Econometric analysis of the factors in the regression model to predict the market value of diversified portfolio of stocks

Nowadays it is more and more important for a company to efficiently allocate and invest its cash flows, when increasing competition and expertise of its rivals lead to a necessity for a more efficient and more reliable forecasting methods.

A company’s cash flow is divided into three categories, operating, financing and investment cash flows. Most of multinational corporations diversify its activities and hedge their future cash flows, whether they are inflows or outflows. Moreover, a company should provide liquidity in order to be able to repay its debts, financial instruments are also used for liquidity creation.

In order to effectively diversify a company’s cash flows and increase funds trading on stock exchanges is widely used. But to achieve a successful trading strategy it is needed to forecast future shifts of price, for which a technical analysis and econometrics are used. Econometrics uses statistical data in order to establish connection with changes in several factors which influence a stock price. For a successful strategy, historical values of certain indicators should be analyzed such as:

• 1.    Financial statements of a company;

• 2.    Central bank key rate;

• 3.    Macroeconomic rates (GDP, FDI, inflation).

Any of the factors could positively or negatively influence the stock price, statistical approach helps to dentify the most significant of them and determine its impact on share price.

Dow Jones Industrial Average is index, which reflects momentum of the whole US economy and can be used to analyze trends delivering broad historical data. The index includes 30 companies from diversified corporations. Therefore, we could estimate the index dependence in some of the US macroeconomic indicators, such as:

Y	Dow Jones Industrial Average
X1	Global price of brent crude, u.s. dollars per barrel
X2	Real gross domestic product, percent change from preceding period, quarterly, USA
X3	National currency unit per troy ounce gold price in us dollar, end of period
X4	Foreign direct investment in u.s.
X5	Effective federal funds rate, percent, not seasonally adjusted

Table 1 – indicators definition

Even though 5 indicators do not provide the full picture of the economy wellbeing, they could be used to increase a company’s efficiency if their significance is high.

For the analysis a least square method is used - one of the most common methods for estimation unknown parameters. The model involves previous (historical) prices, in which should be minimized the sum of deviations [1]. The analysis is commonly used by investment institution to predict stocks and commodities prices.

Hence, the variables could be used to conduct analysis of Dow Jones Industrial Average index dependence on them and their applicability for the index momentum forecasting.

System м

Estimation Method' Least Squares

Date: 02/28/17 Time: 2213

Sample: 1 37

Included observations 37

Total system (balanced) observations 37

	Coefficient	Std. Error	t-Statistic	Prob.
C(1)	-1287.065	193.4223	-5.235529	0.0000
C(2)	3 08723	0.465343	3.874682	0 0217
CG)	34.56456	8.324534	3.928373	0.0081
C(4)	-0.31979	0.211214	-2.934772	0 0271
C(5)	0.00011	5.11E-05	19 024372	0 0000
C(6)	57.34324	16.87293	4.153636	0.0011

Determinant residual covariance 9215.082

Equation: Y=C(1)-C(2)*XVC(3)*X2-C(4)*X3+C(5)*X4+C(6)*X5

Observations 37 __________________________________________

R-squared	0.913242	Mean dependent var	1498421
Adusted R-squared	0 902387	S.D dependent var	370.1024
S E of regression	101.4231	Sum squared resid	379930 9
Durbin-Watson stat	1.128382

Table 2 – Least squares regression results

The regression analysis shows that R – squared = 91.3242%, which implies high significance of the link between Y and X. Probability of variables (Prob.) less than 5% which means high level of significance, moreover t – statistics also confirms it. Durbin – Watson test equals = 1.1283, which implies that our model is in the area of uncertainty, because some of the variables may depend on our final results (correlation).

In order to confirm that the model is useful, a test on heteroscedasticity has to be conducted. It analyzes whether the variance of the random error is constant or focused. We cannot accept the model if errors distributed without similarity. Heteroscedasticity is random distribution of errors, which we cannot use to predict share price. Provided two type of tests on heteroscedasticity.

X1 = 29.37%	Prevails heteroscedasticity.
X2 = 90.21%	Prevails homoscedasticity.
X3 = 37.4%	Prevails heteroscedasticity.
X4 = 19.73%	Prevails heteroscedasticity.
X5 = 58.02%	Prevails homoscedasticity.

Table 3 – Glejser test

Breusch – Pagan – Godfrey test equals 27.41%, that means we can accept appearance of heteroscedasticity with 72,59%. Two variables, demonstrate focused distribution of errors – x2 - real GDP and x5 - effective federal fund rate.

y	x1	x2	x3	x4          x5
y   1
x1 0,245234238	1
x2 0,382342316	0,148032743	1
x3 0,155634522	0,430823125	0,320973946	1
x4 0,782057327	0,229210012	0,369824052	0,292142131	1
x5 0,139832656	0,043209742	0,100349724	0,472139023	0,440912801  1

Table 3 – Correlation

We can draft the equation of Dow Jones Industrial Average:

Dow Jones Industrial Average = -1287.065 + x1 * 3.087 + x2 * 34.564 + x3 * (-0.319) + x4 * 0.001+x5 * 57.343

The equation could be interpreted that Dow Jones Industrial Average will change if one of the «x» factors is changed, most of them has positive impact, the one (x4) has negative. The most significant factors are: real GDP and effective federal fund rate.

The analysis of factors which could affect the Dow Jones Industrial Average index lead us to a conclusion that only two of them have passed all of the tests and could be used for statistical analysis – GDP growth and the effective federal funds will increase Dow Jones Industrial Average stocks’ price. This model is suitable for forecasting, but for the more precise and reliable analysis additional factors should be added.