance respectively
Table 3 The binary choice logit model for «money laundering» license revocation before varying time lags
|
Variable |
Estimate |
Standard Error |
|
|
Const *** |
-18,37 |
5,10 |
|
|
Change in GDP *** |
0,09 |
0,01 |
|
|
Deflator Index *** |
0,15 |
0,04 |
|
|
Key Interest Rate |
-0,02 |
0,07 |
|
|
The Exchange Rate of the Ruble to the US Dollar *** |
0,05 |
0,01 |
|
|
RONA (Return On Net Assets) *** |
-0,03 |
0,01 |
|
|
Natural Logarithm of Net Assets *** |
-0,42 |
0,09 |
|
|
Highly Liquid Assets to Net Assets |
-0,01 |
0,01 |
|
|
Loans to Individuals to Net Assets |
0,00 |
0,01 |
|
|
Overdue Debt to Loans to Individuals ** |
0,01 |
0,01 |
|
|
Interbank Loans Given to Net Assets *** |
-0,05 |
0,01 |
|
|
Loans to Enterprises and Organizations to Net Assets |
-0,01 |
0,01 |
|
|
Overdue Debt to Loans to Enterprises and Organizations |
-0,00 |
0,01 |
|
|
Investments in Shares to Net Assets ** |
0,07 |
0,03 |
|
|
Investments in Bonds to Net Assets |
-0,02 |
0,02 |
|
|
Investments in Promissory Notes to Net Assets * |
0,05 |
0,03 |
|
|
Deposits of Individuals to Liabilities ** |
0,02 |
0,01 |
|
|
Funds of Enterprises and Organizations to Liabilities |
0,01 |
0,01 |
|
|
Interbank Loans Gained to Liabilities |
-0,01 |
0,02 |
|
|
Bonds and Promissory Notes Issued to Liabilities |
-0,08 |
0,07 |
|
|
McFadden R-squared |
0,18 |
||
|
Akaike info criterion |
0,06 |
||
|
Schwarz info criterion |
0,08 |
Note. *, **, *** mean 10%, 5%, 1% of statistical significance respectively
Then we varied time lags from «minus one quarter» to «minus four quarters» for all predictors, starting from the «Change in GDP» and finishing with «Bonds and Principal Notes to Liabilities». The time lag was chosen for every variable in accordance with the principle of McFadden R-squared maximization. Then we repeated the procedure before McFadden R-squared stopped showing an increase.
Table 4 The binary choice logit model for «economic reasons» license revocation after varying time lags
|
Variable |
Estimate |
Standard Error |
|
|
Const *** |
41,06 |
13,00 |
|
|
Change in GDP(-1) *** |
-0,12 |
0,03 |
|
|
Deflator Index(-3) *** |
-0,59 |
0,17 |
|
|
Key Interest Rate(-4) *** |
-0,71 |
0,15 |
|
|
The Exchange Rate of the Ruble to the US Dollar *** |
0,52 |
0,12 |
|
|
RONA (Return On Net Assets) *** |
-0,04 |
0,01 |
|
|
Natural Logarithm of Net Assets(-4) *** |
-0,42 |
0,11 |
|
|
Highly Liquid Assets to Net Assets *** |
-0,06 |
0,02 |
|
|
Loans to Individuals to Net Assets(-3) *** |
-0,04 |
0,01 |
|
|
Overdue Debt to Loans to Individuals |
0,01 |
0,01 |
|
|
Interbank Loans Given to Net Assets *** |
-0,08 |
0,02 |
|
|
Loans to Enterprises and Organizations to Net Assets(-1) |
-0,01 |
0,01 |
|
|
Overdue Debt to Loans to Enterprises and Organizations(-1) |
0,01 |
0,01 |
|
|
Investments in Shares to Net Assets(-1) |
0,06 |
0,04 |
|
|
Investments in Bonds to Net Assets *** |
-0,05 |
0,02 |
|
|
Investments in Promissory Notes to Net Assets *** |
0,09 |
0,02 |
|
|
Deposits of Individuals to Liabilities *** |
0,03 |
0,01 |
|
|
Funds of Enterprises and Organizations to Liabilities |
-0,02 |
0,01 |
|
|
Interbank Loans Gained to Liabilities |
-0,02 |
0,02 |
|
|
Bonds and Promissory Notes Issued to Liabilities(-3) *** |
0,07 |
0,02 |
|
|
McFadden R-squared |
0,30 |
||
|
Akaike info criterion |
0,09 |
||
|
Schwarz info criterion |
0,10 |
Note. *, **, *** mean 10%, 5%, 1% of statistical significance respectively
Table 5 The binary choice logit model for «money laundering» license revocation after varying time lags
|
Variable |
Estimate |
Standard Error |
|
|
Const *** |
-34,44 |
6,04 |
|
|
Change in GDP(-2) *** |
0,13 |
0,01 |
|
|
Deflator Index *** |
0,28 |
0,05 |
|
|
Key Interest Rate(-1) *** |
0,23 |
0,07 |
|
|
The Exchange Rate of the Ruble to the US Dollar (-2) *** |
0,06 |
0,02 |
|
|
RONA (Return On Net Assets) *** |
-0,05 |
0,01 |
|
|
Natural Logarithm of Net Assets *** |
-0,42 |
0,09 |
|
|
Highly Liquid Assets to Net Assets(-3) *** |
-0,03 |
0,01 |
|
|
Loans to Individuals to Net Assets(-2) *** |
-0,03 |
0,01 |
|
|
Overdue Debt to Loans to Individuals(-1) |
0,01 |
0,01 |
|
|
