STARBUCKS
PRECIO DE LA ACCION DE STARBUCKS Y SU COMPORTAMIENTO
Figura 1. Precio de Cierre de STARBUCKS(SBUX): enero 2015 - marzo 2020
Fuente: elaboracion propia realizada en RGRÁFICA EN MODALIDAD DE RENDIMIENTOS DE STARBUCKS
Modelos ARCH
primera prueba ARCH
Tabla 1. Prueba de efectos ARCH
| Prueba | Valor p | Ho | Resultado |
|---|---|---|---|
| ARCH test | 2.2e-16 | No existe ausencia de efectos ARCH | RECHAZO H0 |
Fuente: elaboracion propia realizada en R
MODELO ARCH 2
Figura 4: MODELO ARCH(2)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(2,0)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000118 0.000007 17.8458 0.0e+00
alpha1 0.378990 0.055085 6.8801 0.0e+00
alpha2 0.144385 0.034722 4.1584 3.2e-05
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000118 0.000019 6.0821 0.000000
alpha1 0.378990 0.088286 4.2928 0.000018
alpha2 0.144385 0.071977 2.0060 0.044858
LogLikelihood : 3846.895
Information Criteria
------------------------------------
Akaike -5.7414
Bayes -5.7298
Shibata -5.7415
Hannan-Quinn -5.7371
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.5502 0.4582
Lag[2*(p+q)+(p+q)-1][2] 0.5516 0.6699
Lag[4*(p+q)+(p+q)-1][5] 1.7060 0.6893
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.4234 0.5153
Lag[2*(p+q)+(p+q)-1][5] 1.6459 0.7040
Lag[4*(p+q)+(p+q)-1][9] 2.8388 0.7853
d.o.f=2
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[3] 1.073 0.500 2.000 0.3003
ARCH Lag[5] 1.172 1.440 1.667 0.6828
ARCH Lag[7] 2.325 2.315 1.543 0.6488
Nyblom stability test
------------------------------------
Joint Statistic: 1.442
Individual Statistics:
omega 0.1925
alpha1 0.3772
alpha2 1.2645
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 0.846 1.01 1.35
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.5659 0.5715
Negative Sign Bias 0.6426 0.5206
Positive Sign Bias 0.2175 0.8278
Joint Effect 0.5181 0.9149
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 100.5 4.334e-13
2 30 111.3 1.395e-11
3 40 127.1 2.913e-11
4 50 118.8 1.006e-07
Elapsed time : 1.312842
MODELO ARCH 3
Figura 5: MODELO ARCH(3)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(3,0)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000116 0.000007 17.2957 0.000000
alpha1 0.313152 0.051670 6.0606 0.000000
alpha2 0.075293 0.031812 2.3668 0.017943
alpha3 0.098584 0.033669 2.9280 0.003412
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000116 0.000020 5.6929 0.000000
alpha1 0.313152 0.089592 3.4953 0.000474
alpha2 0.075293 0.057593 1.3073 0.191099
alpha3 0.098584 0.051940 1.8981 0.057689
LogLikelihood : 3855.821
Information Criteria
------------------------------------
Akaike -5.7533
Bayes -5.7377
Shibata -5.7533
Hannan-Quinn -5.7475
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.6169 0.4322
Lag[2*(p+q)+(p+q)-1][2] 0.6228 0.6387
Lag[4*(p+q)+(p+q)-1][5] 1.4799 0.7449
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.1683 0.6817
Lag[2*(p+q)+(p+q)-1][8] 1.0348 0.9716
Lag[4*(p+q)+(p+q)-1][14] 1.7327 0.9953
d.o.f=3
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[4] 0.1018 0.500 2.000 0.7496
ARCH Lag[6] 0.8734 1.461 1.711 0.7843
ARCH Lag[8] 1.0095 2.368 1.583 0.9229
Nyblom stability test
------------------------------------
Joint Statistic: 1.9957
Individual Statistics:
omega 0.1989
alpha1 0.1358
alpha2 1.5637
alpha3 1.0110
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 1.07 1.24 1.6
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.496968 0.6193
Negative Sign Bias 0.534470 0.5931
Positive Sign Bias 0.007762 0.9938
Joint Effect 0.395995 0.9411
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 85.3 2.230e-10
