A Financial Crash: a Devastating Effect on the Economy - Essay Example

Stimulus (Second Project) Table of Contents Introduction 3 The Bottleneck 3 Data Collection and Methodology of Estimation 5 Empirical Analysis 8 Conclusion 10 Reference 10 Appendix 11 Introduction A financial crash can have a devastating effect on the fundamentals of an economy. The effects are further proliferated if the nation is a developing one, due to the absence of abundant funds to tackle the crisis. It is because of the resource crunch that many developmental projects in growing nations are obliterated, while the funds allotted for their commencement are utilized in some other urgent operation. Urgency of an issue however, is decided on the basis of the future utilities it is supposed to fetch to the society. The present paper is an attempt to redeem the government from a similar situation, where discussions are on to inject the money previously meant for a promotion in the number of secondary enrolments in the nation, in the financial sector of the economy which is going through a difficult phase right now. The target of the administration is to enhance the GDP situation in the nation that will serve as a gateway towards a stable growth path in future. This study is an effort to disentangle the tight spot, with the help of empirical analyses, availing various statistical tools. The Bottleneck The Finance Ministry of the nation had already decided upon the implementation of policy measures to improvise the country’s rate of enrolment in secondary education, as the desired channel of fund inflow. However, the financial shock that overwhelmed the economy meanwhile, and led to imperative demands for financial support from banks around the country, has compelled it to reconsider the outcome of its previous meeting. The deadlock that the nation is facing at present could be logically solved by means of statistical analysis. The idea is to carry on a comparative analysis through estimating a regression equation, with a function of change in GDP as the dependent variable and the changes in both the levels of secondary school enrolments and infusion of credit to the banking sector, as the independent ones along with a number of other variables, expected to produce significant impact on the dependent variable. To be precise, the model being designed for estimation is, dlypci = 1 + 2lypc90i + 3lsecedi + 4govgdpi + 5openi + 6infli + 7crediti + ui Where, lypc90 = natural logarithm of per capita GDP (constant prices: chain series) in 1990, dlypc = difference between natural logarithms of per capita GDP in 2005 and for 1990, at constant prices, lseced = natural logarithm of percentage of secondary school age population enrolled at secondary school in 1990, govgdp = government share of real GDP per capita in 1990, open = openness of the economy, i.e., ratio of the aggregate of exports and imports to the GDP in 1990, infl = five-year inflation rate defined as the difference between the natural logarithms of CPI for the years 1990 and 1985 and credit = ratio of private credit by deposit money banks and other financial institutions to GDP in 1990 The question that might arise at this point is that, what should be the basis of deductions regarding the validity of the above model as well as for the reliability of the estimated coefficients. The subsequent sections will hopefully help the reader to reach a suitable answer to any such queries. Data Collection and Methodology of Estimation The research aims to completely explore the average outcome of similar situations faced by other developing nations of the world and that is why will use information core to 50 growing economies, sampled across all continents, rather than using previously assumed domestic data. Data on all the above specified variables (dependent and explanatory) have been collected to carry on with the analysis. After running a regression in the above specified format, the next step would be to test for the level of significance of the estimated coefficients so as to be certain about the impact of the corresponding explanatory variable on the dependent one. The null and alternative hypotheses which needs to be tested in this context are, 1. H01: 2=3=4=5=6=7=0 against H11:j0 for at least one j The ideal way of testing in this case is through estimating an F-statistic, obtained from the ANOVA table and defined as, F = (MSR/ MSE) Where, MSR = Mean sum of squares of the regressed model and MSE = Mean sum of squares of the predicted error terms. 2. H0i: i = 0 against H1i:i  0 for, i = 2, 3 and 7 The ideal way of testing in this case is through estimating a Student’s t-statistic as follows – T = (i - 0i)/ se (i) Where, i = Estimated coefficient of the ith term in the regression model and 0i = Hypothesized value of the ith term. With the degrees of freedom of the estimated statistics available, it will be easy to obtain their respective probability values or rather the level of significance at which the respective null hypothesis will be accepted. If this level of significance exceeds that which has been assumed as a standard, then the related null hypothesis cannot be rejected at the specified level of significance. For F-statistic, this standard level is considered to be 0.05; so is the value for the first two cases of hypothesis testing using Student’s t-statistic. However, in the third test, the standard level of significance is considered to be 0.01. However, completion of the hypothesis testing process does not accomplish the fact that the regression model so estimated is a reliable one. To ensure this, the model needs to undergo certain diagnostic tests necessary to establish or guarantee its compliance with the Classical Linear Regression (CLRM) assumptions. Three such important assumptions, which might also be regarded as the most vital, have been briefly addressed. Linearity Assumption, that assumes that the underlying model is a linear one, in the sense that the coefficients being estimated are linear in nature. The best way to identify a linear model is to plot down the predicted residuals against the fitted model and trace down the trend that the residuals follow. If it appears to be haphazardly spread over a linear scale, then the linearity assumption is supposed to have been satisfied. Normality Assumption, that assumes that the predicted residuals follow a normal