# Foreign Exchange Risk Management in Commercial Banks in Pakistan

Where:

Log[p/1-p] = Logit [p] =ln[p/1-p]

a = Constant

b = slope values (for independent variables 1 to 5)

X = Independent variables ( 1 to 5)

P can be calculated using the below formula, which is simply another

rearrangement of the above equation:

Where:

� = ��� ^{a+b}_{1}^{X}

1^{+b}

2^{X}2^{+b}

3^{X}

3^{+b}

4^{X}

4^{+b}

5^{X}_{5 }/ 1+ ��� ^{a+b}

1^{X}1^{+b}

2^{X}2^{+b}

3^{X}

3^{+b}4^{X}

4^{+b}

5^{X}

5

Model 4: Calculation of P using Binary Logistic Regression for Currency Derivative Usage

p = odds ratio

The Logits (Log Odds) are the b coefficients (Slope values) of the regression

equation. b coefficients (slope values) can be interpreted as the change in the Log

Odds due to a unit change in the independent variable. b for the Net Assets (in

billions rupees) can be interpreted as the change in Log Odds due to a one billion

change in net assets of a commercial bank. Similarly, the b’s for all other

independent variables can be interpreted in the same way.

p here refers to as the Odds ratio, which is very important to interpret in this

model here. Odds ratio estimate the change in odds of the membership in the

target group (which is usage of tools other than forwards in this case) for a unit

increase in the independent variable. For example, changes in odds due to a unit

change in Net Assets, i.e. one billion rupees of net assets. It can be calculated as

the exponential of the regression coefficient, b, of the relative independent

variable.

Using this model, it has to be found out that what role does above mentioned

independent variables play for the selection of tools. Selection of tools means that

whether the banks use only forward exchange contracts or use swaps and options

along with forward exchange contracts. Using SPSS, Binary Logistic Regression

is run using “Backward Stepwise based on Likelihood Ratio Test”. All the

variables are entered in the model and then using backward stepwise method

based on Likelihood ratio tests, insignificant independent variables are removed

and finally only the significant variables are retained in the model.

Page | 31

Foreign Exchange Risk Management & its Impact on Income

Foreign Exchange Risk Management practices vary from bank to bank and so

does the income of banks. An important objective of this study is to study income

from dealing in foreign currencies and do different factors affect income from

dealing in foreign currencies.

Study the descriptive of Income from dealing in foreign currencies in

both Islamic & Conventional Banks & Comparison Between them

Income from dealing in foreign currencies of commercial banks with respect to its

type i.e. conventional or Islamic is studied using descriptive analysis. For this

purpose, the variable used is Income from Dealing in foreign currencies as a

percentage of total income. Further income from dealing in foreign currencies in

absolute terms is also descriptively analyzed. The actual purpose of using the both

variables in this descriptive analysis is to check descriptive in absolute terms as

well as in relative terms to its overall income.

To compare both these groups independent sample t-test is used and the below

mentioned hypothesis are formed:

H0: There is no significant difference between IFX of conventional banks and

Islamic Banks

H1: There is a significant difference between IFX of conventional banks and

Islamic Banks

Study the descriptive of Income from dealing in foreign currencies in

both Public Sector Commercial Banks and Local Private Banks &

Compare them

Income from dealing in foreign currencies of commercial banks with respect to its

ownership status i.e. whether it is a public sector commercial bank or local private

bank, is studied using descriptive analysis. Both the variables Income from

dealing in foreign currencies in absolute terms and in relative terms to overall

income.

To compare both these groups independent sample t-test is used and the below

mentioned hypothesis are formed:

H0: There is no significant difference between IFX of public sector commercial

banks and local private banks

H1: There is a significant difference between IFX of public sector commercial

banks and Islamic Banks

Page | 32

Income from dealing in foreign currencies and size of bank

It is very important to check whether size of bank have any effect on income from

dealing in foreign currencies. Simple Linear Regression model is used to check

this relationship. Two separate models are constructed to check the effect of size

of bank on income from dealing in foreign currencies. First model check the

effect of size on Income from dealing in foreign currencies in absolute terms (in

rupees) and second one check the effect of size on the Income from dealing in

foreign currencies relative to total income.

Independent Variable: Size of bank, as measured throughout this study using Net

Assets.

