Sunteți pe pagina 1din 8

Journal of Money, Investment and Banking

ISSN 1450-288X Issue 17 (2010)


© EuroJournals Publishing, Inc. 2010
http://www.eurojournals.com/JMIB.htm

Distributional Characteristics of Ratios:


Evidence from Turkish Banking Sector

Mahmut Karğın
Accounting and Finance Department, Celal Bayar University
Uncubozköy, Manisa, Turkey
E-mail: mahmut.kargin@bayar.edu.tr

Hüseyin Aktaş
Accounting and Finance Department, Celal Bayar University
Uncubozköy, Manisa, Turkey
E-mail: huseyin.aktas@bayar.edu.tr

Koray Kayalıdere
Accounting and Finance Department, Celal Bayar University
Uncubozköy, Manisa, Turkey
E-mail: koray.kayalidere@bayar.edu.tr

Abstract

This study investigates the distributional characteristics and appropriate remedial


actions of selected ratios from Turkish banking sector. A total of 24 listed banks with 21 of
6 group ratios were examined for the period from 2007 to 2009. Normality test was carried
out to the data using Shapiro-Wilk test. The findings for three years demonstrate that five
ratios conformed to non-normal distribution. However, significant improvement on
normality was observed after remedial actions (transformations). We found that the F2 ratio
(Non-Operating Expense / Total Operating Income) was generally normally distributed in
all years but the F1 ratio (Interest Income / Total Income) was generally not normally
distributed at the significance level of 0.01. Transformation process was carried out using
four techniques namely cube root, inverse, natural log, and square root. All of these
transformation techniques reduced non-normality. The cube root and square root
transformation techniques slightly outperform the others.

Keywords: Financial Ratios, Normality, Distributional Properties, Transformation


Techniques, Banking Sector.
JEL Classification: M40, M41, M49

1. Introduction
The use of financial ratios in a variety of performance evaluation and decision-making process is an
important area of accounting and finance (Frecka and Hopwood, 1983). The researchers indicate well
known samples of the applications: financial distress prediction (Altman, 1968); bond rating (Pinches
and Mingo, 1973); commercial credit scoring (Ewert, 1977); security analysis (Bernhard, 1979); and
audit evaluation (Altman and McGough, 1974). One more important empirical development is
Journal of Money, Investment and Banking - Issue 17 (2010) 48

Beaver’s (1967) study, accepted as a landmark of ratio studies that attempt to investigate ability of
ratios to predict the failure of firms (Horrigan, 1968).
Adoption of financial ratios as a tool of financial statement analysis is assumed relatively recent
development in accounting and finance literature by Horrigan in 1968 (Horrigan, 1968). Since last half
of the Nineteenth century, financial ratios have been commonly used by financial statement users to
evaluate the performance of a firm with its competitors and also to assess the firm’s progress from one
accounting period to the next (McLeay and Stevenson, 2009). Over the years, many studies have
repeatedly showed the usefulness of financial ratios. For example, financially-distressed firms can be
separated from the non-failed firms in the year before the bankruptcy at an accuracy rate of better than
90% by analyzing ratios (Chen and Shimerda, 1981, p. 51).
Financial statement users (i.e., managers, shareholders, investors, and creditors) use financial
ratios in their decision-making process. An important point in this issue is whether the simple ratio
metric that is commonly used as a financial indicator provides an adequate measure for the financial
statements users in order to make rational decisions. Statistical information about financial ratios can
be more useful for decision makers if ratios meet certain statistical criteria such as if they are
distributed normally. In accounting and finance literature ratios are assumed to be normally distributed.
In particular, we ask in this study whether financial ratios are normally distributed and how this
characteristic of ratios can be improved by taking proper remedial actions.

