• Ejercicio Autocorrelación (1)
“Desempleo y pobreza afligen el mundo”, así de contundente se mostraba un titular
de CNNExpansión[i] en febrero de 2011. Junto con la injusticia social, la pobreza y
el desempleo se encuentran entre las mayores preocupaciones de la población
según la encuesta Ipsos/Reuters a 18,676 adultos en 24 países de Asia, África,
Europa, Norteamérica y Sudamérica. La falta de puestos de trabajo genera pobreza
y estas son dos de las grandes preocupaciones mundiales que no han cambiado
desde 2010.
Con el fin de valorar empíricamente esta última afirmación se han estimado distintas
propuestas que muestran la relación entre los índices de pobreza y desempleo de
la economía americana durante el periodo 1980- 2003.
Determine para las distintas estructuras, cuál es la estimación más adecuada.
Se ha estimado un primer modelo que considere como único origen de la pobreza
el desempleo y otro modelo alternativo en el que además del desempleo sea la
propia pobreza del periodo anterior la que determine el nivel de pobreza actual.
Indique cuál es la estimación más adecuada para cada uno de estos modelos.
Identifique la utilidad de cada uno de los resultados presentados.
Primer modelo: único origen de la pobreza el desempleo
Estimación 1
Source | SS df MS Number of obs = 24
-------------+---------------------------------- F(1, 22) = 38.35
Model | 17.5384184 1 17.5384184 Prob > F = 0.0000
Residual | 10.0611652 22 .457325692 R-squared = 0.6355
-------------+---------------------------------- Adj R-squared = 0.6189
Total | 27.5995836 23 1.1999819 Root MSE = .67626
------------------------------------------------------------------------------
poverty | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .5866142 .0947262 6.19 0.000 .3901641 .7830644
_cons | 9.792052 .6111865 16.02 0.000 8.524528 11.05957
Estimación 2
Regression with Newey–West standard errors Number of obs = 24
Maximum lag = 1 F( 1, 22) = 37.43
Prob > F = 0.0000
------------------------------------------------------------------------------
| Newey–West
poverty | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .5866142 .0958879 6.12 0.000 .3877549 .7854736
_cons | 9.792052 .6704428 14.61 0.000 8.401638 11.18247
Estimación 3
Cochrane–Orcutt AR(1) regression with iterated estimates
Source | SS df MS Number of obs = 23
-------------+---------------------------------- F(1, 21) = 35.73
Model | 4.94418359 1 4.94418359 Prob > F = 0.0000
Residual | 2.90574405 21 .138368764 R-squared = 0.6298
-------------+---------------------------------- Adj R-squared = 0.6122
Total | 7.84992765 22 .356814893 Root MSE = .37198
------------------------------------------------------------------------------
poverty | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .5834373 .0976036 5.98 0.000 .3804594 .7864151
_cons | 9.890137 .7132942 13.87 0.000 8.406761 11.37351
-------------+----------------------------------------------------------------
rho | .8074629
------------------------------------------------------------------------------
Durbin–Watson statistic (original) = 0.323725
Durbin–Watson statistic (transformed) = 1.724004
Información adicional del modelo estimado por MCO (estimación 1)
Durbin–Watson d-statistic( 2, 24) = .3237247
Breusch–Godfrey LM test for autocorrelation
---------------------------------------------------------------------------
lags(p) | chi2 df Prob > chi2
-------------+-------------------------------------------------------------
1 | 15.353 1 0.0001
2 | 16.033 2 0.0003
3 | 16.060 3 0.0011
4 | 17.034 4 0.0019
5 | 17.086 5 0.0043
---------------------------------------------------------------------------
H0: no serial correlation
Source | SS df MS Number of obs = 24
-------------+---------------------------------- F(2, 21) = 18.64
Model | 6.43630495 2 3.21815247 Prob > F = 0.0000
Residual | 3.62485995 21 .172612379 R-squared = 0.6397
-------------+---------------------------------- Adj R-squared = 0.6054
Total | 10.0611649 23 .437441952 Root MSE = .41547
------------------------------------------------------------------------------
errores | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .0744444 .0594593 1.25 0.224 -.0492079 .1980968
errores_1 | .8461998 .1385768 6.11 0.000 .5580135 1.134386
_cons | -.4963625 .3841865 -1.29 0.210 -1.295322 .302597
Source | SS df MS Number of obs = 24
-------------+---------------------------------- F(3, 20) = 13.42
Model | 6.72126486 3 2.24042162 Prob > F = 0.0001
Residual | 3.33990004 20 .166995002 R-squared = 0.6680
-------------+---------------------------------- Adj R-squared = 0.6182
Total | 10.0611649 23 .437441952 Root MSE = .40865
------------------------------------------------------------------------------
errores | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .114726 .0661154 1.74 0.098 -.0231883 .2526403
errores_1 | .6311015 .2137582 2.95 0.008 .1852097 1.076993
errores_2 | .3295937 .2523125 1.31 0.206 -.196721 .8559085
_cons | -.7682263 .4314039 -1.78 0.090 -1.668119 .1316664
Segundo modelo: el nivel de pobreza actual está originado tanto por el desempleo actual como la propia
pobreza del periodo anterior.