Interbank Loans Given to Net Assets *** |
-0,06 |
0,01 |
|
|
Loans to Enterprises and Organizations to Net Assets(-2) *** |
-0,04 |
0,01 |
|
|
Overdue Debt to Loans to Enterprises and Organizations(-4) |
0,01 |
0,01 |
|
|
Investments in Shares to Net Assets ** |
0,08 |
0,03 |
|
|
Investments in Bonds to Net Assets(-3) *** |
-0,05 |
0,02 |
|
|
Investments in Promissory Notes to Net Assets * |
0,06 |
0,03 |
|
|
Deposits of Individuals to Liabilities(-1) *** |
0,03 |
0,01 |
|
|
Funds of Enterprises and Organizations to Liabilities(-1) ** |
0,02 |
0,01 |
|
|
Interbank Loans Gained to Liabilities |
-0,03 |
0,02 |
|
|
Bonds and Promissory Notes Issued to Liabilities(-2) |
0,03 |
0,03 |
|
|
McFadden R-squared |
0,25 |
||
|
Akaike info criterion |
0,07 |
||
|
Schwarz info criterion |
0,08 |
Note. *, **, *** mean 10%, 5%, 1% of statistical significance respectively
As we see for «economic reasons» default model after varying time lags McFadden R-squared has increased from 0,236 to 0,297 (Table 2 and Table 4). However we can not choose better model on the basis of information criteria: Akaike info criterion has insignificantly decreased and Schwarz info criterion has insignificantly increased. Also the number of statistically significant variables at least at level of 10% has risen from 13 to 14.
As for «money laundering» default model, we can see that after varying time lags McFadden R-squared has increased from 0,177 to 0,247 (Table 3 and Table 5). The information criteria does not lead us to better model: both Akaike and Schwarz info criteria have insignificantly increased. Also the number of statistically significant variables at least at level of 10% has risen from 11 to 16.
Table 6 «Economic reasons» default model before varying time lags: the number of correctly and incorrectly predicted outcomes
|
Forecasted Default |
Forecasted Survival |
||
|
Real Default |
4 |
5 |
|
|
Real Survival |
114 |
12107 |
Table 7 «Money laundering» default before varying time lags: the number of correctly and incorrectly predicted outcomes
|
Forecasted Default |
Forecasted Survival |
||
|
Real Default |
2 |
1 |
|
|
Real Survival |
72 |
12111 |
Table 8 «Economic reasons» default after varying time lags: the number of correctly and incorrectly predicted outcomes
|
Forecasted Default |
Forecasted Survival |
||
|
Real Default |
7 |
3 |
|
|
Real Survival |
88 |
8998 |
Table 9 «Money laundering» default after varying time lags: the number of correctly and incorrectly predicted outcomes
|
Forecasted Default |
Forecasted Survival |
||
|
Real Default |
3 |
0 |
|
|
Real Survival |
59 |
9001 |
As we see, varying time lags for «economic reasons» default model increases the sensitivity from 44, 44 percent to 70 percent, the specificity insignificantly falls from 99, 07 percent to 99, 03 percent, countable R-squared insignificantly falls from 99, 03 percent to 99, 00 percent.
As for another case, varying time lags for «economic reasons» default model increases the sensitivity from 44, 44 percent to 70 percent, the specificity insignificantly falls from 99, 41 percent to 99, 35 percent, countable R-squared insignificantly falls from 99, 40 percent to 99, 35 percent.
All comparisons made above lead us to the conclusion that varying time lags helps to improve models predictive ...
Подобные документы
Development banking, increasing the degree of integration of the banking sector of Ukraine in the international financial community, empowerment of modern financial markets, increasing range of banking products. The management mechanism of bank liquidity.
реферат [17,2 K], добавлен 26.05.2013The principal types of banking in the modern world are commercial banking and central banking. The provision of safe deposit facilities for money and valuables. Establishing a bank account. Cashier’s checks. Characteristic of the central bank in the UK.
презентация [1,1 M], добавлен 23.03.2015General information about Asya Participation Bank. Offering uninterrupted, rapid and effective service via Online Banking. Capital and Shareholder Structure. Affiliates and subsidiaries. The leader of participation banking. Bank Asya’s Objectives.
