Further analysis Directory UMM :Data Elmu:jurnal:J-a:Journal of Energy Finance & Development:Vol4.Issue1.Jun1999:

R.W. Faff, T.J. Brailsford Journal of Energy Finance and Development 4 1999 69–87 81 comparison to Eq. 1, the Oil and Gas, and Diversified Resources industries have a consistent positive sensitivity see Table 2. Similarly, the Paper and Packaging, and Transport industries have a consistent negative sensitivity. In none of these cases is the exchange rate factor also significant. 10 Furthermore, the Wald test rejects the null hypothesis that the two coefficients are jointly equal to zero, although we do not reject the null hypothesis that the two coefficients are equal. The additional sensitivity is observed for the Food and Household Goods industry, which has a negative sign. However, we note that the Wald test does not reject the null hypothesis that the two coefficients are jointly equal to zero. Table 3 offers an insight into the anomalous result earlier observed for the Banks industry. Recall that the Banks industry experienced a significant negative sensitivity to the oil price factor, contrary to expectations. We now observe that this sensitivity disappears with the inclusion of the foreign exchange factor and that the coefficient on this factor is significant. Recall that rejection of the null hypothesis that the two coefficients are equal implies that the exchange rate factor offers additional explanatory information over and above the oil price factor implying that Eq. 1 is mis-specified. In Table 3, we observe that this is the case for the Banks industry, implying that there is an omitted variable that is in part attributable to the foreign exchange factor. Similarly, the Wald test also rejects the null hypothesis for the Building Materials and Engineering industry. These results suggest that only the exchange rate factor should be included to augment the market model for these industries. We do not explore these results further given the focus of the paper but note that they offer insight for more general multi-factor tests. However, it should be noted that in these cases the AUD oil price sensitivity is of very similar magnitude and statistical significance.