2 30 109.1 3.271e-11
3 40 100.2 2.753e-07
4 50 115.6 2.675e-07
Elapsed time : 0.685998
MODELO ARCH 4
Figura 6. MODELO ARCH(4)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(4,0)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000112 0.000007 15.2758 0.000000
alpha1 0.317708 0.051900 6.1216 0.000000
alpha2 0.048226 0.033041 1.4596 0.144414
alpha3 0.090772 0.034176 2.6560 0.007907
alpha4 0.059416 0.046837 1.2686 0.204596
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000112 0.000021 5.40873 0.000000
alpha1 0.317708 0.092964 3.41752 0.000632
alpha2 0.048226 0.054941 0.87777 0.380070
alpha3 0.090772 0.053873 1.68492 0.092004
alpha4 0.059416 0.113765 0.52227 0.601484
LogLikelihood : 3856.762
Information Criteria
------------------------------------
Akaike -5.7532
Bayes -5.7338
Shibata -5.7532
Hannan-Quinn -5.7459
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.4617 0.4969
Lag[2*(p+q)+(p+q)-1][2] 0.4776 0.7042
Lag[4*(p+q)+(p+q)-1][5] 1.4351 0.7559
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.1901 0.6628
Lag[2*(p+q)+(p+q)-1][11] 1.4374 0.9869
Lag[4*(p+q)+(p+q)-1][19] 2.5977 0.9978
d.o.f=4
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[5] 0.04041 0.500 2.000 0.8407
ARCH Lag[7] 0.91381 1.473 1.746 0.7816
ARCH Lag[9] 1.02768 2.402 1.619 0.9275
Nyblom stability test
------------------------------------
Joint Statistic: 2.1278
Individual Statistics:
omega 0.1981
alpha1 0.1196
alpha2 1.6414
alpha3 0.9528
alpha4 0.4610
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 1.28 1.47 1.88
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.47629 0.6339
Negative Sign Bias 0.47925 0.6318
Positive Sign Bias 0.04499 0.9641
Joint Effect 0.33117 0.9541
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 90.77 2.424e-11
2 30 120.87 3.500e-13
3 40 105.27 5.286e-08
4 50 121.46 4.378e-08
Elapsed time : 0.639015
Modelos GARCH
Ajuste con MODELO GARCH (1,1):
Figura 7: MODELO GARCH(1,1) - PRESENTACION DE RESULTADOS OBTENIDOS (PARAMETROS)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(1,1)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000045 0.000011 4.0661 4.8e-05
alpha1 0.212572 0.047219 4.5019 7.0e-06
beta1 0.582188 0.086889 6.7004 0.0e+00
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000045 0.00003 1.5116 0.130645
alpha1 0.212572 0.13277 1.6010 0.109372
beta1 0.582188 0.23839 2.4422 0.014597
LogLikelihood : 3858.577
Information Criteria
------------------------------------
Akaike -5.7589
Bayes -5.7472
Shibata -5.7589
Hannan-Quinn -5.7545
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.4819 0.4876
Lag[2*(p+q)+(p+q)-1][2] 0.4822 0.7020
Lag[4*(p+q)+(p+q)-1][5] 1.4740 0.7463
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.006766 0.9344
Lag[2*(p+q)+(p+q)-1][5] 0.664725 0.9296
Lag[4*(p+q)+(p+q)-1][9] 1.003602 0.9861
d.o.f=2
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[3] 0.07776 0.500 2.000 0.7804
ARCH Lag[5] 0.35914 1.440 1.667 0.9242
ARCH Lag[7] 0.51640 2.315 1.543 0.9769
Nyblom stability test
------------------------------------
Joint Statistic: 1.2811
Individual Statistics:
omega 0.1520
alpha1 0.4544
beta1 0.2684
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 0.846 1.01 1.35
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.5246 0.6000
Negative Sign Bias 0.7936 0.4276
Positive Sign Bias 0.3400 0.7339
Joint Effect 0.9217 0.8202
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 88.47 6.190e-11
2 30 113.08 7.153e-12
3 40 120.09 3.448e-10
4 50 132.14 1.453e-09
Elapsed time : 0.635998
Fuente: elaboracion propia realizada en R
Ajuste con un MODELO GARCH (1,2)