курсовая работа [1,4 M], добавлен 01.11.2011A bank: nature of activity, main business-processes and organizational structure, the market place and history. Definitions of the project and project management, the project life cycle. Management of development projects in a bank, the expected results.
реферат [20,6 K], добавлен 14.02.2016Financial position of the "BTA Bank", prospects, business strategy, management plans and objectives. Forward-looking statements, risks, uncertainties and other factors that may cause actual results of operations; strategy and business environment.
презентация [510,7 K], добавлен 17.02.2013The Banking System of USA. Central, Commercial Banking and the Development of the Federal Reserve and Monetary Policy. Depository Institutions: Commercial Banks and Banking Structure. Banking System in Transition. Role of the National Bank of Ukraine.
научная работа [192,0 K], добавлен 22.01.2010Commercial banks as the main segment market economy. Principles and functions of commercial banks. Legal framework of commercial operation banks. The term "banking risks". Analysis of risks and methods of their regulation. Methods of risk management.
дипломная работа [95,2 K], добавлен 19.01.2014Краткая финансово-экономическая характеристика деятельности ОАО "Optima Bank", адекватность капитала. Процедура учета и организация документооборота расчетно-кассовых операций. Коэффициенты эффективности использования обязательств коммерческого банка.
отчет по практике [42,3 K], добавлен 29.01.2015Рoль вклaдoв клиентoв в фoрмирoвaние реcурcнoй бaзы бaнкa. Клaccификaция бaнкoвcкиx депoзитoв. Xaрaктериcтика АО "Kaspi Bank", анализ его финaнcoвo-xoзяйcтвенной деятельнocти. Aнaлиз депoзитнoгo пoртфеля бaнкa, его прoблемы и перcпективы развития.
дипломная работа [289,2 K], добавлен 21.05.2012History of introduction of a modern banking system to the Muslim countries, features of their development and functioning in today's market economy. Perspectives of future development of Islamic banking in the world and in the Republic of Kazakhstan.
курсовая работа [1,3 M], добавлен 19.04.2012Сущность понятия "ипотечное кредитование". Объемы ипотечного кредитования в Казахстане. Основные источники финансирования жилищного строительства Астаны. Кредитный портфель АО "Kaspi Bank". Предложения по совершенствованию ипотечного кредитования.
доклад [14,2 K], добавлен 09.12.2010Asian Development Fund. Poverty reduction in Asia and the Pacific. Promotion of pro poor, sustainable economic growth. Supporting social development. Facilitating good governance. Long-term Strategic Framework. Private, financial sector development.
презентация [298,7 K], добавлен 08.07.2013History of the online payment systems. Payment service providers. Online bill payments and bank transefrs. Pros and cons for using online payment systems. Card Holder Based On Biometrics. Theft in online payment system. Online banking services, risk.
реферат [37,2 K], добавлен 26.05.2014The concept and general characteristics of the banking system and its main elements of the claimant. Current trends and prospects of development of the banking system, methods of its realization, legal foundation. Modern banking services in Ukraine.
контрольная работа [21,7 K], добавлен 02.10.2013Оценка современного состояния и перспектив дальнейшего развития банковской системы Казахстана, причины опережения развития по сравнению с постсоветскими странами. Характеристика "HSBC Bank Kazakhstan", анализ и оценка его сервисов, микро- и медиасреда.
презентация [125,7 K], добавлен 17.02.2011Раскрытие сущности и характеристика основных видов кредитования населения. Общие условия и методы кредитования. Кредитная политика и анализ структуры кредитного портфеля в КФ АО "Kaspi bank". Кредитный мониторинг проблемных потребительских кредитов.
дипломная работа [312,2 K], добавлен 25.10.2015Внедрение CRM и его преимущества. Общая характеристика Сбербанка, стратегия и элементы бизнес-модели. Задекларированные высокоуровневые цели и направления развития CRM в исследуемом банке. Ожидаемые результаты реализации стратегии и критерии успеха.
дипломная работа [2,2 M], добавлен 15.01.2017Оценка основных показателей деятельности банка, величина собственного капитала, коэффициенты доходности и прибыльности АО "Kaspi bank". Анализ динамики и структуры его кредитного портфеля. Финансовые отношения банка с клиентами и расчетные операции.
курсовая работа [648,3 K], добавлен 08.12.2014The history of the development of Internet banking in Kazakhstan and abroad. Analysis of the problems faced by banks in the development of this technology. Description of statistical of its use and the dynamics of change. Security practices for users.
презентация [1,3 M], добавлен 24.05.2016Анализ и оценка финансово-экономической деятельности банка на примере ОАО "Кaspi bank". Организационно-экономическая характеристика и финансовые показатели деятельности коммерческого банка, разработка рекомендаций по его финансовому оздоровлению.
курсовая работа [1,2 M], добавлен 05.05.2015
