You are on page 1of 29

Serial no: 36 FRAM assignment

Project Report
on
Time Series Forecasting
of
Given Instrument

BY

MOHD RAHIL RAHMAN


2017A4PS1633H

Under the supervision of


DR. THOTA NAGARAJU

Submitted
In partial fulfilment of
The requirements of the course of

Financial Risk and Management (FIN F414)

Allotted companies:
L&T FINANCE HOLDINGS
LIC HOUSING FINANCE
Serial no: 36 FRAM assignment

LIC HOUSING FINANCE

It was incorporated in the year 1989 and is one of the largest housing finance companies of India
solely owned by the government of India. Its key aim is to provide loans or finance to individuals
interested in construction, purchase of house/flat for residential purposes in India. It also provides
finance on existing property for purchase of equipment or other requirements for the office, clinic,
hospitals etc. Also provides finance to builders and contractors.

The company went public in the year 1994 and is actively traded on NSE. It has served about 25
lakh customers till date, with 284 marketing offices and 24000 marketing intermediates.

It has a cumulative disbursement of rupees 3.35 lakh crore since inception and more than 2 lakh
crores is the loan book.

DAILY ANALYSIS OF LICHSGFIN

1. ESTIMATION OF BETA OF THE STOCK USING CAPM


The CAPM model can be formulated

E(R) = Rf + Beta (Rm-Rf)

E(R) is the expected return of the firm

Rf is the risk- free rate

Rm is market return

Beta is estimated by regression. By assuming returns of the firm as dependent variable y and returns
of the market as the independent variable x. The slope obtained from this regression is the beta of
the CAPM.

In this evaluation we considered time from 1st November 2017 to 28 th October 2020. The return of
instrument Axis bank returns are to be calculated. Here for finding beta we regressed our respective
firm returns with nifty return.

Plot of daily closing price:


Serial no: 36 FRAM assignment

The stock price has been steady until covid crisis.


Serial no: 36 FRAM assignment

Plot of daily returns

Regression analysis:

P value is less than 0.05 indicates 95% confidence interval. The beta of the stock is 1.314 for daily
frequency, implying its quite a risky asset. 1% increase in market results in 1.314% change in the
Serial no: 36 FRAM assignment

stock price. The intercept is quite low indicating the performance of stock is quite dependent on the
market.

ESTIMATING AR AND MA COEFFICIENTS

The AR and MA coefficients are obtained from ACF and PACF plots. By running plots, we can
determine the order of the AR and MA terms.

To check the stationarity of the series we do augmented dickey-fuller test.

A series is said to be weakly stationary if: -

The mean E(yt) is the same for all time t.

The variance of yt is the same for all time t.

The covariance and the correlation between yt  and yt−1  are the same for all t.

ADF test summary

The testing p-value is 0.01 hence we fail to accept null hypothesis, therefore it is stationary.

ACF PLOT FOR RETURNS

Highly significant at the beginning and tending to zero at the end. Hence it is MA(0) model.

PACF PLOT FOR DAILY


Serial no: 36 FRAM assignment

Less significant at the beginning but keeps increasing as the lag increases. Therefore it’s a AR(1)
model.

Now ARIMA(p,q,r) model is to be estimated by varying the p,q,r and checking for what values the aic
values are minimum. In this case aic is minimum for ARIMA (1,0,1).
Serial no: 36 FRAM assignment

ARIMA PLOTS

Future stock price can be predicted using ARIMA model and the predicted values are given above.
The p-values are more than 0.05 hence the test is very much reliable. The ACF is significant for the
first lags and then insignificant. The residuals are normally distributed.

WEEKLY ANALYSIS

CAPM FOR BETA:

The CAPM model can be formulated

E(R) = Rf + Beta (Rm-Rf)

E(R) is the expected return of the firm

Rf is the risk- free rate

Rm is market return

Beta is estimated by regression. By assuming returns of the firm as dependent variable y and returns
of the market as the independent variable x. The slope obtained from this regression is the beta of
the CAPM.
Serial no: 36 FRAM assignment

In this evaluation we considered time from 1st November 2017 to 28 th October 2020. The return of
instrument Axis bank returns are to be calculated. Here for finding beta we regressed our respective
firm returns with nifty return.

Plots of closing price and returns:

The stock has been significantly decreasing with its own ups and went too low during covid crisis.

