Hi,
I am using a bivariate garch in mean model to examine the effect of oil price uncertainty on real stock returns. I am using monthly data for real oil prices and real stock returns for the period 1973:1 - 2009:12.
The results show a positive effect of oil price uncertainty on stock returns (weird results), and I wonder whether I have errors in the code!
VAR-GARCH-M
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
VAR-GARCH-M
Last edited by economics2012 on Wed May 02, 2012 12:38 pm, edited 3 times in total.
Re: VAR-GARCH-M
You'll probably need to provide more details if you want anyone to be able to help. For example, can you post the complete estimation results, along with a description of what the various variables are (unless it is completely obvious from the variable names)? You may also want to post the data set so others can try to repeat the estimation.
Also, can we assume that you've tested the OLS residuals for GARCH effects?
Thanks,
Tom Maycock
Estima
Also, can we assume that you've tested the OLS residuals for GARCH effects?
Thanks,
Tom Maycock
Estima
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
Hi Tom,
Here are my complete estimation results.
The variables used are: oilgrow which stand for: real oil price growth , and stock returns.
I also attached the data file with three columns: date, oil growth and stock returns.
I believe line 53. of the last table shows the effect of oil price uncertainty on stock returns with a positive coefficient of 0.08152284 and t-statistics of 1.31957.
In regard of testing the OLS residuals for GARCH effects, I am not sure whether you meant what I got in the first table above!
Is there any RATS wizard for testing the OLS residuals for GARCH effects?
Thanks a lot,
Here are my complete estimation results.
The variables used are: oilgrow which stand for: real oil price growth , and stock returns.
Code: Select all
VAR/System - Estimation by Least Squares
Monthly Data From 1974:01 To 2009:12
Usable Observations 432
Dependent Variable OILGROW
Mean of Dependent Variable 0.2369802905
Std Error of Dependent Variable 7.0912551842
Standard Error of Estimate 6.1203682923
Sum of Squared Residuals 15245.775570
Durbin-Watson Statistic 1.9786
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. OILGROW{1} 0.520215612 0.049304475 10.55108 0.00000000
2. OILGROW{2} -0.056023448 0.055598300 -1.00765 0.31422274
3. OILGROW{3} -0.047396997 0.055730447 -0.85047 0.39556464
4. OILGROW{4} -0.012540444 0.055509720 -0.22591 0.82138142
5. OILGROW{5} -0.013774081 0.055804895 -0.24683 0.80516759
6. OILGROW{6} -0.156653647 0.056013501 -2.79671 0.00540738
7. OILGROW{7} 0.076372676 0.056348410 1.35537 0.17605290
8. OILGROW{8} -0.007260344 0.056523610 -0.12845 0.89785791
9. OILGROW{9} -0.013793027 0.056821586 -0.24274 0.80832699
10. OILGROW{10} 0.034771912 0.057106989 0.60889 0.54293661
11. OILGROW{11} 0.077203440 0.057275340 1.34794 0.17842902
12. OILGROW{12} -0.110411073 0.050948530 -2.16711 0.03080585
13. STOCKRETURNS{1} 0.071504797 0.063697545 1.12257 0.26228298
14. STOCKRETURNS{2} 0.009777206 0.063760930 0.15334 0.87820486
15. STOCKRETURNS{3} 0.174765054 0.063528975 2.75095 0.00620757
16. STOCKRETURNS{4} -0.079545208 0.064202272 -1.23898 0.21606765
17. STOCKRETURNS{5} 0.019224822 0.063627290 0.30215 0.76269402
18. STOCKRETURNS{6} -0.031107333 0.063744588 -0.48800 0.62581291
19. STOCKRETURNS{7} -0.017301692 0.063611910 -0.27199 0.78576897
20. STOCKRETURNS{8} 0.006381151 0.063626547 0.10029 0.92016292
21. STOCKRETURNS{9} 0.022956458 0.063569084 0.36113 0.71819253
22. STOCKRETURNS{10} 0.031585505 0.063681602 0.49599 0.62016845
23. STOCKRETURNS{11} -0.008874725 0.063669535 -0.13939 0.88921305
24. STOCKRETURNS{12} -0.054992978 0.063458035 -0.86660 0.38666969
25. Constant 0.163708463 0.296075836 0.55293 0.58061664
26. SQRTHOIL 0.000000000 0.000000000 0.00000 0.00000000
F-Tests, Dependent Variable OILGROW
Variable F-Statistic Signif
*******************************************************
OILGROW 13.3676 0.0000000
STOCKRETURNS 0.9399 0.5066601
Dependent Variable STOCKRETURNS
Mean of Dependent Variable 0.0211076390
Std Error of Dependent Variable 4.8051892243
Standard Error of Estimate 4.7505842819
Sum of Squared Residuals 9185.1967649
Durbin-Watson Statistic 1.9954
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. OILGROW{1} -0.062631622 0.038269767 -1.63658 0.10249076
2. OILGROW{2} 0.081653285 0.043154986 1.89209 0.05918799
3. OILGROW{3} 0.004278068 0.043257558 0.09890 0.92126827
4. OILGROW{4} -0.071395226 0.043086231 -1.65703 0.09828379
5. OILGROW{5} 0.016035537 0.043315344 0.37020 0.71142267
6. OILGROW{6} -0.025565076 0.043477262 -0.58801 0.55685153
7. OILGROW{7} 0.000388776 0.043737216 0.00889 0.99291212
8. OILGROW{8} -0.087855469 0.043873205 -2.00249 0.04589541