Estimación 1
Source | SS df MS Number of obs = 23
-------------+---------------------------------- F(2, 20) = 79.44
Model | 24.3008047 2 12.1504023 Prob > F = 0.0000
Residual | 3.0591956 20 .15295978 R-squared = 0.8882
-------------+---------------------------------- Adj R-squared = 0.8770
Total | 27.3600003 22 1.24363638 Root MSE = .3911
------------------------------------------------------------------------------
poverty | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
unemploy | .3711363 .0667902 5.56 0.000 .2318143 .5104582
|
poverty |
L1. | .5714112 .0917011 6.23 0.000 .380126 .7626964
|
_cons | 3.455135 1.066045 3.24 0.004 1.231404 5.678867
Análisis basados en la estimación 1
Durbin–Watson d-statistic( 3, 23) = .931197
Source | SS df MS Number of obs = 23
-------------+---------------------------------- F(3, 19) = 2.44
Model | .850265979 3 .283421993 Prob > F = 0.0961
Residual | 2.20892963 19 .116259454 R-squared = 0.2779
-------------+---------------------------------- Adj R-squared = 0.1639
Total | 3.05919561 22 .139054346 Root MSE = .34097
-------------------------------------------------------------------------------
residuos_din | Coefficient Std. err. t P>|t| [95% conf. interval]
------------+------------------------------------------------------------------
unemploy | -.0168542 .0485568 -0.35 0.732 -.1184847 .0847762
residuos_din_1| .4967362 .2296018 2.16 0.043 .0161741 .9772982
residuos_din_2| .0638445 .2296254 0.28 0.784 -.4167671 .544456
_cons | .1025082 .3114458 0.33 0.746 -.5493553 .7543716
Estimación 2
ARIMA regression
Sample: 1981 thru 2003 Number of obs = 23
Wald chi2(3) = 85.77
Log likelihood = -4.494719 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
| OPG
poverty | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
poverty |
unemploy | .4604097 .1185338 3.88 0.000 .2280877 .6927317
|
poverty |
L1. | .357229 .1795382 1.99 0.047 .0053407 .7091174
|
_cons | 5.767352 1.967567 2.93 0.003 1.910993 9.623712
-------------+----------------------------------------------------------------
ARMA |
ar |
L1. | .6935483 .2526162 2.75 0.006 .1984296 1.188667
-------------+----------------------------------------------------------------
/sigma | .2900286 .0523726 5.54 0.000 .1873803 .3926769
------------------------------------------------------------------------------
Note: The test of the variance against zero is one sided, and the two-sided
confidence interval is truncated at zero.
Estimación 3
ARIMA regression
Sample: 1981 thru 2003 Number of obs = 23
Wald chi2(3) = 103.21
Log likelihood = -6.710011 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
| OPG
poverty | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
poverty |
unemploy | .3928368 .1082616 3.63 0.000 .180648 .6050255
|
poverty |
L1. | .5161342 .1430562 3.61 0.000 .2357492 .7965192
|
_cons | 4.070126 1.551603 2.62 0.009 1.029041 7.111211
-------------+----------------------------------------------------------------
ARMA |
ma |
L1. | .4148316 .2378259 1.74 0.081 -.0512987 .8809618
-------------+----------------------------------------------------------------
/sigma | .3226148 .0621365 5.19 0.000 .2008296 .4444
------------------------------------------------------------------------------
Note: The test of the variance against zero is one sided, and the two-sided
confidence interval is truncated at zero.
Estimación 4
ARIMA regression
Sample: 1981 thru 2003 Number of obs = 23
Wald chi2(4) = 117.88
Log likelihood = -4.240011 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
| OPG
poverty | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
poverty |
unemploy | .4714504 .1120669 4.21 0.000 .2518033 .6910974
|
poverty |
L1. | .3623122 .1579135 2.29 0.022 .0528074 .671817
|
_cons | 5.609868 1.715168 3.27 0.001 2.248201 8.971535
-------------+----------------------------------------------------------------
ARMA |
ar |
L1. | .7816919 .2327097 3.36 0.001 .3255892 1.237794
|
ma |
L1. | -.1754888 .3850394 -0.46 0.649 -.9301523 .5791746
-------------+----------------------------------------------------------------
/sigma | .2866772 .0507762 5.65 0.000 .1871576 .3861968
------------------------------------------------------------------------------
Note: The test of the variance against zero is one sided, and the two-sided
confidence interval is truncated at zero.
http://www.cnnexpansion.com/economia/2011/02/01/problema-mundial-encuesta-injusticia-eu
