Universidad de Costa Rica, EC4300 Microeconometría
Las variables de este modelo siguen de cerca las definiciones de Kaminsky (1999) Currency and Banking Crises– The Early Warnings of Distress.
Tenemos dos variables dependientes, que modelamos por separado:
crisisbank: =1 si hay una crisis bancaria en los siguientes 24 mesescrisisbop: =1 si hay una crisis de balanza de pagos en los siguientes 24 mesesConsideramos los siguientes indicadores macroeconómicos como variables explicativas:
dolar: Dolarización (nivel)dolarg: Dolarización (Δ%)dom: Crédito doméstico / PIBrdegr: Depósitos domésticos (Δ%)expgr: Exportaciones (Δ%)impgr: Importaciones (Δ%)omega: Déficit comercial (nivel)resgr: Reservas (Δ%)rerres: Tipo de cambio real (desv. tendencia)inf: Inflación (nivel)mmgr: Multiplicador dinero (Δ%)mresg: M2 / Reservas (Δ%)rus: Interés EEUU (nivel)En este cuaderno usamos el comando labsumm, que es similar a summarize pero que muestra las etiquetas de las variables en lugar de sus nombres. Para instalarlo, ejecutar net install labsumm.
Los datos están en un libro de Excel, que importamos con import excel.
cd "C:\Users\randa\OneDrive\Documents\Teaching\UCR\EC4300 Microeconometría\Materiales\Binary"
import excel "crisis-binary.xlsx", sheet("Sheet1") firstrowC:\Users\randa\OneDrive\Documents\Teaching\UCR\EC4300 Microeconometría\Material
> es\Binary
(34 vars, 3,024 obs)Para simplificar el código, definimos algunas variables globales. Estas variables nos servirán para asegurarnos de que mantenemos consistencia en el listado de variables explicativas, y en el formato de los gráficos.
global regresores dolar dolarg dom expgr impgr inf mmgr mresg omega rdegr rerres resgr rus
global noiseplot1 mcolor(blue%18) msymbol(circle) mfcolor(blue) mlcolor(none)
global noiseplot2 yscale(range(0 1)) yline(0.5, lwidth(thick) lpattern(dash) lcolor(cranberry)) ylabel(#3) xtitle(, size(large)) xscale(range(-0.1 0.6)) xlabel(0 "No" 0.5 "Sí", labels labsize(medlarge)) xsize(4) ysize(4)
global kernelplot recast(area) fcolor(%25) ytitle(Densidad) ylabel(, nogrid) xscale(range(0 1)) xlabel(0(0.25)1) legend(order(1 "crisis" 2 "tranquilo") position(2) ring(0)) scheme(meta) xsize(4) ysize(4)Además, definimos los datos como de panel (indexados por meses y por países).
encode country, generate(pais)
gen meses = ym(year(dateid), month(dateid))
tsset pais meses, monthly
Panel variable: pais (strongly balanced)
Time variable: meses, 1970m1 to 2005m12
Delta: 1 monthFinamente, etiquetamos los datos para obtener resultados con las etiquetas apropiadas (serán más fáciles de recordar, al no obligarnos a recordar los nombres de las variables).
label variable dolar "Dolarización (nivel)"
label variable dolarg "Dolarización (Δ%)"
label variable dom "Crédito doméstico / PIB"
label variable rdegr "Depósitos domésticos (Δ%)"
label variable expgr "Exportaciones (Δ%)"
label variable impgr "Importaciones (Δ%)"
label variable omega "Déficit comercial (nivel)"
label variable resgr "Reservas (Δ%)"
label variable rerres "Tipo de cambio real (desv. tendencia)"
label variable inf "Inflación (nivel)"
label variable mmgr "Multiplicador dinero (Δ%)"
label variable mresg "M2 / Reservas (Δ%)"
label variable rus "Interés EEUU (nivel)"
label define crisis 0 "Tranquil times" 1 "Crisis times"
label values crisisbank crisis
label values crisisbop crisis
format %8.4f $regresores crisisbank crisisbopConservamos únicamente las variables de interés.
keep crisisbank crisisbop $regresores paisObtenemos un cuadro de estadísticas básicas para las variables.