9. OILGROW{9} 0.045515656 0.044104491 1.03200 0.30268686
10. OILGROW{10} -0.073375723 0.044326019 -1.65536 0.09862138
11. OILGROW{11} -0.003679140 0.044456693 -0.08276 0.93408477
12. OILGROW{12} -0.056065357 0.039545870 -1.41773 0.15703480
13. STOCKRETURNS{1} 0.111850757 0.049441560 2.26228 0.02420656
14. STOCKRETURNS{2} -0.069348343 0.049490759 -1.40124 0.16190489
15. STOCKRETURNS{3} 0.031606247 0.049310717 0.64096 0.52190875
16. STOCKRETURNS{4} -0.003463371 0.049833325 -0.06950 0.94462648
17. STOCKRETURNS{5} 0.047549704 0.049387029 0.96280 0.33622087
18. STOCKRETURNS{6} -0.064160867 0.049478074 -1.29675 0.19545069
19. STOCKRETURNS{7} 0.023529918 0.049375091 0.47655 0.63393519
20. STOCKRETURNS{8} -0.014721206 0.049386452 -0.29808 0.76579278
21. STOCKRETURNS{9} -0.032231345 0.049341850 -0.65323 0.51397987
22. STOCKRETURNS{10} 0.047474881 0.049429185 0.96046 0.33739271
23. STOCKRETURNS{11} -0.020479281 0.049419819 -0.41439 0.67880383
24. STOCKRETURNS{12} 0.044874198 0.049255654 0.91105 0.36281006
25. Constant 0.066856129 0.229811858 0.29092 0.77126315
26. SQRTHOIL 0.000000000 0.000000000 0.00000 0.00000000
F-Tests, Dependent Variable STOCKRETURNS
Variable F-Statistic Signif
*******************************************************
OILGROW 1.7123 0.0617950
STOCKRETURNS 0.8971 0.5499893
Simplex Optimization, Trial 0. Function Calls: 60
Old Function = -2778.401291 New Function = -2777.386240
New Coefficients:
0.000000 0.520216 -0.056023 -0.047397 -0.012540
-0.013774 -0.156654 0.076373 -0.007260 -0.013793
0.034772 0.077203 -0.110411 0.071505 0.009777
0.157289 -0.079545 0.019225 -0.031107 -0.017302
0.006381 0.022956 0.031586 -0.008875 -0.054993
0.163708 0.000000 -0.062632 0.081653 0.004278
-0.071395 0.016036 -0.025565 0.000389 -0.087855
0.045516 -0.073376 -0.003679 -0.056065 0.111851
-0.069348 0.031606 -0.003463 0.047550 -0.064161
0.023530 -0.014721 -0.032231 0.047475 -0.020479
0.044874 0.066856 0.000000 7.058229 0.200000
0.600000 4.252406 0.200000 0.600000
Simplex Optimization, Trial 1. Function Calls: 62
Old Function = -2777.386240 New Function = -2735.504304
New Coefficients:
0.000508 0.517570 -0.055739 -0.047156 -0.012477
-0.013704 -0.155857 0.075984 -0.007223 -0.013723
0.034595 0.076811 -0.109850 0.071141 0.009727
0.173876 -0.079141 0.019127 -0.030949 -0.017214
0.006349 0.022840 0.031425 -0.008830 -0.054713
0.162876 0.000508 -0.062313 0.081238 0.004256
-0.071032 0.015954 -0.025435 0.000387 -0.087409
0.045284 -0.073003 -0.003660 -0.055780 0.111282
-0.068996 0.031446 -0.003446 0.047308 -0.063835
0.023410 -0.014646 -0.032067 0.047233 -0.020375
0.044646 0.066516 0.000508 8.469875 0.198983
0.596949 4.230784 0.198983 0.596949
Simplex Optimization, Trial 16. Function Calls: 137
Old Function = -2735.504304 New Function = -2704.644081
New Coefficients:
0.000000 0.520216 -0.056023 -0.047397 -0.012540
-0.013774 -0.156654 0.076373 -0.007260 -0.013793
0.034772 0.077203 -0.110411 0.071505 0.009777
0.174765 -0.079545 0.019225 -0.031107 -0.017302
0.006381 0.022956 0.031586 -0.008875 -0.054993
0.163708 0.000000 -0.062632 0.081653 0.004278
-0.071395 0.016036 -0.025565 0.000389 -0.087855
0.045516 -0.073376 -0.003679 -0.056065 0.111851
-0.069348 0.031606 -0.003463 0.047550 -0.064161
0.023530 -0.014721 -0.032231 0.047475 -0.020479
0.044874 0.066856 0.010000 9.881521 0.200000
0.600000 4.252406 0.200000 0.600000
Simplex Optimization, Trial 32. Function Calls: 213
Old Function = -2704.644081 New Function = -2683.913998
New Coefficients:
0.000508 0.517570 -0.055739 -0.047156 -0.012477
-0.013704 -0.155857 0.075984 -0.007223 -0.013723
0.034595 0.076811 -0.109850 0.071141 0.009727
0.173876 -0.079141 0.019127 -0.030949 -0.017214
0.006349 0.022840 0.031425 -0.008830 -0.054713
0.162876 0.000508 -0.062313 0.081238 0.004256
-0.071032 0.015954 -0.025435 0.000387 -0.087409
0.045284 -0.073003 -0.003660 -0.055780 0.111282
-0.068996 0.031446 -0.003446 0.047308 -0.063835
0.023410 -0.014646 -0.032067 0.047233 -0.020375
0.044646 0.066516 0.000508 11.293167 0.198983
0.596949 4.230784 0.198983 0.596949
Simplex Optimization, Trial 577. Function Calls: 1819
Old Function = -2610.909923 New Function = -2608.539051
New Coefficients:
0.002574 0.509967 -0.054616 -0.047515 -0.012541
-0.013641 -0.155869 0.078931 -0.006997 -0.013490
0.031247 0.075883 -0.123796 0.069329 0.009509
0.142970 -0.078219 0.019927 -0.027683 -0.016070
0.006350 0.022800 0.031360 -0.009205 -0.008627
0.169289 -0.003568 -0.063623 0.079332 0.004336
-0.068768 0.016136 -0.026307 0.000382 -0.084914
0.044794 -0.073696 -0.003742 -0.054623 0.106221
-0.066982 0.031509 -0.003578 0.046036 -0.065307
0.023652 -0.015249 -0.032593 0.047764 -0.020895
0.044653 0.066156 0.011425 14.862771 0.785629
0.301812 4.130916 0.196437 0.657537
Non-Linear Optimization, Iteration 1. Function Calls 1893.