Regression summary:

The beta for weekly frequency is 1.31 with a very low intercept. This indicates the stock is very much
dependent on markets performance, intuitive to say so since it’s a financial institution. P- value is
also less than 0.05 indicating there’s less than 5% margin of error.

AR AND MA COEFFICIENTS AND ARIMA MODEL


Serial no: 36 FRAM assignment

ADF TEST:

P<0.01 indicates we reject to accept null hypothesis indicating it is stationary

ACF AND PACF PLOTS:

ACF is significant at beginning and tending to


zero as lag increases indicating an MA(0) model.

PACF is increasing as lag increases indicating an AR(1) model

The ARIMA(p,q,r) model is now ARIMA(1,0,1)

Test Plots:
Serial no: 36 FRAM assignment

P values are greater than 0.05 indicating the residuals are independent and since ACF is quite
significant beginning and tending to zero the testing is aptly accurate.

The future stock prices of weekly frequency could be predicted using ARIMA.
Serial no: 36 FRAM assignment

MONTHLY ANALYSIS:

BETA USING CAPM

CAPM FOR BETA:

The CAPM model can be formulated

E(R) = Rf + Beta (Rm-Rf)

E(R) is the expected return of the firm

Rf is the risk- free rate

Rm is market return

Beta is estimated by regression. By assuming returns of the firm as dependent variable y and returns
of the market as the independent variable x. The slope obtained from this regression is the beta of
the CAPM.

STOCK PRICE AND RETURNS PLOT:

REGRESSION SUMMARY:
Serial no: 36 FRAM assignment

P value less than 0.05 implies there is less than 5% error in the regression. The beta is 1.422
indicating it is very much dependent on market, 1% rise is market will result in 1.42% rise in the
stock price.

AR AND MA COEFFICIENTS AND ARIMA:

ADF TEST:

P< 0.05 indicating the returns are stationary.

PACF AND ACF PLOTS:

The pacf plot indicates it is AR(1) model and acf plot indicates MA(0) model.

It is an ARIMA(1,0.1) model.

DIAGNOSTIC PLOTS AND PREDICTION:


Serial no: 36 FRAM assignment

The p values suggest residuals to be highly independent but since residuals hence the test is quite
accurate

The future stock values can be predicted using the ARIMA model
Serial no: 36 FRAM assignment

L&T FINANCE HOLDINGS LTD:

L&T Finance Holdings Limited is a holding company. The Company has a range of financial products
and services across the corporate, retail and infrastructure finance sectors through its completely
owned subsidiaries. It functions primarily in the business of investment activity. Its product
diversifies across four business groups includes retail finance, which includes construction
equipment finance, transportation equipment finance, rural products finance, rural enterprise
finance, micro finance and financial products marketing, corporate finance, which includes corporate
loans and leases and supply chain finance; infrastructure finance, which includes project finance and
corporate loans, financial consultancy services and equity investments, and investment
management.
Serial no: 36 FRAM assignment

DAILY ANALYSIS:

BETA USING CAPM:

BETA USING CAPM

CAPM FOR BETA:

The CAPM model can be formulated

E(R) = Rf + Beta (Rm-Rf)

E(R) is the expected return of the firm

Rf is the risk- free rate

Rm is market return
Serial no: 36 FRAM assignment

Beta is estimated by regression. By assuming returns of the firm as dependent variable y and returns
of the market as the independent variable x. The slope obtained from this regression is the beta of
the CAPM.

STOCK PRICE AND RETURNS PLOT:

REGRESSION SUMMARY:

The beta of 1.59 indicates for every 1% rise in market the stock price will increase by 1.54%. p- value
less than 0.05 leaving a less than 5% margin for error.

AR AND MA COEFFICIENTS AND ARIMA

ADF TEST:
Serial no: 36 FRAM assignment

The ADF test indicates the returns to be stationary.

ACF AND PACF PLOTS:

The acf plot indicates it to be a MA(0) MODEL


and in pacf plot mod(0.05) is being exceeded indicating it is a AR(1) model

Therefore the model is ARIMA(1,0,1)

PREDICTION AND DIAGNOSTIC TEST PLOT:

WEEKLY ANALYSIS:

STOCK PRICE AND RETURN PLOTS


Serial no: 36 FRAM assignment

REGRESSION SUMMARY:

The beta is 1.59 indicating 1% rise in market indicates 1.59% rise in stock price.

AR AND MA COEFFICIENTS AND ARIMA

ADF TEST AND ACF AND PACF PLOTS:


Serial no: 36 FRAM assignment

in pacf values are less than mod(0.05) hence AR(0) model, and acf indicates MA(0) model.