distribution; in simple words, have zero mean and constant variance. Homoscedasticity Assumption, which could be considered as a corollary of the normality assumption, since it stresses upon the constant variance part of the predicted residuals. There are certain tests, known as diagnostic tests, that help to examine whether the assumptions are actually satisfied or not. Though there are no such hard-core tests to guarantee the fulfillment of the linearity supposition, some mathematical tests are available that guarantees the remaining two. Two such tests being employed in the present paper are Jarque-Bera test for normality and Breusch-Pagan Test for homoscedasticity. Jarque-Bera test for normality estimates the statistic JB as, Where, S = Skewness in the distribution of the estimated residuals, K = Kurtosis of the estimated cumulative distribution of the estimated residuals and n = Degrees of freedom Given that in case of a normal distribution, skewness = 0 and kurtosis = 3, the estimated JB statistic is supposed to be 0. The relevant null and alternative hypotheses in this perspective are, H0J: Estimated JB = 0 against, H1J: JB  0. On the other hand, Breusch-Pagan test for homoscedasticity estimates a linear regression model of the squares of the predicted residuals on the original explanatory variables. If it is found that the independent variables are able to explain a large chunk of total variations in squared predicted residuals, then the assumption of homoscedasticity is believed to be violated. Hence, the null and alternative hypotheses in this context are, H0B: homoscedastic variance against, H1B: heteroscedastic variance. The test statistic that evaluates the validity of both the assumptions is chi-square (χ2). Thus in either of the two cases discussed above, fate of hypothesis testing will depend upon the following rules – If the level of significance (p-value) of the estimated chi-square statistics > 0.05 (standard level of significance), reject the null hypothesis and If p-value of estimated chi-square statistics < 0.05, do not reject the null hypothesis at the assumed level of significance. In the former case, the estimated regression model violates the assumptions of CLRM, while the latter is found to abide by them. Thus, in the first case, the estimated model needs to be re-estimated after some corrective measures have been applied. One such common problem is that about the presence of outliers in the model, which often swerve the model from its actual form. Outliers could be detected through noting the values of the estimated residuals at each individual observation. Observations at which its value is found to exceed three times of that of the standard error of the estimated regression model (square root of the Mean Sum of Squares of the model), are the ones featuring outliers. An ideal way of addressing such typical discrepancies in analysis is through the construction of dummy variables defined as, D = 1, if an observation displays an outlier and 0 otherwise The original regression model has to be estimated once again after such an impulse dummy is created. Thus, the form of the new estimated model would be, dlypci = 1 + 2lypc90i + 3lsecedi + 4govgdpi + 5openi + 6infli + 7crediti + 7Di + ui After this model is estimated, the diagnostic tests have to be repeated on it, so as to ensure that there are no more interruptions in CLRM assumptions. Empirical Analysis The sample data on different macroeconomic variables for fifty developing nations have been subjected to an ordinary least square (OLS) regression method and the model so evolved is, dlypci = 1.9094 – 0.1998 lypc90i + 0.2503 lsecedi + 0.0100 govgdpi + 0.0002 openi + 0.1143 infli + 0.2125 crediti + ui The estimated Student’s t-statistics corresponding to the explanatory variables and F-statistic for testing the significance of the overall model has been presented in the appendix to the chapter (A.1). It is found that, the estimated F-statistic in this case is, 1.77, and the corresponding p-value at 6, 43 degrees of freedom is, 0.1289. Clearly, 0.1289 (p-value) > 0.05 (standard level of significance) Thus, the null hypothesis, H01 cannot be rejected at 5% level of significance. Again, the p-values of the estimated coefficients of variables ‘lnypc90’, ‘lseced’ and ‘credit’ reveal the following information. P-valuelnypc90 = 0.011 < 0.05, P-valuelseced = 0.0065 < 0.05 and P-valuecredit = 0.068 > 0.01. Hence, though null hypotheses, H02 and H03 can be rejected at 5% level of significance, H07 cannot be rejected at 1% level of significance. The implications of these results is that, though the variables lnypc90 and lseced are found to have a significant impact on explanatory variables dlypc, but, the variable credit does not have any such considerable effect on the same. To clarify any doubts about the validity of the estimated equation, diagnostic tests are performed as additional measures. Linearity The adjoining diagram assures that the linearity assumption is followed by the predicted residuals, since they appear to be scattered along a straight line. Normality On the other hand, estimated JB = 0.138147, with a p-value of the corresponding chi-square statistic equal to 0.9333. Thus, p-value (0.9333) > 0.05, implying that the null hypothesis, H0J cannot be rejected at 5% level of significance. Homoscedasticity The chi-square test statistics meant for Breusch-Pagan test, so estimated has a p-value equal to 0.2653, where, 0.2653 > 0.05, indicating that the hull hypothesis, H0B cannot be rejected at 5% level of significance. Thus, all the three diagnostic tests conducted above reveal that the CLRM assumptions are perfectly satisfied by the estimated model. Conclusion The study reveals that, it would be wiser for the Finance Ministry to transfuse its funds to uplift the secondary education situation in the nation. The consequences of instilling money into the financial sector might not be optimal for the economy’s future, as far as an ascent in its GDP value is concerned. Though the outcome might seem quite disparaging at the wake of the shock, it would prove to be rather a complacent one for future growth. Reference Heston, A., Summers, R. and Aten, B. (August 2009) Penn World Table Version 6.3, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Appendix A.1 Estimated Regression Model A.2 Jarque-Bera Test A.3 Breusch-Pagan Test Read More