Cosine of Angle between Direction and Gradient 0.6036428. Alpha used was 0.000000
Adjusted squared norm of gradient 19.30759
Diagnostic measure (0=perfect) 0.0000
Subiterations 1. Distance scale 1.000000000
Old Function = -2608.539051 New Function = -2599.349002
New Coefficients:
0.055356 0.507335 -0.056781 -0.050454 -0.007777
-0.011781 -0.155780 0.084212 -0.020437 -0.041736
0.027890 0.068036 -0.117581 0.041246 -0.014164
0.105471 -0.086404 0.018226 -0.019992 0.011165
0.015916 0.037352 -0.034349 -0.024337 0.027013
0.332242 0.000000 -0.094167 0.045066 0.008161
-0.050047 0.029877 -0.035787 0.013769 -0.072354
0.035727 -0.075542 -0.007173 -0.031695 0.088860
-0.020880 0.013768 -0.038687 0.033296 -0.067836
0.015371 -0.029024 -0.036324 0.041881 -0.008040
0.015062 0.060676 0.027382 14.945463 0.819208
0.000000 4.108014 0.169310 0.655544
Non-Linear Optimization, Iteration 2. Function Calls 1954.
Cosine of Angle between Direction and Gradient 0.2010777. Alpha used was 0.000000
Adjusted squared norm of gradient 8.660941
Diagnostic measure (0=perfect) 0.0000
Subiterations 1. Distance scale 1.000000000
Old Function = -2599.349002 New Function = -2596.634363
New Coefficients:
0.074416 0.501488 -0.067600 -0.061792 -0.007641
-0.010216 -0.153992 0.103579 -0.014355 -0.059174
0.021676 0.048216 -0.112987 0.021575 -0.028925
0.080070 -0.097519 0.024511 -0.008702 0.016514
0.016357 0.036656 -0.039998 -0.015240 0.043004
0.352439 0.000000 -0.046618 0.057530 0.024202
-0.046646 0.026829 -0.055762 0.021026 -0.064462
0.032103 -0.057572 0.002259 -0.008807 0.060172
-0.019467 0.012204 -0.042978 0.041637 -0.058994
0.017351 -0.027583 -0.040515 0.033590 -0.006256
-0.006670 0.052338 0.045787 14.542781 0.857614
0.000000 4.111264 0.154819 0.655882
Non-Linear Optimization, Iteration 3. Function Calls 2015.
Cosine of Angle between Direction and Gradient 0.0780976. Alpha used was 0.000000
Adjusted squared norm of gradient 6.895031
Diagnostic measure (0=perfect) 0.0000
Subiterations 1. Distance scale 1.000000000
Old Function = -2596.634363 New Function = -2594.167546
New Coefficients:
0.083703 0.509456 -0.071877 -0.069334 -0.006062
-0.009672 -0.161064 0.105383 -0.012909 -0.070766
0.020302 0.035980 -0.107331 0.017244 -0.027163
0.071929 -0.104700 0.031984 -0.001025 0.011712
0.022082 0.035854 -0.026055 -0.010232 0.036253
0.300744 0.000000 -0.036466 0.043943 0.016455
-0.052065 0.033340 -0.058015 0.036193 -0.055850
0.026906 -0.052720 -0.013253 -0.025869 0.073019
-0.020417 0.032967 -0.031669 0.043272 -0.059409
0.021111 -0.018681 -0.041689 0.034904 -0.005521
-0.007006 -0.018426 0.050201 13.868919 0.891285
0.000000 4.156036 0.157391 0.659005
Non-Linear Optimization, Iteration 27. Function Calls 3485.
Cosine of Angle between Direction and Gradient 0.1037753. Alpha used was 0.000000
Adjusted squared norm of gradient 0.1728131
Diagnostic measure (0=perfect) 0.7724
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.675250 New Function = -2588.623525
New Coefficients:
0.095899 0.524238 -0.027220 -0.119051 0.027656
-0.017375 -0.155211 0.056500 0.022205 -0.085367
0.063674 -0.014771 -0.108334 -0.004291 0.011084
0.041451 -0.108275 0.040994 0.004199 0.023873
0.037537 0.020265 -0.055681 -0.005112 0.029246
0.228057 0.000000 -0.031989 0.023866 0.044970
-0.061442 0.038157 -0.064987 0.056494 -0.082748
0.032033 -0.037939 0.012323 -0.032148 0.087160
-0.012474 0.022819 -0.037020 0.041326 -0.055407
0.031656 -0.033960 -0.012982 0.043190 -0.015109
-0.011858 -0.256649 0.077899 11.156172 0.982766
0.000000 0.822560 0.129347 0.846681
Non-Linear Optimization, Iteration 28. Function Calls 3546.