ARIMA prediction and diag test


Serial no: 36 FRAM assignment

From the above figure we can see that standardized residuals are randomly distributed.

ACF is significant only for the first lag and insignificant.

Thereafter and the p values which are obtained from lung-Box is greater than 0.05. Therefore, model
is good fit.
Serial no: 36 FRAM assignment

MONTHLY ANALYSIS

PLOT OF STOCK PRICE AND RETURNS

The stock has been significantly decreasing since a long time, during covid it has taken a heavy toll,

RGERESSION SUMMARY:

The beta is predicted by the CAPM model which is regression between market returns and the stoc
returns. It signifies how vulnerable is a stock to the market performance.

In this case the beta is 1.59 which indicates 1% rise in market will result in around 1.54% rise in stock
price, the p value is also less than 0.05 saying the test lies in 95% confidence interval.
Serial no: 36 FRAM assignment

AR AND MA COEFFICIENTS AND ARIMA

The AUGEMENTED DICKY FULLER test is done to check if it is a stationary time series,

The test rejects null hypothesis and indicates it to be a stationary time series.

ACF AND PACF PLOTS:

PACF is the partial autocorrelation function. From the


starting none of the lags are significant(i.e.) all of theirs value is less than 0.05. Therefore, it is AR (0)
model. As can be seen from the graph that the PACF is not significant for any value of lag, the order
of the auto regressive model can be taken as zero. 

The property of ACF is such that it identifies pattern of any autocorrelations. From the above ACF
plot for a positive value of lag, the ACF keeps on decreasing with alternative change in sign of ACF
value and tending towards 0 with the increase in lag h. From the above diagram ACF is significant for
the first lag, therefore it is MA (0) model.

Therefore it is an ARIMA(0,0,1) MODEL

ARIMA PREDICTION AND DIAG TEST PLOTS:


Serial no: 36 FRAM assignment

From the above figure we can see that standardized residuals are randomly distributed.

ACF is significant only for the first lag and insignificant thereafter

And the p values which are obtained from lung-Box is greater than 0.05. Therefore, model is good fit.

GARCH MODELS: LICHSGFIN AND L&TFH


Serial no: 36 FRAM assignment

DAILY:

LANCO:

From above figure we can see that GARCH (1,1) is best model and thereon the mean model ARFIMA
(1,0,1) is chosen. SGARCH has been the preferable model

DAILY FORECAST AFTER 28TH OCT 2020


Serial no: 36 FRAM assignment

LICHSGFIN

To estimate best GARCH model the AIC and BIC values are taken into consideration and check for
what values of GARCH the AIC and BIC values are minimum.

Here Omega, Alpha and Beta are obtained from estimated standard error are given in the figure
below.

From the figure below we see that In GARCH (1,1) is better as its AIC and BIC values are less than
other GARCH values. Therefore GARCH (1,1) is best estimate
Serial no: 36 FRAM assignment

FORECAST AFTER 28TH OCT FOR LICHSGFIN

MONTHLY:

LANCO:
Serial no: 36 FRAM assignment

From above it is clear SGARCH(1,1) is the best fit model and ARFIMA(1,0,1)

The future forecast after 28th oct 2020 for lanco using the above model will be
Serial no: 36 FRAM assignment

LICHSGFIN

To estimate best GARCH model the AIC and BIC values are taken into consideration and check for
what values of GARCH the AIC and BIC values are minimum.

Here Omega, Alpha and Beta are obtained from estimated standard error are given in the figure
below.

From the figure below we see that In GARCH (1,1) is better as its AIC and BIC values are less than
other GARCH values. Therefore GARCH (1,1) is best estimate

MONTHLY FORECAST AFTER 28TH OCT 2020 FOR LIC USING GARCH MODEL
Serial no: 36 FRAM assignment

CONCLUSION:

Betas calculated from monthly, daily, and weekly frequencies are almost identical. Surprisingly
frequency didn’t matter much for these two stocks. LANCO showed more riskiness than LIC
HOSUING AND FINANCE

ARIMA (0,0,1) model is the best model for L&T FINANCE HOLDINGS whereas ARIMA (1,0,1) is the
best model for LICHSGFIN

GARCH (1,1) model is the best model to forecast volatility for both for all three frequencies.

Prediction with ARIMA model gave NEGATIVE return in future for both LANCO AND LICHSGFN

With much higher losses for LICHSGFIN.

You might also like