2. Review of Literature
Several studies have examined possible distributional forms for ratios (e.g., Horrigan, 1965; Deakin,
1976; Frecka and Hopwood, 1983; Ricketts and Stover, 1978; Ezzamel et al., 1987). Horrigan (1965)
suggested that financial ratios tend to be approximately normal but positively skewed.
The results indicating non-normality of financial ratio distributions have led researchers and
practitioners into looking for methods of restoring normality to warrant standard parametric statistical
analyses (Salmi and Martikainen, 1994).
Deakin (1976) concludes that an assumption of normality cannot be supported from his work
on selected firms. Transformation techniques, square root and lognormal techniques reduced non-
normality. However, Frecka and Hopwood (1983) examined large sample of manufacturing firms for
the period of 1950-1979 by 11 ratios that Deakin (1976) analyzed. They found that outliers had
tremendous influence on the distribution of normality and after deleting outliers normality or
approximate normality was achieved.
Ricketts and Stover (1978) interpret their study’s results as supporting the normality assumption
in the selected data. Ezzamel et al. (1987) investigated the distributional properties of financial ratios of
selected firms for the period of 1980 to 1981 by using the Kolgomorov-Smirnov and the Shapiro-Wilk
tests for the normality. Their results were consistent with literature and they found that the square root
transformation technique outperformed logarithmic technique for reaching normality.
A review of the literature describing models and theories for evaluating and predicting financial
performance reveals that although models have become increasingly complex, few studies adequately
focuses on the problems associated with the sample used. For instant, most studies use multivariate
analysis that is based on the assumption of a normal distribution of ratios. Without confirming the
approximation of normality of ratio distribution, the researcher may not eliminate the risk of drawing
erroneous inferences (Nenide et al., p. 4). Most of well known researchers in the literature and their
work, such as Beaver (1966), Altman (1981), Ohlso (1980), Zavgren (1985), Deakin (1972), and Blum
(1974) are criticized for using samples that have not been checked for normal distribution or for the
removal of errors (Nenide et al., p. 7).
Ashton et al. (2004), note that other studies in this area have been conducted by McLeay (1986
and 1997), Ezzamel and Mar-Malinero (1990), Kolari, McInish and Saniga (1989) and Lau, Lau and
Gribbin (1995). Common approach of these works is the use of statistical distributions that have
49 Journal of Money, Investment and Banking - Issue 17 (2010)

unbounded moments.
Understanding firms’ financial performance by ratios has motivated a significant number of
research efforts in the area of financial statement analysis in banking. In those studies banking data are
usually assumed to have a normal distribution. However, a number of studies show that the distribution
of many financial ratios departs significantly from normality because of the presence of skewness and
extreme outliers (Bedingfield et al., 1985, p. 77).
Most of researchers and practitioners have continued to use untransformed ratios. Why do
researchers still use untransformed financial ratios? “The continued and pervasive use of
untransformed, accounting-based financial ratios despite the associated problems, suggest that (1) the
convenience of using ratios may exceed the cost of errors in analysis caused by ratio-related model
mis-specification, and/or (2) in general, no equally convenient, or superior alternative to transformed
ratios has been developed and applied to financial ratio analysis” (Kane and Meade, 1998, p. 59).
In this study we check normality of selected ratios from Turkish banking sector and investigate
appropriate remedial techniques (transformation techniques).

3. Data and Methodology


When performing ratios analysis, the following problems need to be accounted for and addressed in the
research design if the results are to be meaningful and useful to researchers (Nenide et al p. 9): Data
entry errors, negative denominators, outlier influence, and normality of distribution. First of all
researchers need to assume that data entry process from financial statements is not error free and it is
essential that database must be corrected before the analysis. Second of all, in case of negative
denominators, investigators need to make a decision as to whether this represents a reasonable part of
the financial distribution that should be included or it is abnormal and need to be excluded. Similarly,
analysts must consider that whether outlier represents actual performance of firm or an extreme
situation. If it is believed that outlier does not represent the actual performance it must be excluded.
Finally researchers must check normality of distribution. Even with the elimination of data
entry and negative denominators problems, the sample may have extreme observations that distort
mean and standard deviation calculations. For these reason, the data must be transformed to achieve
normality parameters if parametric statistics are to be used (Nenide et al, p. 15).