Cosine of Angle between Direction and Gradient 0.1380861. Alpha used was 0.000000
Adjusted squared norm of gradient 0.03112548
Diagnostic measure (0=perfect) 0.4635
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.623525 New Function = -2588.606437
New Coefficients:
0.096216 0.525013 -0.028507 -0.117273 0.026092
-0.017299 -0.154819 0.057367 0.021613 -0.084859
0.062868 -0.014193 -0.108342 -0.003726 0.010981
0.041400 -0.109128 0.041704 0.003784 0.024443
0.037761 0.020251 -0.055902 -0.005129 0.028963
0.230678 0.000000 -0.032645 0.024039 0.044853
-0.061384 0.038701 -0.065337 0.056465 -0.082263
0.031888 -0.037842 0.013421 -0.032111 0.087456
-0.012475 0.022429 -0.036202 0.041372 -0.055659
0.032305 -0.034142 -0.012703 0.042915 -0.015224
-0.012735 -0.266757 0.078550 11.152920 0.983035
0.000000 0.789128 0.129023 0.848427
Non-Linear Optimization, Iteration 29. Function Calls 3607.
Cosine of Angle between Direction and Gradient 0.1099209. Alpha used was 0.000000
Adjusted squared norm of gradient 0.01860308
Diagnostic measure (0=perfect) 0.2781
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.606437 New Function = -2588.593624
New Coefficients:
0.096049 0.527689 -0.030678 -0.114108 0.024130
-0.017395 -0.153497 0.057169 0.022349 -0.084860
0.063444 -0.014105 -0.108920 -0.003474 0.010523
0.041069 -0.110081 0.043103 0.003158 0.025429
0.038678 0.019679 -0.056866 -0.004638 0.028775
0.232362 0.000000 -0.033464 0.024792 0.044498
-0.061396 0.039629 -0.065708 0.056220 -0.081343
0.031873 -0.038185 0.014704 -0.032278 0.087148
-0.012402 0.022267 -0.035258 0.041599 -0.055840
0.033048 -0.034354 -0.012268 0.042830 -0.015040
-0.013391 -0.284639 0.081450 11.127535 0.984034
0.000000 0.743873 0.128685 0.851075
Non-Linear Optimization, Iteration 30. Function Calls 3668.
Cosine of Angle between Direction and Gradient 0.1691721. Alpha used was 0.000000
Adjusted squared norm of gradient 0.01406628
Diagnostic measure (0=perfect) 0.1668
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.593624 New Function = -2588.588710
New Coefficients:
0.096010 0.528449 -0.032010 -0.112945 0.023528
-0.017492 -0.152657 0.057021 0.023263 -0.085722
0.063686 -0.014062 -0.108955 -0.003283 0.010703
0.041631 -0.110629 0.043447 0.002855 0.025647
0.038392 0.019845 -0.056421 -0.004637 0.028683
0.233952 0.000000 -0.033783 0.025058 0.044110
-0.061423 0.039724 -0.065961 0.055941 -0.081099
0.031459 -0.038355 0.014862 -0.032396 0.087328
-0.012624 0.022412 -0.035569 0.041432 -0.056015
0.032973 -0.034338 -0.012509 0.042979 -0.015144
-0.013247 -0.289128 0.081189 11.131470 0.983780
0.000000 0.752197 0.128427 0.850742
Non-Linear Optimization, Iteration 31. Function Calls 3729.
Cosine of Angle between Direction and Gradient 0.1740419. Alpha used was 0.000000
Adjusted squared norm of gradient 0.008398739
Diagnostic measure (0=perfect) 0.1001
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.588710 New Function = -2588.584702
New Coefficients:
0.095843 0.529501 -0.033363 -0.111246 0.022815
-0.017597 -0.151715 0.056354 0.024495 -0.086965
0.064606 -0.013881 -0.108822 -0.003212 0.010711
0.042046 -0.110873 0.043867 0.002312 0.025749
0.038222 0.020246 -0.056083 -0.004744 0.028726
0.235773 0.000000 -0.033591 0.025278 0.043948
-0.061458 0.039604 -0.065882 0.055786 -0.080990
0.031194 -0.038491 0.014734 -0.032652 0.087369
-0.012709 0.022672 -0.035685 0.041466 -0.055961
0.032803 -0.034368 -0.012727 0.043195 -0.014982
-0.012706 -0.289962 0.081976 11.141576 0.983306
0.000000 0.764931 0.128209 0.850256
Non-Linear Optimization, Iteration 32. Function Calls 3790.
Cosine of Angle between Direction and Gradient 0.1228157. Alpha used was 0.000000
Adjusted squared norm of gradient 0.003486388
Diagnostic measure (0=perfect) 0.0601
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.584702 New Function = -2588.583851
New Coefficients:
0.095750 0.530206 -0.033756 -0.110685 0.022799
-0.017573 -0.151230 0.055522 0.025648 -0.087552
0.065456 -0.014301 -0.108872 -0.003584 0.010735
0.042100 -0.110536 0.043990 0.002258 0.025817
0.038255 0.019871 -0.056085 -0.004501 0.028762
0.235850 0.000000 -0.033460 0.025208 0.044047
-0.061549 0.039407 -0.065595 0.055766 -0.081332
0.031269 -0.038407 0.014493 -0.032639 0.087266
-0.012644 0.022623 -0.035901 0.041332 -0.055746
0.032668 -0.034296 -0.012808 0.043117 -0.015065
-0.012342 -0.292479 0.082289 11.135539 0.983340
0.000000 0.770127 0.128129 0.850151
Non-Linear Optimization, Iteration 33. Function Calls 3851.
Cosine of Angle between Direction and Gradient 0.1184611. Alpha used was 0.000000
Adjusted squared norm of gradient 0.0004671975
Diagnostic measure (0=perfect) 0.0360
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583851 New Function = -2588.583619
New Coefficients:
0.095826 0.530151 -0.033883 -0.110572 0.022753
-0.017493 -0.151313 0.055514 0.025637 -0.087601
0.065337 -0.014119 -0.108758 -0.003573 0.010716
0.042172 -0.110491 0.043963 0.002129 0.025775
0.038208 0.020024 -0.055973 -0.004673 0.028782
0.236169 0.000000 -0.033249 0.025064 0.044186
-0.061529 0.039295 -0.065472 0.055911 -0.081450
0.031364 -0.038281 0.014386 -0.032603 0.087311
-0.012619 0.022565 -0.035819 0.041412 -0.055660
0.032627 -0.034322 -0.012749 0.043077 -0.015036
-0.012407 -0.290088 0.082092 11.140185 0.983187
0.000000 0.766869 0.128094 0.850346
Non-Linear Optimization, Iteration 34. Function Calls 3912.