Figura 8: MODELO GARCH(1,2) - PRESENTACION DE RESULTADOS OBTENIDOS (PARAMETROS)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(1,2)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000042 0.000010 4.0231 0.000057
alpha1 0.235338 0.044258 5.3174 0.000000
beta1 0.107459 0.071305 1.5070 0.131805
beta2 0.464981 0.092897 5.0053 0.000001
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000042 0.000025 1.6578 0.097356
alpha1 0.235338 0.119495 1.9694 0.048904
beta1 0.107459 0.098828 1.0873 0.276892
beta2 0.464981 0.180632 2.5742 0.010047
LogLikelihood : 3865.585
Information Criteria
------------------------------------
Akaike -5.7679
Bayes -5.7523
Shibata -5.7679
Hannan-Quinn -5.7620
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.3092 0.5782
Lag[2*(p+q)+(p+q)-1][2] 0.3108 0.7898
Lag[4*(p+q)+(p+q)-1][5] 1.0830 0.8406
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.01322 0.9085
Lag[2*(p+q)+(p+q)-1][8] 0.50506 0.9962
Lag[4*(p+q)+(p+q)-1][14] 0.97529 0.9997
d.o.f=3
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[4] 0.2179 0.500 2.000 0.6407
ARCH Lag[6] 0.2945 1.461 1.711 0.9466
ARCH Lag[8] 0.6441 2.368 1.583 0.9689
Nyblom stability test
------------------------------------
Joint Statistic: 1.3613
Individual Statistics:
omega 0.1624
alpha1 0.3580
beta1 0.2633
beta2 0.2267
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 1.07 1.24 1.6
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.4907 0.6237
Negative Sign Bias 0.5989 0.5493
Positive Sign Bias 0.2727 0.7851
Joint Effect 0.6331 0.8888
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 88.08 7.247e-11
2 30 115.99 2.333e-12
3 40 113.40 3.473e-09
4 50 137.51 2.472e-10
Elapsed time : 0.713994
Fuente: elaboracion propia realizada en R
Ajuste con un MODELO GARCH(2,1)
Figura 9: MODELO GARCH 2,1 - PRESENTACION DE RESULTADOS OBTENIDOS (PARAMETROS)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(2,1)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000045 0.000025 1.7660 0.077398
alpha1 0.212370 0.036959 5.7461 0.000000
alpha2 0.000000 0.092111 0.0000 1.000000
beta1 0.582136 0.217228 2.6798 0.007366
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000045 0.000083 0.54212 0.587734
alpha1 0.212370 0.067231 3.15881 0.001584
alpha2 0.000000 0.344928 0.00000 1.000000
beta1 0.582136 0.683990 0.85109 0.394720
LogLikelihood : 3858.56
Information Criteria
------------------------------------
Akaike -5.7574
Bayes -5.7418
Shibata -5.7574
Hannan-Quinn -5.7516
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.4828 0.4872
Lag[2*(p+q)+(p+q)-1][2] 0.4831 0.7016
Lag[4*(p+q)+(p+q)-1][5] 1.4760 0.7458
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.006653 0.9350
Lag[2*(p+q)+(p+q)-1][8] 0.923096 0.9791
Lag[4*(p+q)+(p+q)-1][14] 1.428476 0.9980
d.o.f=3
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[4] 0.3312 0.500 2.000 0.5650
ARCH Lag[6] 0.4651 1.461 1.711 0.9017
ARCH Lag[8] 0.6773 2.368 1.583 0.9655
Nyblom stability test
------------------------------------
Joint Statistic: 6.9323
Individual Statistics:
omega 0.1522
alpha1 0.4573
alpha2 1.8651
beta1 0.2694
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 1.07 1.24 1.6
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.5239 0.6005
Negative Sign Bias 0.7952 0.4266
Positive Sign Bias 0.3420 0.7324
Joint Effect 0.9251 0.8194
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 88.44 6.265e-11
2 30 112.72 8.205e-12
3 40 119.73 3.908e-10
4 50 132.14 1.453e-09
Elapsed time : 0.5830059
Ajuste con un MODELO GARCH(2,2)
Figura 10: GARCH(2,2) - PRESENTACION DE RESULTADOS OBTENIDOS (PARAMETROS)
*---------------------------------*
* GARCH Model Fit *