3.1. Ratios
Chen and Shimerda (1981) state that there are many useful ratios reported in the literature and different
researchers often include different ratios. As a result, critics on the usefulness of specific ratios vary.
The financial ratios examined in this study are 21 of the 6 group of ratios that have most frequently
been used by The Banks Association of Turkey.
The ratios are obtained from Finnet Data Base cover 6 major field of interest including Capital
Adequacy, Balance Sheet Structure, Quality of Assets, Liquidity, Profitability, and Income –
Expenditure Structure presented in Table 1.
Journal of Money, Investment and Banking - Issue 17 (2010) 50
Table 1: Financial Ratios Used

Group of Ratios
Capital Adequacy
Equity / ( Loans + Market+ Amount Based on Operational Risk) A1
Equity / Total Assets A2
(Equity –Fixed Assets) / Total Assets A3
Balance Sheet Structure
Assets in Turkish Lira / Total Assets B1
Liabilities in Turkish Lira / Total Liabilities B2
Deposits in Turkish Lira / Total Deposits B3
Total Deposits / Total Assets B4
Quality of Assets
Financial Assets (Net) / Total Assets C1
Total Loans and Receivables / Total Assets C2
Fixed Assets / Total Assets C3
Liquidity
Liquid Assets / Total Assets D1
Liquid Assets / Short-term Liabilities D2
Liquid Assets in Turkish Lira / Total Assets D3
Liquid Assets / (Deposits + Non-Deposits Funds) D4
Liquid Assets in Foreign Currency / Liabilities in Foreign Currency D5
Profitability
Net Profit (Loss) / Total Assets E1
Net Profit (Loss) / Total Equity E2
Net Profit (Loss) / Paid in Capital E3
Income – Expenditure Structure
Interest Income / Total Income F1
Non-operating Expense / Total Operating Income F2
Interest Income / Interest Expense F3

3.2. Methodology
In this study we selected 21 ratios of 24 banks out of 25 listed on Turkish banking sector from 2007 to
2009. After screening data, only one bank is omitted from the list because of inadequate data.
Distributional characteristics of 21 ratios of 24 banks are investigated. To improve distributional
characteristics of raw ratios, four transformation techniques are applied for all observations for each
year. Normality test was carried out to the data using Shapiro-Wilk test, which was probably used in
the field of ratio analysis for the first time by Bird and McHugh (1977) for testing normality of small
samples.

3.3. Transformation
As stated, many researchers and practitioners have used transformation techniques to reduce non-
normality. This study conducts transformations techniques similar to Buijink and Jegers (1986). The
researchers used four transformation techniques in their study; cube root (CBR(X)), inverse (1/X),
natural log (lnX), and square root (SQR(X)). They note that transformation is obvious way to improve
the normality. However, appropriate technique of transformation need to be applied. For example,
while cube root transformation is applicable for all observations, the logarithmic, inverse, and square
root techniques could be only performed for ratios that are not negative. In the current study to improve
the normality of a ratio distribution, selected transformation techniques are applied to all observations.

4. Results
For each ratio in the sector in each year, the distribution was tested for normality using the Shapiro-
Wilk test. Horrigan (1968) has suggested that financial ratios tend to be approximately normal. We
51 Journal of Money, Investment and Banking - Issue 17 (2010)

found that only F2 ratio is normally distributed in three years. What could be said about the other ratios
and details of all tests are given under each Table.