Cosine of Angle between Direction and Gradient 0.1295045. Alpha used was 0.000000
Adjusted squared norm of gradient 0.0001071594
Diagnostic measure (0=perfect) 0.0216
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583619 New Function = -2588.583564
New Coefficients:
0.095847 0.530103 -0.033819 -0.110586 0.022734
-0.017466 -0.151331 0.055482 0.025643 -0.087535
0.065272 -0.014108 -0.108736 -0.003564 0.010725
0.042102 -0.110448 0.043979 0.002112 0.025796
0.038202 0.019964 -0.055984 -0.004665 0.028764
0.236069 0.000000 -0.033226 0.024976 0.044243
-0.061504 0.039245 -0.065396 0.055947 -0.081540
0.031438 -0.038204 0.014369 -0.032538 0.087324
-0.012576 0.022512 -0.035798 0.041442 -0.055618
0.032658 -0.034323 -0.012682 0.043024 -0.015062
-0.012512 -0.289192 0.081928 11.138885 0.983160
0.000000 0.764623 0.128073 0.850462
Non-Linear Optimization, Iteration 35. Function Calls 3973.
Cosine of Angle between Direction and Gradient 0.0959132. Alpha used was 0.000000
Adjusted squared norm of gradient 4.232889e-005
Diagnostic measure (0=perfect) 0.0130
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583564 New Function = -2588.583544
New Coefficients:
0.095864 0.530113 -0.033825 -0.110561 0.022693
-0.017448 -0.151339 0.055465 0.025661 -0.087503
0.065245 -0.014117 -0.108761 -0.003561 0.010732
0.042088 -0.110441 0.043976 0.002121 0.025838
0.038231 0.019928 -0.055996 -0.004671 0.028759
0.236100 0.000000 -0.033215 0.024946 0.044270
-0.061476 0.039214 -0.065362 0.055961 -0.081561
0.031500 -0.038173 0.014319 -0.032504 0.087326
-0.012588 0.022479 -0.035800 0.041463 -0.055621
0.032657 -0.034323 -0.012644 0.043006 -0.015059
-0.012570 -0.288540 0.081781 11.137164 0.983179
0.000000 0.764080 0.128073 0.850494
Non-Linear Optimization, Iteration 36. Function Calls 4034.
Cosine of Angle between Direction and Gradient 0.0980764. Alpha used was 0.000000
Adjusted squared norm of gradient 2.070718e-005
Diagnostic measure (0=perfect) 0.0078
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583544 New Function = -2588.583533
New Coefficients:
0.095856 0.530123 -0.033825 -0.110521 0.022633
-0.017448 -0.151322 0.055453 0.025672 -0.087489
0.065246 -0.014128 -0.108789 -0.003547 0.010734
0.042058 -0.110454 0.043989 0.002130 0.025865
0.038240 0.019893 -0.056018 -0.004658 0.028759
0.236050 0.000000 -0.033218 0.024933 0.044257
-0.061454 0.039200 -0.065354 0.055944 -0.081571
0.031528 -0.038172 0.014302 -0.032478 0.087328
-0.012576 0.022483 -0.035799 0.041468 -0.055651
0.032656 -0.034317 -0.012618 0.043006 -0.015067
-0.012603 -0.287781 0.081664 11.135754 0.983196
0.000000 0.764295 0.128075 0.850481
Non-Linear Optimization, Iteration 37. Function Calls 4095.
Cosine of Angle between Direction and Gradient 0.1084149. Alpha used was 0.000000
Adjusted squared norm of gradient 1.173717e-005
Diagnostic measure (0=perfect) 0.0047
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583533 New Function = -2588.583527
New Coefficients:
0.095852 0.530133 -0.033823 -0.110486 0.022582
-0.017453 -0.151307 0.055455 0.025671 -0.087485
0.065266 -0.014140 -0.108806 -0.003537 0.010740
0.042044 -0.110464 0.044001 0.002150 0.025882
0.038252 0.019886 -0.056027 -0.004653 0.028765
0.236057 0.000000 -0.033228 0.024943 0.044249
-0.061438 0.039198 -0.065370 0.055926 -0.081551
0.031536 -0.038187 0.014295 -0.032481 0.087324
-0.012590 0.022478 -0.035796 0.041466 -0.055682
0.032646 -0.034328 -0.012615 0.043015 -0.015063
-0.012592 -0.287130 0.081571 11.134854 0.983211
0.000000 0.764711 0.128080 0.850456
Non-Linear Optimization, Iteration 38. Function Calls 4156.
Cosine of Angle between Direction and Gradient 0.1684459. Alpha used was 0.000000
Adjusted squared norm of gradient 3.899427e-006
Diagnostic measure (0=perfect) 0.0028
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583527 New Function = -2588.583525
New Coefficients:
0.095849 0.530129 -0.033826 -0.110474 0.022565
-0.017453 -0.151305 0.055461 0.025663 -0.087487
0.065281 -0.014153 -0.108805 -0.003539 0.010748
0.042043 -0.110463 0.043998 0.002159 0.025880
0.038250 0.019887 -0.056033 -0.004656 0.028771
0.236065 0.000000 -0.033225 0.024943 0.044240
-0.061441 0.039206 -0.065385 0.055924 -0.081548
0.031524 -0.038199 0.014305 -0.032488 0.087326
-0.012590 0.022489 -0.035790 0.041458 -0.055694
0.032640 -0.034328 -0.012626 0.043022 -0.015069
-0.012577 -0.286838 0.081541 11.134762 0.983218
0.000000 0.764857 0.128082 0.850447
Non-Linear Optimization, Iteration 39. Function Calls 4217.