set linesize 90
labsumm, format Variable | Obs Mean Std. Dev. Min Max
--------------------------------------+----------------------------------------------
crisisbank | 3024 0.1157 0.3200 0.0000 1.0000
crisisbop | 3024 0.2189 0.4136 0.0000 1.0000
Dolarización (nivel) | 3023 0.0852 0.0900 0.0000 0.7979
Dolarización (Δ%) | 2939 0.0181 0.0815 -0.6645 0.6328
Crédito doméstico / PIB | 2785 0.0263 0.2516 -1.0000 2.3342
Exportaciones (Δ%) | 2940 0.1391 0.3336 -0.8529 3.4702
Importaciones (Δ%) | 2940 0.1472 0.3390 -0.7616 2.3421
Inflación (nivel) | 2898 0.0413 0.1144 -0.0968 3.9696
Multiplicador dinero (Δ%) | 2929 0.0411 0.2840 -0.8667 3.3476
M2 / Reservas (Δ%) | 2940 0.0657 0.5398 -0.9478 7.9688
Déficit comercial (nivel) | 2947 -0.1138 0.4279 -2.0877 0.6573
Depósitos domésticos (Δ%) | 2821 0.0812 0.3036 -0.8503 3.3052
Tipo de cambio real (desv. tendencia) | 2905 0.0000 31.3367 -53.6713 343.6609
Reservas (Δ%) | 2940 0.2654 0.8081 -0.9285 10.8824
Interés EEUU (nivel) | 3017 6.2596 3.1325 0.8608 17.2877
country | 3024 4 2.00033 1 7Veamos que el cuadro de arriba es más conveniente que el que obtendríamos con summarize
summarize, format
Variable | Obs Mean Std. dev. Min Max
-------------+---------------------------------------------------------
crisisbank | 3,024 0.1157 0.3200 0.0000 1.0000
crisisbop | 3,024 0.2189 0.4136 0.0000 1.0000
dolar | 3,023 0.0852 0.0900 0.0000 0.7979
dolarg | 2,939 0.0181 0.0815 -0.6645 0.6328
dom | 2,785 0.0263 0.2516 -1.0000 2.3342
-------------+---------------------------------------------------------
expgr | 2,940 0.1391 0.3336 -0.8529 3.4702
impgr | 2,940 0.1472 0.3390 -0.7616 2.3421
inf | 2,898 0.0413 0.1144 -0.0968 3.9696
mmgr | 2,929 0.0411 0.2840 -0.8667 3.3476
mresg | 2,940 0.0657 0.5398 -0.9478 7.9688
-------------+---------------------------------------------------------
omega | 2,947 -0.1138 0.4279 -2.0877 0.6573
rdegr | 2,821 0.0812 0.3036 -0.8503 3.3052
rerres | 2,905 0.0000 31.3367 -53.6713 343.6609
resgr | 2,940 0.2654 0.8081 -0.9285 10.8824
rus | 3,017 6.2596 3.1325 0.8608 17.2877
-------------+---------------------------------------------------------
pais | 3,024 4 2.000331 1 7rename crisisbank crisis
logit crisis $regresores ibn.pais, vce(robust) noci noconstant
estimates store bank_logit
Iteration 0: log pseudolikelihood = -1922.0971
Iteration 1: log pseudolikelihood = -863.80044
Iteration 2: log pseudolikelihood = -825.02213
Iteration 3: log pseudolikelihood = -823.96348
Iteration 4: log pseudolikelihood = -823.96177
Iteration 5: log pseudolikelihood = -823.96177
Logistic regression Number of obs = 2,773
Wald chi2(20) = 1125.81
Log pseudolikelihood = -823.96177 Prob > chi2 = 0.0000
-----------------------------------------------------
| Robust
crisis | Coefficient std. err. z P>|z|