Table 2: 2009 Results

Transformed Ratios
Raw Ratios
Shapiro-Wilk Statistic
Shapiro-
Skewness Kurtosis Cube Square
N Std. Dev. Wilk Inv. Log
Statistic Statistic root. root
Statistic
A1 24 34.48 4.55 21.49 .360 .543 .926* .660 .489
A2 24 15.01 4.19 19.03 .470 .683 .975* .794 .626
A3 24 14.48 4.20 19.16 .477 .747 .980* .874 .675
B1 24 10.85 -0.21 0.27 .976* .964* .916* .955* .968*
B2 24 11.36 0.80 2.35 .951* .970* .944* .972* .967*
B3 24 17.58 -1.00 3.31 .889* .733 .353 .624 .782
B4 24 16.94 -1.59 2.14 .824 .726 .519 .672 .753
C1 24 17.64 1.18 1.36 .903* .973* .910* .982* .961*
C2 24 20.24 -1.04 0.68 .910* .667 .395 .691 .861
C3 24 1.78 0.05 -1.19 .949* .945* .815 .927* .949*
D1 24 21.46 1.54 2.15 .833 .945* .813 .963* .924*
D2 24 133.28 4.66 22.37 .360 .665 .857 .847 .569
D3 24 21.75 2.16 4.67 .742 .930* .726 .966* .891*
D4 24 135.22 4.69 22.53 .338 .635 .817 .831 .537
D5 24 46.26 4.00 17.60 .499 .777 .969* .900* .704
E1 24 1.15 1.03 3.03 .921* .928* .469 .832 .951*
E2 24 8.84 0.37 -0.30 .967* .940* .556 .864 .962*
E3 24 54.63 1.75 3.51 .813 .972* .577 .932* .953*
F1 24 14.66 -3.29 14.09 .651 .517 .320 .456 .550
F2 24 15.66 0.70 -0.33 .934* .962* .969* .970* .956*
F3 24 212.78 4.11 17.98 .434 .594 .894* .686 .549
*Significant at the 0.01 level.

Some of the raw ratios (A1, A2, A3, D1, D3, D5, E3, and F3) are not normally distributed but
they can be improved by the proper transformation technique(s) as reported in Table 2. After remedial
actions assumption of normality is not rejected. However, B4, D2, D4, and F1 raw ratios are not
normally distributed and improvement by selected transformation techniques is not possible.
Assumption of normal distribution is rejected for those ratios.

Table 3: 2008 Results

Transformed Ratios
Raw Ratios
Shapiro-Wilk Statistic
Shapiro-
Skewness Kurtosis Cube Square
N Std. Dev. Wilk Inv. Log
Statistic Statistic root. root
Statistic
A1 24 34.86 4.59 21.81 .344 .508 .833 .609 .460
A2 24 16.94 3.26 11.42 .549 .727 .969* .815 .680
A3 24 15.31 3.67 15.22 .550 .840 .788 .943* .768
B1 24 11.30 0.10 0.31 .982* .980* .941* .975* .982*
B2 24 11.80 1.26 3.15 .899* .936* .941* .946* .928*
B3 24 16.90 -1.34 5.38 .824 .614 .259 .484 .678
B4 24 17.23 -1.36 2.24 .874 .743 .478 .668 .780
C1 24 16.69 1.55 2.77 .865 .969* .889* .985* .950*
C2 24 17.26 -1.43 1.73 .845 .571 .734 .796 .850
C3 24 3.66 3.48 14.41 .616 .855 .961* .939* .800
D1 24 18.51 1.91 4.12 .803 .919* .949* .956* .895*
D2 24 138.11 4.63 22.13 .363 .655 .953* .834 .564
Journal of Money, Investment and Banking - Issue 17 (2010) 52
D3 24 18.60 2.69 9.00 .714 .932* .454 .918* .895*
D4 24 133.53 4.66 22.29 .336 .598 .968* .778 .513
D5 24 38.81 3.65 15.00 .551 .775 .992* .873 .719
E1 24 1.26 1.94 4.57 .782 .922* .419 .861 .912*
E2 24 6.70 0.47 0.80 .968* .916* .390 .791 .953*
E3 24 49.37 3.34 13.30 .625 .962* .436 .941* .902*
F1 24 11.25 -0.10 6.79 .803 .782 .671 .762 .790
F2 24 12.84 0.47 0.38 .974* .986* .953* .984* .985*
F3 24 207.98 4.54 21.33 .363 .515 .869 .612 .471
*Significant at the 0.01 level.

Non-normality characteristics of nine raw ratios in Table 3, (A2, A3, C3, D2, D3, D4, D5, E1,
and E3) can be reduced by at least one of the transformation techniques. And assumption of normal
distribution is not rejected after transformation. However, six raw ratios that are non-normally
distributed (A1, B3, B4, C2, F1, and F3) can not be improved by any selected transformation
techniques. Assumption of normality is rejected for those ratios after transformation.

