A study on relationship between tail risk on earning management in Iranian banking industry

Article history: Received July 18, 2012 Received in revised format 10 November 2012 Accepted 15 November 2012 Available online November 19 2012 Risk management plays an important role in banking industry and there are literally many investigations to reduce any risk components in this industry. In this paper, we present a study on relationship between tail risk on earning management in Iranian banking industry. In this survey, we use two series of data. The first set is associated with yearly information of 19 different banks over the period 2005-2011 and it contains 114 observations. The second set of data includes weekly historical data of eight banks over the same period 2005-2011. In this survey, there are four objectives to be investigated. The first hypothesis considers the effects of seven independent variables on loan loss allowance as a fraction of total loans. The second model is associated with the effects of two independent variables on realized gains and losses on securities. The third objective is to study the effects of different independent variables with various interruptions on return of banking sectors. Finally, the last model investigates the effects of revenue management on tail risk. The result of this survey indicates that there is no relationship between tail risk and earning management. © 2013 Growing Science Ltd. All rights reserved.


Introduction
There have been growing concerns among banks' managers as well as investors on tail risk associated with the shares of banking industry (Dimson, 1979;Dechow et al., 1995;Dechow et al., 1996;Demski, 1998;Amihud, 2002;Cornett et al. 2009;Hutton et al., 2009).Ahmed et al. (1999), for instance, reexamined capital management, earnings management, and signaling effects for bank loanloss provisions.Arya et al. (1998) investigated the relationship between earnings management and the revelation principle.Beatty et al. (1995) investigated the impact of taxes, regulatory capital, and earnings on managing financial reports of commercial banks.Beatty et al. (2002) investigated the impact of earnings management to avoid earnings declines across publicly and privately held banks.Beaver and Engel (1996) studied discretionary behavior with respect to allowances for loan losses and the behavior of security prices.Cohen et al. (2004) studied trends in earnings management and informativeness of earnings announcements in the pre-and post-Sarbanes Oxley periods.Cohen et al. (2004), in an assignment, investigated trends in earnings management and informativeness of earnings announcements in the pre-and post-Sarbanes Oxley periods.Gunther and Moore (2003) investigated loss underreporting and the auditing role of bank exams.Healy (1985) studied the effect of bonus schemes on the selection on accounting decisions.Sloan (1996) investigated whether stock prices fully reflect information in accruals and cash flows about future earnings.Scholes et al. (1990) studied tax planning, regulatory capital planning, and financial reporting strategy for commercial banks.Wahlen (1994) performed an empirical investigation on the nature of information in commercial bank loan loss disclosures.Houshmand Neghabi and Morshedian Rafiee (2013) investigated the relationship between capital structure as dependent variable and seven independent variables including tax rate, firms' growth rate, fixed assets, firms' size, operating risk, profitability and industry type.They used the financial information of 107 selected companies from 18 different industries listed on Tehran Stock Exchange over the period of 2004-2011 covering 40% of total number of companies listed in this stock exchange.They used ordinary least square technique to study the relationships.The results of the survey indicate that the there was a positive relationship between tax rate and firm's growth rate, and capital structure.The result of the survey also indicated there was a negative relationship between firm's profitability and capital structure.However, there was no evidence to believe that there was any relationship between fixed assets and capital structure.Farzinfar (2013) investigated the relationship between auditor's opinion and stock return in the companies listed at Tehran stock exchange market.In this study, all required data were collected from aware shareholders and provided a sampling of 130 questionnaires, the data collected over the period 2010-2011 using test methods such as computer software, data analysis and statistical methods to answer research questions.According to research result through questionnaires and tests, there was a significant relationship between stock returns and the auditor's opinion, in fact, for aware shareholders of the company the auditor's opinion had a special message.Sohrabi Araghi and Attari (2013) investigated the effect of accruals and operating cash flows in decisions of financial statement users in listed companies on Tehran stock exchange, information content of operating cash flows and accruals in the connection with decision-making criteria used by various groups using financial statement has been examined.They reported that there was a significant different between accruals and operating cash flows information content in relation to various decision-making criteria but utilizing accruals and operating cash flows supplementary and simultaneously in profit frame depending on the selection criteria may or may not be include information value-added.