Cosine of Angle between Direction and Gradient 0.2833513. Alpha used was 0.000000
Adjusted squared norm of gradient 8.387825e-007
Diagnostic measure (0=perfect) 0.0017
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583525 New Function = -2588.583525
New Coefficients:
0.095850 0.530128 -0.033821 -0.110473 0.022563
-0.017453 -0.151306 0.055462 0.025657 -0.087485
0.065292 -0.014163 -0.108796 -0.003543 0.010752
0.042042 -0.110455 0.043998 0.002164 0.025876
0.038248 0.019888 -0.056032 -0.004658 0.028772
0.236066 0.000000 -0.033225 0.024945 0.044244
-0.061445 0.039213 -0.065394 0.055928 -0.081543
0.031514 -0.038204 0.014316 -0.032498 0.087326
-0.012595 0.022485 -0.035789 0.041456 -0.055691
0.032639 -0.034336 -0.012638 0.043023 -0.015070
-0.012566 -0.286732 0.081527 11.134812 0.983218
0.000000 0.764829 0.128082 0.850449
Non-Linear Optimization, Iteration 40. Function Calls 4278.
Cosine of Angle between Direction and Gradient 0.2353028. Alpha used was 0.000000
Adjusted squared norm of gradient 2.31118e-007
Diagnostic measure (0=perfect) 0.0010
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583525 New Function = -2588.583525
New Coefficients:
0.095852 0.530126 -0.033822 -0.110474 0.022566
-0.017451 -0.151308 0.055462 0.025655 -0.087481
0.065294 -0.014168 -0.108791 -0.003545 0.010755
0.042041 -0.110451 0.043996 0.002166 0.025875
0.038246 0.019888 -0.056032 -0.004659 0.028772
0.236075 0.000000 -0.033224 0.024943 0.044246
-0.061448 0.039217 -0.065396 0.055933 -0.081545
0.031510 -0.038203 0.014321 -0.032502 0.087326
-0.012595 0.022485 -0.035789 0.041457 -0.055687
0.032641 -0.034337 -0.012642 0.043022 -0.015071
-0.012567 -0.286718 0.081524 11.134813 0.983218
0.000000 0.764773 0.128081 0.850452
Non-Linear Optimization, Iteration 41. Function Calls 4339.
Cosine of Angle between Direction and Gradient 0.3163347. Alpha used was 0.000000
Adjusted squared norm of gradient 4.465044e-008
Diagnostic measure (0=perfect) 0.0006
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583525 New Function = -2588.583525
New Coefficients:
0.095852 0.530128 -0.033822 -0.110474 0.022567
-0.017451 -0.151308 0.055461 0.025654 -0.087479
0.065294 -0.014169 -0.108790 -0.003546 0.010755
0.042041 -0.110450 0.043996 0.002165 0.025875
0.038245 0.019886 -0.056032 -0.004659 0.028771
0.236072 0.000000 -0.033223 0.024942 0.044248
-0.061449 0.039218 -0.065396 0.055935 -0.081547
0.031511 -0.038203 0.014322 -0.032502 0.087326
-0.012595 0.022484 -0.035790 0.041459 -0.055686
0.032642 -0.034338 -0.012642 0.043021 -0.015071
-0.012570 -0.286719 0.081523 11.134805 0.983218
0.000000 0.764757 0.128081 0.850453
Non-Linear Optimization, Iteration 42. Function Calls 4400.
Cosine of Angle between Direction and Gradient 0.1460616. Alpha used was 0.000000
Adjusted squared norm of gradient 9.944952e-009
Diagnostic measure (0=perfect) 0.0004
Subiterations 1. Distance scale 1.000000000
Old Function = -2588.583525 New Function = -2588.583525
New Coefficients:
0.095852 0.530128 -0.033823 -0.110474 0.022568
-0.017452 -0.151307 0.055461 0.025653 -0.087478
0.065294 -0.014170 -0.108789 -0.003546 0.010756
0.042041 -0.110450 0.043996 0.002165 0.025875
0.038246 0.019887 -0.056032 -0.004659 0.028771
0.236073 0.000000 -0.033223 0.024942 0.044248
-0.061449 0.039218 -0.065396 0.055936 -0.081547
0.031512 -0.038203 0.014323 -0.032503 0.087326
-0.012595 0.022484 -0.035791 0.041459 -0.055685
0.032643 -0.034338 -0.012642 0.043021 -0.015071
-0.012571 -0.286718 0.081523 11.134788 0.983218
0.000000 0.764757 0.128081 0.850453
MAXIMIZE - Estimation by BFGS
Convergence in 42 Iterations. Final criterion was 0.0000016 <= 0.0000100
Monthly Data From 1974:01 To 2009:12
Usable Observations 432
Function Value -2588.5835
Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. B 0.09585250 0.03127439 3.06489 0.00217752
2. BVEC(1)(1) 0.53012820 0.05028247 10.54300 0.00000000
3. BVEC(1)(2) -0.03382281 0.03860605 -0.87610 0.38097509
4. BVEC(1)(3) -0.11047403 0.04279874 -2.58124 0.00984447
5. BVEC(1)(4) 0.02256768 0.04294121 0.52555 0.59920207
6. BVEC(1)(5) -0.01745154 0.02681079 -0.65091 0.51510146
7. BVEC(1)(6) -0.15130741 0.03258400 -4.64361 0.00000342
8. BVEC(1)(7) 0.05546113 0.03518137 1.57643 0.11492564
