Advanced Time Series Analysis

Question: Discuss about the Advanced Time Series Analysis. Answer: Introduction: One of the most important aspects of a scientific work is the ability of being reproduced (Popper, 2009). For an experiment to be regarded as a valid one it must possess attributes of replicability. The validity of an experiment can be determined by reviewing the methodology used, how the data was collected, whether enough literature review was undertaken, review the quantitative analysis done the researchers and lastly review the conclusions made by the authors. It is for this reason that this paper attempts to get a clear understanding and scientific authenticity of the paper entitled Exploitable Predictable Irrationality: The FIFA World Cup Effect on the U.S. Stock Market. This paper validates the fact that the experiment was independent of the local conditions, and that what was written clearly describes the findings of the experiment. To get a clear picture of the study, the following aspects of the paper were reviewed; the objectives of the study (i.e. whether the researchers h ad SMART objectives), the literature review conducted by the researchers, the methodology of the study, data collection procedures, how analysis was done and lastly the conclusions made by the researchers. In the review, despite the fact that the authors were articulate in conducting the experiment, a number of issues were found to have either been compromised or overlooked. For instance, the authors failed to include sensitivity analysis that would help validate potential errors in the results and findings. Also the literature review was not adequate enough to the existing gaps in the previously done studies. Conclusion was not as per the required standards; the researchers failed to highlight how their newly developed model was important as compared to the other models and as such the conclusion did not make any economic sense. Critical evaluation The paper under review sought to develop a practical method to exploit the asymmetric characteristic of the soccer sentiment effect. This objective is a SMART objective since it is measurable as is required of a SMART objective. The paper however does not provide enough literature review to show where the research fits into the existing body of nor to outline the existing gaps in the previously done studies. Even though the paper highlights the dataset used, it does not clearly describe where and how the data was obtained. It is important to clearly describe the methodology involved in obtaining the data as this would help verifying whether there could be any sort of bias associated with it that may end up biasing the results. On a positive note, the study clearly highlights the theoretical background but does not relate to the study under investigation. In terms of analysis, though the researchers did present several analytic tests, they have failed to include important components that relates to the kind of data under investigation. One such test is the test of stationarity and the test on absence or non-absence of autocorrelation that is associated with time series. The researchers ought to have presented these tests to tell the readers whether there data was affected with the mentioned issues or not. Another important aspect of the paper that the authors overlooked is the sensitivity analysis; clearly this paper has not mentioned anything to do with sensitivity analysis neither was one done by the authors. Sensitivity analysis is important in investigating the potential errors or changes of the parameter values or assumptions and the impact they have on the conclusions that can be made (Baird, 2009). Lastly, even though the authors gave their conclusion, the conclusions given have not highlighted how important their proposed model is as compared to the previous models. They have also not highlighted the shortcomings of the proposed model. These are important aspects that needs to be highlighted to enable the readers to make a quick and wise decision on the best model to utilize. In summary, the authors conclusions don't make any economic sense and it does not matter economically. Replication Exercise In this section, we attempt to replicate the analysis done by the authors. We ran OLS regression where we estimated the REW which is the stock return from equal weighted index (dependent variable) using 11 independent variables. Results showed that the value of R-Squared is 0.1136; which implies that 11.36% of the variation in the dependent variable Stock return) is explained by the 11 variables in the model. It can also be observed that out of the 11 variables, 6 of them were statistically significant in the model at 5% level of significance (p-value 0.05). 3 of the significant variables had negative regression coefficient while the other three had a positive regression coefficient. The negative coefficient shows that a unit increase in the variable results to a decrease in the stock returns. Dependent Variable: REW Method: Least Squares Date: 03/19/17 Time: 12:24 Sample: 1 16819 Included observations: 16819 Variable Coefficient Std. Error t-Statistic Prob. C 0.001442 0.000809 1.781849 0.0748 D1 -0.002603 0.000821 -3.172375 0.0015 D2 -0.001523 0.000820 -1.858215 0.0632 D3 -0.000351 0.000820 -0.427743 0.6688 D4 -0.000339 0.000820 -0.413655 0.6791 D5 0.000440 0.000820 0.536553 0.5916 E -0.001581 0.000431 -3.670864 0.0002 H 0.001137 0.000368 3.085931 0.0020 J1 -0.074294 0.002387 -31.12017 0.0000 J2 0.064294 0.002387 26.93237 0.0000 P -0.000145 0.000165 -0.877557 0.3802 T 0.003861 0.000419 9.203592 0.0000 R-squared 0.113574 Mean dependent var 0.000629 Adjusted R-squared 0.112993 S.D. dependent var 0.008012 S.E. of regression 0.007546 Akaike info criterion -6.934938 Sum squared resid 0.956975 Schwarz criterion -6.929423 Log likelihood 58331.36 Hannan-Quinn criter. -6.933119 F-statistic 195.7638 Durbin-Watson stat 1.567376 Prob(F-statistic) 0.000000 Further Analysis The last part of this review is the presentation of the further analysis that the authors ought to have included in their paper. In addition to the analysis this reviewed added analysis of the ARCH model for the stock returns. To estimate a standard GARCH(1,1) model that has no independent variables we applied the following equation models: We then fitted a GARCH(1,1) model to the first difference of stock returns (REW), based on backcast values. The output is presented below: Table 2: GARCH (1,1) model Dependent Variable: REW Method: ML - ARCH (Marquardt) - Normal distribution Date: 03/19/17 Time: 17:36 Sample: 1 16819 Included observations: 16819 Convergence achieved after 39 iterations Presample variance: backcast (parameter = 0.7) GARCH = C(2) + C(3)*RESID(-1)^2 + C(4)*GARCH(-1) Variable Coefficient Std. Error z-Statistic Prob. C 0.000928 4.52E-05 20.53041 0.0000 Variance Equation C 2.12E-06 5.39E-08 39.39110 0.0000 RESID(-1)^2 0.174876 0.004373 39.98830 0.0000 GARCH(-1) 0.792184 0.004446 178.1676 0.0000 R-squared -0.001394 Mean dependent var 0.000629 Adjusted R-squared -0.001394 S.D. dependent var 0.008012 S.E. of regression 0.008018 Akaike info criterion -7.290314 Sum squared resid 1.081093 Schwarz criterion -7.288476 Log likelihood 61311.90 Hannan-Quinn criter. -7.289707 Durbin-Watson stat 1.586013 In the above table, the co-efficient C(2) shows the last period (t-1) volatility while C(3) shows the impact of long term volatility and lastly C(4) shows the leverage effect. As can be seen, the symmetry term is positive which indicates a positive shock has a much greater volatility impact as compared to the negative shocks that bear the same magnitude. References Baird, B. F. (2009). Managerial Decisions Under Uncertainty, An Introduction to the Analysis of Decision Making. Popper, K. R. (2009). The Logic of Scientific Discovery.