The proposed study
The proposed study of this paper considers three models to investigate 12 hypotheses.The first model is as follows, (1) where GAINS is realized gains and losses on securities as a fraction of total assets, LNASSET is the natural log of total assets and UGAINS is the unrealized gains and losses on securities as a fraction of total assets.The second set of data considers 2496 weekly observations associated with eight banks and Table 2 demonstrates details of some basic statistics.In this survey, there are four objectives to be investigated.The first hypothesis considers the effects of seven independent variables on loan loss allowance as a fraction of total loans.The second model is associated with the effects of two independent variables on realized gains and losses on securities.
The third objective is to study the effects of different independent variables with various interruptions on return of banking sectors.Finally, the last model investigates the effects of revenue management on tail risk.We have used Hausman Test to find out whether we should use fixed or variable method and the test examines the following hypothesis,  Based on the results of three mentioned tests in Table 4, we can conclude that the data are not normally distributed.In order to use ordinary least square, we need to make sure there is no autocorrelation between residuals.Table 5 demonstrates details of our findings, As we can observe from the results of Table 5, all F-statistics are meaningful when the level of significance is one percent and we can conclude that there is a linear relationship between independent variables and dependent variable.In addition, all Durbin-Watson values are within acceptable limit, which means there is no correlation among residuals.Therefore, we can use ordinary least square to estimate the model.Finally, we need to make sure there is no correlation between each pairs of independent variables and our investigation are summarized in Table 6.It is obvious from the results of Table 6 that there is no strong correlation among each pairs of statistics and we may use all data for regression model.

The results
In this section, we present details of our survey on different main hypotheses of the survey.The main hypothesis of this survey is associated with the relationship between tail risk on earning management in Iranian banking industry, which is as follows, Main hypothesis: There is meaningful relationship tail risk and earning management in Iranian banking industry.0 1 : 0There is no relationship between tail risk and earning management.: 0 There is some relationship between tail risk and earning management.
Table 7 shows details of our findings of our regression analysis.

Table 7
The relationship between tail risk and earning management As we can observe from the results of Table 7, t-student is not statistically significant and we cannot reject the null hypothesis.There are also eleven sub hypotheses investigated in this survey.

The hypotheses associated with loan loss provisions as a fraction of total loans
The first seven hypotheses are associated with Eq. ( 1) as follows, : , , , , , , 0 : , , , , , , 0 The results of regression analysis are summarized in Table 8 where F-value and Durbin-Watson are 13.93262 and 1.559447, respectively.

The relationship between LNASSET and LOSS
The first sub-hypothesis is associated with the relationship of natural log of total assets with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we can conclude that there is a meaningful relationship between these two variables when the level of significance is five percent.In other words, when all other conditions are remained constant, an increase of one unit in log of total assets will yield an increase of 0.065524 in loan loss provisions as a fraction of total loans.

The relationship between NLP and LOSS
The second sub-hypothesis is associated with the relationship of nonperforming loans with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we cannot conclude that there is any meaningful relationship between these two variables when the level of significance is five or even ten percent.

The relationship between LLR and LOSS
The third sub-hypothesis is associated with the relationship of loan loss allowance with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we can conclude that there is a meaningful relationship between these two variables when the level of significance is five percent.In other words, when all other conditions are remained constant, an increase of one unit in log of total assets will yield an increase of 0.318997 in loan loss provisions as a fraction of total loans.

The relationship between LOANR and LOSS
The fourth sub-hypothesis is associated with the relationship of real estate loans with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we cannot conclude that there is any meaningful relationship between these two variables when the level of significance is five or even ten percent.

The relationship between LOANC and LOSS
The fifth sub-hypothesis is associated with the relationship of commercial and industrial loans with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we cannot conclude that there is any meaningful relationship between these two variables when the level of significance is five or even ten percent.

The relationship between LOANA and LOSS
The sixth sub-hypothesis is associated with the relationship of agriculture loans with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we can conclude that there is a meaningful relationship between these two variables when the level of significance is five percent.In other words, when all other conditions are remained constant, an increase of one unit in agriculture loans will yield an increase of 0.277374 in loan loss provisions as a fraction of total loans.