9. BVEC(1)(8) 0.02565334 0.04283871 0.59884 0.54928255
10. BVEC(1)(9) -0.08747772 0.03943955 -2.21802 0.02655347
11. BVEC(1)(10) 0.06529450 0.03946081 1.65467 0.09799209
12. BVEC(1)(11) -0.01416972 0.03790634 -0.37381 0.70854675
13. BVEC(1)(12) -0.10878948 0.02934713 -3.70699 0.00020974
14. BVEC(1)(13) -0.00354579 0.03691264 -0.09606 0.92347369
15. BVEC(1)(14) 0.01075567 0.04309607 0.24957 0.80291658
16. BVEC(1)(15) 0.04204084 0.04085463 1.02903 0.30346331
17. BVEC(1)(16) -0.11045016 0.03602444 -3.06598 0.00216959
18. BVEC(1)(17) 0.04399617 0.03620323 1.21526 0.22426875
19. BVEC(1)(18) 0.00216526 0.04073171 0.05316 0.95760513
20. BVEC(1)(19) 0.02587499 0.04265024 0.60668 0.54406413
21. BVEC(1)(20) 0.03824552 0.04041933 0.94622 0.34403705
22. BVEC(1)(21) 0.01988663 0.04163203 0.47768 0.63288071
23. BVEC(1)(22) -0.05603180 0.04212411 -1.33016 0.18346564
24. BVEC(1)(23) -0.00465893 0.03895077 -0.11961 0.90479144
25. BVEC(1)(24) 0.02877134 0.03609311 0.79714 0.42536840
26. BVEC(1)(25) 0.23607305 0.18802043 1.25557 0.20927144
27. BVEC(1)(26) 0.00000000 0.00000000 0.00000 0.00000000
28. BVEC(2)(1) -0.03322330 0.03748393 -0.88633 0.37543728
29. BVEC(2)(2) 0.02494214 0.03483301 0.71605 0.47396120
30. BVEC(2)(3) 0.04424796 0.03268036 1.35396 0.17574850
31. BVEC(2)(4) -0.06144929 0.03087430 -1.99031 0.04655731
32. BVEC(2)(5) 0.03921773 0.03170393 1.23700 0.21608746
33. BVEC(2)(6) -0.06539563 0.02848592 -2.29572 0.02169205
34. BVEC(2)(7) 0.05593624 0.03049274 1.83441 0.06659292
35. BVEC(2)(8) -0.08154739 0.03420693 -2.38394 0.01712822
36. BVEC(2)(9) 0.03151176 0.02892760 1.08933 0.27600745
37. BVEC(2)(10) -0.03820279 0.03021077 -1.26454 0.20603543
38. BVEC(2)(11) 0.01432259 0.03342856 0.42845 0.66832083
39. BVEC(2)(12) -0.03250296 0.03313018 -0.98107 0.32655930
40. BVEC(2)(13) 0.08732603 0.05644643 1.54706 0.12184870
41. BVEC(2)(14) -0.01259509 0.05471921 -0.23018 0.81795444
42. BVEC(2)(15) 0.02248448 0.04855820 0.46304 0.64333428
43. BVEC(2)(16) -0.03579113 0.04925730 -0.72662 0.46746135
44. BVEC(2)(17) 0.04145924 0.04796702 0.86433 0.38740773
45. BVEC(2)(18) -0.05568539 0.04603051 -1.20975 0.22637500
46. BVEC(2)(19) 0.03264276 0.05147225 0.63418 0.52596227
47. BVEC(2)(20) -0.03433788 0.04887338 -0.70259 0.48231213
48. BVEC(2)(21) -0.01264198 0.03402305 -0.37157 0.71021234
49. BVEC(2)(22) 0.04302130 0.04149468 1.03679 0.29983343
50. BVEC(2)(23) -0.01507078 0.04863744 -0.30986 0.75666767
51. BVEC(2)(24) -0.01257103 0.04561292 -0.27560 0.78285346
52. BVEC(2)(25) -0.28671814 0.40142566 -0.71425 0.47507283
53. BVEC(2)(26) 0.08152284 0.06178002 1.31957 0.18697983
54. GARCHP(1)(1) 11.13478778 1.41768408 7.85421 0.00000000
55. GARCHP(1)(2) 0.98321750 0.02284533 43.03800 0.00000000
56. GARCHP(1)(3) 0.00000000 0.00000000 0.00000 0.00000000
57. GARCHP(2)(1) 0.76475712 0.47982621 1.59382 0.11097611
58. GARCHP(2)(2) 0.12808063 0.01898348 6.74695 0.00000000
59. GARCHP(2)(3) 0.85045337 0.02957501 28.75581 0.00000000
SIC for VAR 5632.52461
SIC for GARCH-M 5523.06731I believe line 53. of the last table shows the effect of oil price uncertainty on stock returns with a positive coefficient of 0.08152284 and t-statistics of 1.31957.
In regard of testing the OLS residuals for GARCH effects, I am not sure whether you meant what I got in the first table above!
Is there any RATS wizard for testing the OLS residuals for GARCH effects?
Thanks a lot,
Last edited by economics2012 on Wed May 30, 2012 1:03 pm, edited 1 time in total.
Re: VAR-GARCH-M
Note that what you're estimating isn't a standard VAR-GARCH-M, but Elder's specialized version. The "weird" result is a positive but insignificant coefficient so you probably shouldn't be that concerned. BTW, you are including in your data set the 1970's when (U.S.) oil prices were controlled and the stock market was weak.
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
Hi Tom,
Correct, I am following Elder's representation. I am examining oil price uncertainty on aggregate and sectoral level stock returns.
Results for automobile, retail and steel returns, for instance, show positive and significant results! I will attach the data set for some of the sectors.