Dependent Variable (Model 1): IFXRS, Income from dealing in foreign

currencies (in ‘000 rupees)

Dependent Variable (Model 2): IFX, Income from dealing in foreign currencies as

a percentage of total income of bank.

Simple linear regression models used to investigate above relationships can be

given as:

Hypothesis:

����� = � + ��� + ɛ

Model 5: Model 1 of Relationship between Net Assets & IFXRS

��� = � + ��� + ɛ

Model 6: Model 2 of Relationship between Net Assets & IFX

Following are the list of hypothesis constructed

H_{0}: There is no linear relationship between Net Assets & Income from dealing in

foreign currencies in rupees.

H_{1}: There is a linear relationship between Net Assets & Income from dealing in

foreign currencies in rupees.

H_{0}: There is no linear relationship between Net Assets & Income from dealing in

foreign currencies relative to total income.

H_{1}: There is a linear relationship between Net Assets & Income from dealing in

foreign currencies relative to total income

Effects of tools used on Income from Dealing in Foreign Currencies

Page | 33

Another objective of this study is to check whether the use of currency derivatives

have any impact on the income from dealing in foreign currencies of commercial

banks. This can be investigated using Simple Linear Regression.

Independent Variable: In this case independent variable is tools used by

commercial banks. This variable is entered into the simple linear regression model

as a dummy variable. As forwards contracts are used by all the commercial banks

of Pakistan therefore, again here two categories are formed, 0 & 1. 0 for the banks

who use only forward exchange contracts and 1 for the banks who use currency

swaps or foreign currency options or both along with these forwards exchange

contracts.

Dependent Variable: Dependent Variable in this case is Income from dealing in

foreign currencies as a percentage of total income of bank, as obtained from

Income Statement of respective commercial bank. Income from foreign currencies

is mentioned under non markup income head in income statement. This cannot be

directly used in this model due to the differences between the banks in size and

overall income, therefore, income from dealing in foreign currencies is taken as a

percentage of total income of the bank. Total Income here includes total Net

Markup Income and Markup income before deduction of any expenses. Net

markup income means total markup income earned less markup expenses as paid

on deposits. Therefore, using this variable as Income from foreign currencies as a

percentage of total income, denoted by IFX, adjusts for the interbank differences

of characteristics.

Model:

The simple linear regression model as used in this analysis is as below:

Where

��� = � + ������ + �

Model 7: Relationship between Tools used and IFX

IFX = Income from dealing in foreign currencies as a percentage

of Total income

� = Intercept of the model

β = Slope of the model, regression coefficient

Tools =

variable

Independent variable: tools used entered as a dummy

Page | 34

� = Error term

Using the above mentioned model, regression coefficient b, an unbiased estimate

of β, is calculated and its significance is checked. To check the significance,

ANOVA is used and below mentioned hypothesis is constructed;

H_{0}: There is no linear relationship between IFX & Tools

H_{1}: There is a linear relationship between IFX & Tools

ANOVA is used to check the overall significance of this regression model. R

square explains the extent of relationship as explained by the regression model.

Income from dealing in foreign currencies and Exchange Rate Volatility

Exchange Rate Volatility is a very important phenomenon in the forex market. As

commercial banks deal in foreign currencies, it is very important to check whether

commercial banks are affected by the exchange rate volatility. Simple Linear

Regression model is used to check this relationship. Two separate models are

constructed to check the effect of exchange rate volatility on income from dealing

in foreign currencies. First model check the effect of Exchange Rate Volatility on

Income from dealing in foreign currencies in absolute terms (in rupees) and

second one check the effect of Exchange Rate Volatility on the Income from

dealing in foreign currencies relative to total income.

Independent Variable: Exchange Rate Volatility, for this purpose current year

exchange rate volatility is taken into account.

Dependent Variable (Model 1): IFXRS, Income from dealing in foreign

currencies (in ‘000 rupees)

Dependent Variable (Model 2): IFX, Income from dealing in foreign currencies as

a percentage of total income of bank.

Simple linear regression models used to investigate above relationships can be

given as:

Hypothesis:

����� = � + ���� + ɛ

Model 8: Model 1 of Relationship between ERV& IFXRS

��� = � + ���� + ɛ

Model 9: Model 2 of Relationship between ERV & IFX

Following are the list of hypothesis constructed

Page | 35

H_{0}: There is no linear relationship between Exchange Rate Volatility & Income

from dealing in foreign currencies in rupees.