The relationship between LOANI and LOSS
The seventh sub-hypothesis is associated with the relationship of consumer loans with loan loss provisions as a fraction of total loans.Based on the results of Table 8, we can conclude that there is a meaningful relationship between these two variables when the level of significance is five percent.In other words, when all other conditions are remained constant, an increase of one unit in consumer loans will yield an increase of 0.270105 in loan loss provisions as a fraction of total loans.

The second set of hypotheses
The second hypothesis of this survey is associated with the relationship of realized gains and losses on securities as a fraction of total assets (GAINS) as dependent variable with the natural log of total assets (LNASSET) and the unrealized gains and losses on securities as a fraction of total assets (UGAINS) as independent variables given in Eq. ( 2).Table 9 demonstrates the results of our survey,

The relationship between LNASSET and GAINS
The first sub-hypothesis is associated with the relationship of realized gains and losses on securities as a fraction of total assets (GAINS), as dependent variable, with the natural log of total assets (LNASSET).Based on the results of Table 9, we can conclude that there is a meaningful relationship between these two variables when the level of significance is five percent.In other words, when all other conditions are remained constant, an increase of one unit in the natural log of total assets will yield an increase of 1162.681 unit in realized gains and losses on securities as a fraction of total assets.

The relationship between UGAINS and GAINS
The second sub-hypothesis is associated with the relationship between realized gains and losses on securities as a fraction of total assets (GAINS), as dependent variable, with the unrealized gains and losses (UGAINS).Based on the results of Table 9, we cannot conclude that there is any meaningful relationship between these two variables when the level of significance is five percent.

The third set of hypotheses
The third hypothesis of this survey is associated with the relationship of the return of a particular bank with the return of market and other banks.: , , , , , , , , , 0 : , , , , , , , , , 0 Table 10 summarizes the results of our survey when F-value and Durbin-Watson are 41.56312 and 1.807640, respectively.The regression Eq. (3) has been fitted using the information of eight different banks with 52 intervals of data and the results of regression analysis are shown in Table 10.Based on the results of Table 10, two hypotheses are investigated.

The relationship between return of a bank with market return
Based on the results of Table 10 we can observe that t-student value associated with RM(-1) is 0.258440, which is not meaningful and we cannot reject the null hypothesis.

The relationship between return of a bank with banking industry
Based on the results of Table 10 we can observe that t-student is not meaningful and we cannot reject the null hypothesis when the level of significance is five percent but the hypothesis is rejected when the level of significance is ten percent.

Conclusion
In this paper, we have investigated the relationship between tail risk on earning management in Iranian banking industry.There were four objectives to be investigated.The first hypothesis considered the effects of seven independent variables on loan loss allowance as a fraction of total loans.The second model was associated with the effects of two independent variables on realized gains and losses on securities.The third objective was to study the effects of different independent variables with various interruptions on return of banking sectors.Finally, the last model investigated the effects of revenue management on tail risk.The result of this survey indicates that there was no relationship between tail risk and earning management.In summary, Table 11 summarizes details of our findings for different components of this survey.
is the return of bank j in week t, mt r is the CRSP value-weighted market return and jt r the stock market return of bank j in week t.In this survey, we use two series of data.The first set is associated with yearly information of 19 different banks over the period 2005-2011 and it contains 114 observations.The second set of data includes weekly historical data of eight banks over the same period of 2005-2011.Table 1 shows details of different statistics associated with various variables of our study.

Table 2
Descriptive Statistics for the second series of data

Table 3
The results of Chow and Hausman testsBefore we use regression analysis, we need to make sure that the data are normally distributed.The results of the implementation of Kolmogorev-Smirnov, Shapiro-Wilk and Jarque-bera test are summarized in Table4as follows,

Table 5
The results of statistical tests

Table 6
The summary of correlation test

Table 8
The summary of regression analysis

Table 9
The summary of regression analysis for the second model

Table 10
The summary of regression analysis for the third model