What do you suggest in regards of the data?
thanks a lot
Correct, I am following Elder's representation. I am examining oil price uncertainty on aggregate and sectoral level stock returns.
Results for automobile, retail and steel returns, for instance, show positive and significant results! I will attach the data set for some of the sectors.
What do you suggest in regards of the data?
thanks a lot
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
Hi Tom,
Attached are the data sets for three different returns, the automobile, steel and retail.
Best,
Attached are the data sets for three different returns, the automobile, steel and retail.
Best,
Last edited by economics2012 on Wed May 30, 2012 1:04 pm, edited 1 time in total.
Re: VAR-GARCH-M
A GARCH model doesn't describe the dynamics of the oil growth series very well. Instead of volatility clustering, you get about four really large spikes (two up, two down). Feed the generated standard deviation series into a regression and you have basically a dummy variable with four 1's and everything else 0's. That will be the same regardless of the other series that you use for stock returns. You're never going to get good statistical inference from a 4 data point dummy.
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
So you mean the problem is in the real price of oil that I cannot estimate a GARCH model? Elder and Serletis(2010), in their paper "Oil Price Uncertainty", Journal of Money, Credit and Banking, Vol. 42, No. 6, 1137-1159, have used the oil price-GDP bivariate VAR-GARCH-M model. They used quarterly data while mine is monthly.
Last edited by economics2012 on Mon Feb 20, 2012 8:51 pm, edited 1 time in total.
Re: VAR-GARCH-M
I'm not sure their GARCH model fit the oil price any better than yours did, but they are working with a target variable of quarterly GDP, which doesn't react as quickly as your stock price series. If the GARCH model doesn't fit oil prices well, then adding the predicted variance from a poorly-fitting model isn't likely to give useful information, plus much of the information from the past oil prices may already be priced in.
At any rate, the calculations are correct given your data.
At any rate, the calculations are correct given your data.
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
Dear Tom,
I re-ran the GARCH model using daily data instead of monthly data. The results show POSITIVE and significant effect of oil price uncertainty on the Retail returns, and Entertainment returns. However, The results are supposed to be either negative or insignificant. For the aggregate returns and the other sectors, the results came out insignificant.
I attached below the data, code and results for the retail sector.
Your help is greatly appreciated,
I re-ran the GARCH model using daily data instead of monthly data. The results show POSITIVE and significant effect of oil price uncertainty on the Retail returns, and Entertainment returns. However, The results are supposed to be either negative or insignificant. For the aggregate returns and the other sectors, the results came out insignificant.
I attached below the data, code and results for the retail sector.
Your help is greatly appreciated,
Last edited by economics2012 on Wed May 02, 2012 12:42 pm, edited 2 times in total.
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
I need to check whether the mean in the above model is correctly specified or I still have GARCH effects using the daily data.
Any help is greatly appreciated,
Any help is greatly appreciated,
Re: VAR-GARCH-M
You copied a constraint which was specific to the original example which had GDP (which has no GARCH effect) as one of the variables. So you were knocking the GARCH term out of the stock market returns. If you take that out:
nonlin b bvec garchp bvec(1)(38)=0.0
you get a better fitting model. However, you still end up with a positive and (barely) significant coefficient on the oil price volatility. However, other than the error I just pointed out the model is set up correctly.
nonlin b bvec garchp bvec(1)(38)=0.0
you get a better fitting model. However, you still end up with a positive and (barely) significant coefficient on the oil price volatility. However, other than the error I just pointed out the model is set up correctly.
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
So why "nonlin b bvec garchp bvec(1)(38)=0.0" is specific for GDP and not for stock returns?
Re: VAR-GARCH-M
Sorry for the confusion. You copied the following out of the original example (with adjustment for the number of lags)
*
* In an unconstrained estimation, one of the GARCH coefficients goes
* negative. We peg it to zero. The model estimated in the paper also
* constrains the "M" term on the oil price in the oil (1st) equation to
* zero.
*
nonlin b bvec garchp garchp(1)(3)=0.0 bvec(1)(38)=0.0
The first constaint was specific to the Elder-Serletis example. When we estimated the model without it, that parameter went negative, so we re-estimated it constrained. Your data doesn't need it, so your nonlin should read
nonlin b bvec garchp bvec(1)(38)=0.0
*
* In an unconstrained estimation, one of the GARCH coefficients goes
* negative. We peg it to zero. The model estimated in the paper also
* constrains the "M" term on the oil price in the oil (1st) equation to
* zero.
*
nonlin b bvec garchp garchp(1)(3)=0.0 bvec(1)(38)=0.0
The first constaint was specific to the Elder-Serletis example. When we estimated the model without it, that parameter went negative, so we re-estimated it constrained. Your data doesn't need it, so your nonlin should read
nonlin b bvec garchp bvec(1)(38)=0.0
-
economics2012
- Posts: 51
- Joined: Thu Jan 19, 2012 4:41 pm
Re: VAR-GARCH-M
After editing the error, the coefficient for the retail sector stays positive and highly significant coefficient on the oil price volatility as shown in the attached results. Mainly under line 77 of the last table with a t-stat of 3.21.
77. BVEC(2)(38) 0.041078447 0.012770907 3.21656 0.00129735
Also, after using nonlin b bvec garchp bvec(1)(38)=0.0, I got positive and significant coefficient on the oil price volatility for the aggregate level, which used to be insignificant before adjusting it.
How did you get barely significance? Do I have other errors?
77. BVEC(2)(38) 0.041078447 0.012770907 3.21656 0.00129735
Also, after using nonlin b bvec garchp bvec(1)(38)=0.0, I got positive and significant coefficient on the oil price volatility for the aggregate level, which used to be insignificant before adjusting it.
How did you get barely significance? Do I have other errors?
Last edited by economics2012 on Wed May 02, 2012 12:34 pm, edited 1 time in total.