*---------------------------------*
Conditional Variance Dynamics
-----------------------------------
GARCH Model : sGARCH(2,2)
Mean Model : ARFIMA(0,0,0)
Distribution : norm
Optimal Parameters
------------------------------------
Estimate Std. Error t value Pr(>|t|)
omega 0.000042 0.000021 2.034986 0.041852
alpha1 0.235328 0.044785 5.254578 0.000000
alpha2 0.000000 0.068463 0.000003 0.999997
beta1 0.107457 0.177936 0.603909 0.545904
beta2 0.464999 0.093334 4.982109 0.000001
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
omega 0.000042 0.000065 0.645235 0.518775
alpha1 0.235328 0.133451 1.763398 0.077833
alpha2 0.000000 0.183190 0.000001 0.999999
beta1 0.107457 0.469906 0.228678 0.819120
beta2 0.464999 0.166234 2.797251 0.005154
LogLikelihood : 3865.585
Information Criteria
------------------------------------
Akaike -5.7664
Bayes -5.7470
Shibata -5.7664
Hannan-Quinn -5.7591
Weighted Ljung-Box Test on Standardized Residuals
------------------------------------
statistic p-value
Lag[1] 0.3092 0.5782
Lag[2*(p+q)+(p+q)-1][2] 0.3108 0.7898
Lag[4*(p+q)+(p+q)-1][5] 1.0830 0.8406
d.o.f=0
H0 : No serial correlation
Weighted Ljung-Box Test on Standardized Squared Residuals
------------------------------------
statistic p-value
Lag[1] 0.01321 0.9085
Lag[2*(p+q)+(p+q)-1][11] 0.75633 0.9988
Lag[4*(p+q)+(p+q)-1][19] 1.52925 0.9999
d.o.f=4
Weighted ARCH LM Tests
------------------------------------
Statistic Shape Scale P-Value
ARCH Lag[5] 0.05117 0.500 2.000 0.8210
ARCH Lag[7] 0.27075 1.473 1.746 0.9551
ARCH Lag[9] 0.61296 2.402 1.619 0.9751
Nyblom stability test
------------------------------------
Joint Statistic: 3.5422
Individual Statistics:
omega 0.1624
alpha1 0.3580
alpha2 1.2623
beta1 0.2633
beta2 0.2267
Asymptotic Critical Values (10% 5% 1%)
Joint Statistic: 1.28 1.47 1.88
Individual Statistic: 0.35 0.47 0.75
Sign Bias Test
------------------------------------
t-value prob sig
Sign Bias 0.4907 0.6238
Negative Sign Bias 0.5989 0.5493
Positive Sign Bias 0.2728 0.7851
Joint Effect 0.6332 0.8888
Adjusted Pearson Goodness-of-Fit Test:
------------------------------------
group statistic p-value(g-1)
1 20 88.08 7.247e-11
2 30 115.99 2.333e-12
3 40 113.40 3.473e-09
4 50 137.51 2.472e-10
Elapsed time : 0.5679951
Fuente: elaboracion propia realizada en R
Selección de modelo y simulación de los rendimientos
| MODELO | \(\omega\) | \(\alpha_{1}\) | \(\alpha_{2}\) | \(\alpha_{3}\) | \(\alpha_{4}\) | \(\beta_{1}\) | \(\beta_{2}\) | AKAIKE | BAYES |
|---|---|---|---|---|---|---|---|---|---|
| ARCH(1) | 0.000 | 0.050 | 67.665 | 67.672 | |||||
| ARCH(2) | 0.000 | 0.3789 | 0.1443 | -5.7414 | -5.7298 | ||||
| ARCH(3) | 0.000 | 0.3131 | 0.0752 | 0.0985 | -5.7533 | -5.7377 | |||
| ARCH(4) | 0.000 | 0.3177 | 0.0482 | 0.0907 | 0.0594 | -5.7532 | -5.7338 | ||
| GARCH(1,1) | 0.000 | 0.2125 | 0.5821 | -5.7589 | -5.7472 | ||||
| GARCH(1,2) | 0.000 | 0.2353 | 0.1074 | 0.4649 | -5.7679 | -5.7523 | |||
| GARCH(2,1) | 0.000 | 0.2123 | 0.0000 | 0.5821 | -5.7574 | -5.7418 | |||
| GARCH(2,2) | 0.000 | 0.235 | 0.0000 | 0.1074 | 0.4649 | -5.7664 | -5.7470 |
Por medio del método estadístico comparando los resultados de los parámetros obtenido en las pruebas de AKAIKE Y BAYES, se llega a la conclusion de que los mejores modelos son ARCH(3) y GARCH(1,2) respectivamente, ya que, además, sus respectivas graficas de rendimiento-volatilidad se ajustan de una mejor manera en comparación con las demás.
Resultados de la simulación
Figura 11. Los rendimientos reales contra los simulados: MODELOS ARCH(3) y GARCH(1,2) como las mejores opciones
Fuente: elaboracion propia realizada en R