Table 4: 2007 Results

Transformed Ratios
Raw Ratios
Shapiro-Wilk Statistic
Shapiro-
Skewness Kurtosis Cube Square
N Std. Dev. Wilk Inv. Log
Statistic Statistic root. root
Statistic
A1 24 33.29 4.55 21.46 .366 .560 .901* .677 .503
A2 24 16.24 3.60 13.89 .499 .677 .959* .772 .629
A3 24 14.61 4.13 18.59 .487 .804 .601 .909* .721
B1 24 11.44 -0.09 1.52 .963* .944* .844 .927* .951*
B2 24 13.16 0.41 2.30 .943* .936* .798 .916* .942*
B3 24 16.55 -0.15 1.83 .957* .887* .556 .817 .913*
B4 24 16.44 -1.09 1.44 .910* .816 .557 .755 .843
C1 24 15.49 0.66 -0.25 .950* .922* .213 .605 .976*
C2 24 19.15 -1.04 0.58 .899* .636 .756 .855 .889*
C3 24 4.34 2.46 6.97 .722 .912* .923* .971* .870
D1 24 18.42 1.60 2.97 .850 .939* .755 .947* .924*
D2 24 129.23 4.68 22.48 .355 .636 .681 .801 .548
D3 24 18.78 2.30 6.92 .771 .947* .652 .943* .920*
D4 24 128.44 4.70 22.57 .336 .602 .832 .782 .515
D5 24 31.49 2.23 6.51 .789 .952* .864 .982* .922*
E1 24 1.22 0.71 1.19 .960* .944* .426 .833 .972*
E2 24 9.14 0.20 -0.39 .969* .916* .448 .820 .946*
E3 24 97.64 4.14 18.75 .486 .915* .497 .944* .814
F1 24 12.18 -2.40 7.95 .768 .685 .508 .640 .707
F2 24 16.70 0.31 -0.65 .944* .947* .910* .943* .948*
F3 24 206.31 4.76 23.05 .321 .478 .748 .587 .429
*Significant at the 0.01 level.

Eight raw ratios in Table 4 (A1, A2, A3, C3, D1, D3, D5, and E3) are non-normally distributed
but non-normality can be reduced at least by one of the transformation techniques. However, any
remedial action cannot make improvement on four raw ratios that are not normally distributed (D2, D4,
F1, and F3). Normality assumption is obviously rejected for those ratios.
When three years are considered, F1 ratio is non-normally distributed and any of transformation
techniques cannot reduce the non-normality. Raw ratios A2, A3, D3, D5, and E3 are not normally
distributed but at least one of the selected transformation techniques can improve the normality. It
seems that the F2 ratio was generally normally distributed in all years but the F1 ratio was generally
not normally distributed at the significance level of 0.01. Assumption of normality for F1 ratio is
53 Journal of Money, Investment and Banking - Issue 17 (2010)

rejected.
As presented in Table 5, percentage of normality varies by the transformation technique used.
For example, cube root and square root techniques slightly outperform the other techniques. Generally
it can be said that selected transformation techniques increase the percentage of normality except
inverse technique in 2007.

Table 5: Comparison of Normal Distributions Percentages of Raw and Transformed Ratios

N 2009 2008 2007


Raw ratios 24 0.38 0.17 0.38
Cube root 24 0.42 0.38 0.50
Inverse 24 0.38 0.42 0.17
Log 24 0.38 0.38 0.38
Square root 24 0.42 0.38 0.46