5. Further analysis

The analyses in Tables 2 and 3 cover the full sample period of 1983–1996. As such, their length may obscure possible changes in sensitivity to the oil price factor across industries over time. Therefore, to investigate for potential changes in the sensitivity to the oil price factor over time, two sub-periods of 1 July 1983 to October 1989 and 2 November 1989 to March 1996 are examined. The analysis involves augmenting Eq. 1 with dichotomous dummy variables on both the intercept and slope coefficients. The first sub-period produces five industries with a significant oil price sensitivity using the model in Eq. 1 consistent with the full period results reported in Table 2. Four of these industries are consistent with our initial predictions, while we place little emphasis on the Banks industry result given the previous argument that the model is mis-specified in this industry. In the second sub-period, the two industries with a significant positive sensitivity Oil and Gas, and Diversified Resources remain so. However, the industries with a significant negative sensitivity in the first sub-period lose their significance in the second sub-period. This result is consistent with an argument that firms with a signifi- cant negative sensitivity either can effectively pass on price negative changes to their customers or have become increasingly engaged in effective hedging activities. 82 R.W. Faff, T.J. Brailsford Journal of Energy Finance and Development 4 1999 69–87 Table 3 Estimation of the Market Model Augmented by an Oil and Exchange Rate Factor: 1983:07 to 1996:03 Wald Test a Wald Test b g i d i H : g i 5 d i H : g i 5 d i 5 ASX Industry t statistic t statistic p value p value 1. Gold 0.0795 2 0.0014 0.166 1.600 1.24 20.01 0.68 0.45 2. Other Metals 2 0.0339 2 0.0719 0.073 0.776 20.78 20.51 0.79 0.68 3. Solid Fuels 2 0.0197 0.1418 1.640 1.734 20.51 1.13 0.20 0.42 4. Oil and Gas 0.2454 0.0285 4.542 58.021 7.55 0.28 0.03 0.00 5. Diversified 0.1309 0.0642 0.445 18.105 Resources 4.25 0.64 0.50 0.00 6. Developers and 2 0.0231 2 0.0589 0.165 1.024 Contractors 20.85 20.67 0.68 0.60 7. Building 0.0056 2 0.1728 6.088 6.110 Materials 0.25 22.40 0.01 0.05 8. Alcohol and 2 0.0136 2 0.1352 0.703 0.889 Tobacco 20.29 20.93 0.40 0.64 9. Food and 2 0.0735 2 0.0153 0.420 7.132 Household Goods 22.66 20.17 0.52 0.03 10. Chemicals 2 0.0369 2 0.1482 1.141 2.940 21.15 21.43 0.29 0.230 11. Engineering 2 0.0061 2 0.2785 8.690 9.178 20.22 23.02 0.01 0.01 12. Paper and 2 0.0939 0.0156 1.460 11.794 Packaging 23.37 0.17 0.23 0.01 13. Retail 2 0.0342 2 0.2300 4.457 6.889 21.20 22.49 0.04 0.03 14. Transport 2 0.0908 2 0.1945 0.893 9.180 22.69 21.78 0.35 0.01 continued Formally, there is only one case Miscellaneous Services in which the Wald test rejects the null hypothesis of equality of the coefficient across sub-periods. As such, there appears to be little evidence of changing sensitivities across time, implying that oil price sensitivity is a long-term phenomenon, except perhaps for those industries with a significant negative sensitivity. A further issue is the potential effect of the stock market crash of October 1987. We conduct some sensitivity analysis to assess how robust the findings are to this extreme market movement. Interestingly, the estimation of the oil price sensitivities is qualitatively unaffected, and importantly, the conclusions drawn based upon them are robust to the treatment of the crash. Finally, we also conducted GMM tests of the two-factor APT containing the market R.W. Faff, T.J. Brailsford Journal of Energy Finance and Development 4 1999 69–87 83 Table 3 continued Wald Test a Wald Test b g i d i H : g i 5 d i H : g i 5 d i 5 ASX Industry t statistic t statistic p value p value 15. Media 2 0.0947 0.1009 1.081 3.296 21.64 0.54 0.30 0.19 16. Banks 2 0.0740 2 0.3416 6.149 13.246 22.23 23.17 0.01 0.01 17. Insurance 0.0928 0.0266 0.225 4.158 2.04 0.19 0.64 0.13 18. Entrepreneurial 2 0.0052 0.1429 0.595 0.596 Investors 20.09 0.75 0.44 0.74 19. Investment and 0.0576 2 0.0412 1.346 5.189 Financial Services 2.12 20.48 0.25 0.08 20. Property Trusts 2 0.0089 0.0673 1.279 1.341 20.43 1.00 0.26 0.51 21. Miscellaneous 2 0.0794 0.1474 3.551 7.032 Services 2.14 1.23 0.06 0.03 22. Miscellaneous 0.0128 2 0.0056 0.044 0.223 Industrials 0.45 20.06 0.83 0.90 23. Diversified 2 0.0236 2 0.0094 0.035 1.015 Industrials 21.01 20.12 0.85 0.60 24. Tourism and 2 0.0016 2 0.0756 0.424 0.447 Leisure 20.05 20.67 0.52 0.80 Number of Significant Test Statistics 5 2 3 6 This table reports the output from the following regression: R it 5 a i 1 b i R mt 1 g i OILRUSD t 1 d i XR t 1 e it 7 where R it is the return on the ith asset or portfolio in month t, R mt is the return on the market index in month t and OILRUSD t is the oil price return in month t, expressed in U.S. dollars, and XR t is the return from holding U.S. dollars. Statistic is significantly different from zero at the 1 level. a Wald test statistic has a chi square distribution with 1 degree of freedom. The p value is given in parentheses. b Wald test statistic has a chi square distribution with 2 degrees of freedom. The p value is given in parentheses. and the oil price factors. 11 To perform these tests, estimation of the system of zero- intercept two-factor models in excess returns form is undertaken. Furthermore, the tests are performed in two variations: 1 using returns on a value-weighted domestic market index, and 2 using returns on a global market index. In a system of excess return models, the null hypothesis is that the intercept terms are jointly equal to zero. 12 The overwhelming result is that the null model cannot be rejected. This finding is robust to 1 the choice of market portfolio proxy, 2 performing sub-period analysis, 3 performing separate industry sector resource or industrial analysis, and 4 the crash of October 1987. Hence, we accept the restriction placed on the model. 13 84 R.W. Faff, T.J. Brailsford Journal of Energy Finance and Development 4 1999 69–87

6. Conclusion