-------------+---------------------------------------
dolar | 5.576028 .700061 7.97 0.000
dolarg | -.4979393 .9440406 -0.53 0.598
dom | 2.021968 .3337248 6.06 0.000
expgr | -.1659892 .2142752 -0.77 0.439
impgr | -1.511297 .2634938 -5.74 0.000
inf | -1.809969 .8032339 -2.25 0.024
mmgr | .1332535 .1727431 0.77 0.440
mresg | .5647882 .1395775 4.05 0.000
omega | -1.151355 .1910975 -6.02 0.000
rdegr | -.9789297 .2571707 -3.81 0.000
rerres | -.0031407 .0025129 -1.25 0.211
resgr | -.0762417 .1229349 -0.62 0.535
rus | .0985633 .018055 5.46 0.000
|
pais |
ARG | -2.287594 .213332 -10.72 0.000
BRA | -3.09899 .2442557 -12.69 0.000
CHI | -4.901125 .2828562 -17.33 0.000
COL | -2.571798 .2166813 -11.87 0.000
MEX | -3.371177 .2877089 -11.72 0.000
PER | -3.783422 .2379317 -15.90 0.000
VEN | -4.051092 .3361182 -12.05 0.000
-----------------------------------------------------margins, dydx(*) atmeans noci post
estimates store margin_bank_atmeans
estimates restore bank_logit
margins, dydx(*) noci post
estimates store margin_bank
estimates restore bank_logit
Conditional marginal effects Number of obs = 2,773
Model VCE: Robust
Expression: Pr(crisis), predict()
dy/dx wrt: dolar dolarg dom expgr impgr inf mmgr mresg omega rdegr rerres resgr rus
2.pais 3.pais 4.pais 5.pais 6.pais 7.pais
At: dolar = .0911226 (mean)
dolarg = .0190081 (mean)
dom = .0266963 (mean)
expgr = .1372507 (mean)
impgr = .140755 (mean)
inf = .0425624 (mean)
mmgr = .0428351 (mean)
mresg = .0650741 (mean)
omega = -.1289817 (mean)
rdegr = .0813646 (mean)
rerres = .683406 (mean)
resgr = .2681343 (mean)
rus = 6.248104 (mean)
1.pais = .1471331 (mean)
2.pais = .1081861 (mean)
3.pais = .1489362 (mean)
4.pais = .1489362 (mean)
5.pais = .1489362 (mean)
6.pais = .1489362 (mean)
7.pais = .1489362 (mean)
-----------------------------------------------------
| Delta-method
| dy/dx std. err. z P>|z|
-------------+---------------------------------------
dolar | .3878804 .051841 7.48 0.000
dolarg | -.0346377 .0658742 -0.53 0.599
dom | .1406524 .0227766 6.18 0.000
expgr | -.0115466 .0149594 -0.77 0.440
impgr | -.105129 .0176972 -5.94 0.000
inf | -.1259053 .0557227 -2.26 0.024
mmgr | .0092694 .0120071 0.77 0.440
mresg | .0392879 .0098893 3.97 0.000
omega | -.0800907 .0125584 -6.38 0.000
rdegr | -.0680964 .0178888 -3.81 0.000
rerres | -.0002185 .0001728 -1.26 0.206
resgr | -.0053035 .0085807 -0.62 0.537
rus | .0068563 .0012517 5.48 0.000
|
pais |
BRA | -.1030989 .0273864 -3.76 0.000
CHI | -.1882527 .0238854 -7.88 0.000
COL | -.0428045 .0367298 -1.17 0.244
MEX | -.1258528 .028143 -4.47 0.000
PER | -.1517599 .0263421 -5.76 0.000
VEN | -.1641821 .0243284 -6.75 0.000
-----------------------------------------------------
Note: dy/dx for factor levels is the discrete change
from the base level.