5. Conclusions
Statistical information about financial ratios can be more useful for decision makers if ratios meet
certain statistical criteria such as if they are normally distributed. In area of accounting and finance,
ratios are assumed to be normally distributed. In case of non-normality most of researchers and
practitioners have used transformation techniques to improve normality.
Particularly, we ask in this study whether a total of 24 banks’ 21 financial ratios of Turkish
banking sector are normally distributed and how this characteristic of ratios can be improved by taking
proper remedial actions namely cube root, inverse, natural log, and square root. Normality test was
carried out to the data using Shapiro-Wilk test.
The findings for three years demonstrate that some ratios (A2, A3, D3, D5, and E3) conformed
to non-normal distribution. However, significant improvement on normality was observed after
remedial actions (transformations) so the assumption of normality is not rejected.
Specifically, we found that the F2 ratio (Non-Operating Expense / Total Operating Income) was
generally normally distributed in all years but the F1 ratio (Interest Income / Total Income) was
generally not normally distributed at the significance level of 0.01.
It can be concluded that the cube root and square root transformation techniques slightly
outperform the other techniques in reducing non-normality.
Journal of Money, Investment and Banking - Issue 17 (2010) 54

Referrences
[1] Ashton, D., P. Dunmore and M. Tippett, 2004. “Double Entry Bookkeeping and the
Distributional Properties of a Firm’s Financial Ratios”, Journal of Business Finance and
Accounting, June/July, pp. 583-606.
[2] Bedingfield, J.P., P.M.J. Reckers and A.J. Stagliano, 1985. “Distributions of Financial Ratios
in the Commercial Banking Industry”, The Journal of Financial Research, Vol. VIII, No. 1,
Spring, pp. 77-81.
[3] Bird, R.G. and A.J. McHugh, 1977. “Financial Ratios – An Empirical Study”, Journal of
Business Finance and Accounting, 4 (1), pp. 29-46.
[4] Buijink, W. and M. Jegers, 1986. “Cross-Sectional Distributional Properties of Financial
Ratios in Belgian Manufacturing Industries: Aggregation Effects and Persistence Over Time”,
Journal of Business Finance and Accounting, 13 (3), Autumn, pp. 337-363.
[5] Chen, K.H. and T.A. Shimerda, 1981. “An Empirical Analysis of Useful Financial Ratios”,
Financial Management, Spring, pp. 51-60.
[6] Deakin, E., 1976. “Distributions of Financial Accounting Ratios: Some Empirical Evidence”,
The Accounting Review, January, pp. 90-95.
[7] Ezzamel, M., C. Mar-Molinero and A. Beecher, 1987. “On the Distributional Properties of
Financial Ratios”, Journal of Business Finance and Accounting, 14 (4), Winter, pp. 463-481.
[8] Finnet Data Base, http://www.finnet.com.tr
[9] Frecka, T.J. and W.S. Hopwood, 1983. “The Effects of Outliers on the Cross-Sectional
Distributional Properties of Financial Ratios”, The Accounting Review, Vol. LVIII, No. 1,
January, pp. 115-128.
[10] Horrigan, J.O., 1965. “Some Empirical Bases of Financial Ratio Analysis”, The Accounting
Review, Vol. XL, July, pp. 558-568.
[11] Horrigan, J.O., 1968. “A Short History of Financial Ratio Analysis”, The Accounting Review,
April, pp. 284-294.
[12] Kane, GD. and N.L. Meade, 1998. “Ratio Analysis Using Rank Transformation”, Review of
Quantitative Finance and Accounting, 10, pp. 59-74.
[13] McLeay, S. and M. Stevenson, 2009. “Modelling the Longitudinal Properties of Financial
Ratios”, Applied Financial Economics, 19, pp. 305-318.
[14] Nenide, B., R.W Pricer and S.M. Camp, 2010, “The Use of Financial Ratios for Research:
Problems Associated with and Recommendations for Using Large Databases”, Retrieved 5
January 2010, http://www.fintel.us
[15] Ricketts, D. and R. Stover, 1978. “An Examination of Commercial Bank Financial Ratios”,
Journal of Bank Research, 9, Summer, pp. 121-124.
[16] Salmi, T. and T. Martikainen, 1994. “A Review of the Theoretical and Empirical Basis of
Financial Ratio Analysis”, The Finnish Journal of Business Economics, 4, pp. 426-448.
[17] The Banks Association of Turkey, http://www.tbb.org.tr

S-ar putea să vă placă și