H_{1}: There is a linear relationship between Exchange Rate Volatility & Income

from dealing in foreign currencies in rupees.

H_{0}: There is no linear relationship between Exchange Rate Volatility & Income

from dealing in foreign currencies relative to total income.

H_{1}: There is a linear relationship between Exchange Rate Volatility & Income

from dealing in foreign currencies relative to total income

Page | 36

Page | 37

Findings & Analysis

Data obtained from Annual reports of commercial banks, statistical bulletins and

other publications by State Bank of Pakistan is analyzed using the above

methodology mentioned in the previous section. Different research methods and a

variety of statistical techniques are used for this purpose.

All the findings are mentioned in this section, along with analysis.

Page | 38

Findings on Net Foreign Currency Exposure of Commercial Banks in

Pakistan

Foreign Currency Exposure of commercial banks in Pakistan is studied and

findings relative to different research questions are below:

1. Net Foreign Currency Exposure

1. Annual reports of Commercial banks showed that majority of commercial

banks assume foreign currency exposure. All the commercial banks

mentioned their Net Foreign Currency Exposure in Notes to their financial

statements.

2. A new variable “Net Foreign Currency Exposure Relative to Net Assets”

(NFXNA) is constructed and analyzed.

a. Almost majority of the banks have positive figure which shows

that they hold overall Net Asset Position.

b. Few banks have zero as they do not take any exposure and keep

their positions offset and hedged all the time.

c. A few banks even have negative figure, which shows that they

hold overall NET LIABILITY POSITION.

Below is the table showing descriptive of the variable NFXNA:

Descriptive Statistics: Net Foreign Currency Exposure Relative to Net Assets

N Range Minimum Maximum Mean Std. Deviation

NFXNA 108 2.18049 -1.03309 1.14740 .7988934 .39340539

Valid N (listwise) 108

Table ii: NFXNA, Descriptive Statistics

Descriptive statistics as mentioned in the above Table ii, clearly show that Net

foreign currency exposure varies from bank to bank. A standard deviation of

0.3934, as obtained from the data of 108 banks for the period 2005-2009, shows

the degree of dispersion in this variable and hence it can be said that Net foreign

currency exposure of banks vary and all the banks do not have the same ratio of

NFXNA.

2. Findings on “Factors Affecting Foreign Currency Exposure

To study whether different factors as mentioned in previous section affect the Net

Foreign Currency Exposure of Commercial banks in Pakistan or not. Multiple

Linear Regression is used to check the relationship between “Net Foreign

Currency Exposure Relative to Net Assets” as a dependent variable and a set of

Page | 39

independent variables i.e. size of bank, ownership status of bank & exchange rate

volatility. To serve the purpose, below mentioned model (Model # 1) is formed:

����� = � + �_{����}���� + �_{��}�� + �_{���}��� + �

A correlation matrix to check the correlation between independent variables so

that it can be found out if there is any Multicollinearity in the above mentioned

model or not. The correlation matrix is below:

Correlations

Ownership

Status Size

Exchange

Rate Volatility NFXNA

Ownership Status Pearson Correlation 1 -.343^{** }.073 -.056

Sig. (2-tailed) .000 .448 .565

N 110 110 110 108

Size Pearson Correlation -.343^{** }1 .125 .093

Sig. (2-tailed) .000 .193 .341

N 110 110 110 108

Exchange Rate

Volatility

Pearson Correlation .073 .125 1 .142

Sig. (2-tailed) .448 .193 .142

N 110 110 110 108

NFXNA Pearson Correlation -.056 .093 .142 1

Sig. (2-tailed) .565 .341 .142

N 108 108 108 108

**. Correlation is significant at the 0.01 level (2-tailed).

Table iii: Correlation between OS, Size & ERV

It is obvious from the above correlation matrix that the correlation between any of

two independent variables is not greater than 0.5 and is far less than this

threshold. Therefore, it can be said on the basis of the above mentioned findings

that there is no problem of Multicollinearity in this model.

After checking for the Multicollinearity and finding that there is no

Multicollinearity, Multiple Linear Regression is run on the data using the above

mentioned dependents and a set of independent variables, using SPSS Statistics

Processer. The output from SPSS is below:

Page | 40