(results bank_logit are active now)
Average marginal effects Number of obs = 2,773
Model VCE: Robust
Expression: Pr(crisis), predict()
dy/dx wrt: dolar dolarg dom expgr impgr inf mmgr mresg omega rdegr rerres resgr rus
2.pais 3.pais 4.pais 5.pais 6.pais 7.pais
-----------------------------------------------------
| Delta-method
| dy/dx std. err. z P>|z|
-------------+---------------------------------------
dolar | .4868221 .0621842 7.83 0.000
dolarg | -.0434732 .082635 -0.53 0.599
dom | .1765304 .0286189 6.17 0.000
expgr | -.0144919 .0187637 -0.77 0.440
impgr | -.1319457 .0224053 -5.89 0.000
inf | -.1580216 .0701566 -2.25 0.024
mmgr | .0116339 .0151103 0.77 0.441
mresg | .0493095 .0121184 4.07 0.000
omega | -.1005205 .0157516 -6.38 0.000
rdegr | -.0854667 .0224627 -3.80 0.000
rerres | -.0002742 .0002185 -1.25 0.210
resgr | -.0066564 .010754 -0.62 0.536
rus | .0086052 .0015531 5.54 0.000
|
pais |
BRA | -.102001 .0269919 -3.78 0.000
CHI | -.211556 .0222333 -9.52 0.000
COL | -.0402919 .0348531 -1.16 0.248
MEX | -.127607 .0279135 -4.57 0.000
PER | -.1592223 .0261813 -6.08 0.000
VEN | -.1756954 .0247131 -7.11 0.000
-----------------------------------------------------
Note: dy/dx for factor levels is the discrete change
from the base level.
(results bank_logit are active now)predict phatbank, pr(251 missing values generated)kdensity phatbank if crisis==1, $kernelplot addplot((kdensity phatbank if crisis==0)) title(Probabilidad pronosticada de crisis bancaria) name(bank_kdensity, replace)
estat classification
Logistic model for crisis
-------- True --------
Classified | D ~D | Total
-----------+--------------------------+-----------
+ | 85 32 | 117
- | 265 2391 | 2656
-----------+--------------------------+-----------
Total | 350 2423 | 2773
Classified + if predicted Pr(D) >= .5
True D defined as crisis != 0
--------------------------------------------------
Sensitivity Pr( +| D) 24.29%
Specificity Pr( -|~D) 98.68%
Positive predictive value Pr( D| +) 72.65%
Negative predictive value Pr(~D| -) 90.02%
--------------------------------------------------
False + rate for true ~D Pr( +|~D) 1.32%
False - rate for true D Pr( -| D) 75.71%
False + rate for classified + Pr(~D| +) 27.35%
False - rate for classified - Pr( D| -) 9.98%
--------------------------------------------------
Correctly classified 89.29%
--------------------------------------------------generate banknoise = 0.5*crisis + runiform(-0.1,0.1)
twoway (scatter phatbank banknoise, $noiseplot1), $noiseplot2 xtitle(¿Crisis bancaria?) name(bank_clasifica, replace)
lsens, legend(position(3) ring(0) col(1)) name(bank_lsens, replace) xsize(4) ysize(4)
lroc, name(bank_lroc, replace) xsize(4) ysize(4)
Logistic model for crisis
Number of observations = 2773
Area under ROC curve = 0.8106
rename crisis crisisbankrename crisisbop crisis
logit crisis $regresores ibn.pais, vce(robust) noci noconstant
estimates store bop_logit
Iteration 0: log pseudolikelihood = -1922.0971
Iteration 1: log pseudolikelihood = -1215.3122
Iteration 2: log pseudolikelihood = -1196.974
Iteration 3: log pseudolikelihood = -1196.6378
Iteration 4: log pseudolikelihood = -1196.6375
Iteration 5: log pseudolikelihood = -1196.6375
Logistic regression Number of obs = 2,773
Wald chi2(20) = 1058.62
Log pseudolikelihood = -1196.6375 Prob > chi2 = 0.0000
-----------------------------------------------------
| Robust
crisis | Coefficient std. err. z P>|z|
-------------+---------------------------------------
dolar | -9.589695 .874495 -10.97 0.000
dolarg | 1.72331 .7790785 2.21 0.027
dom | 1.698562 .3100398 5.48 0.000
expgr | -.33557 .1769484 -1.90 0.058
impgr | -.5027071 .1888624 -2.66 0.008
inf | 2.02608 1.890753 1.07 0.284
mmgr | .3518654 .1780368 1.98 0.048
mresg | .9852773 .151289 6.51 0.000
omega | .2098591 .1690642 1.24 0.214
rdegr | -1.20878 .2454483 -4.92 0.000
rerres | -.0002649 .0025736 -0.10 0.918
resgr | .2095455 .0926526 2.26 0.024
rus | .1296012 .0160628 8.07 0.000
|
pais |
ARG | -1.750955 .225005 -7.78 0.000
BRA | -.3920284 .2136607 -1.83 0.067
CHI | -2.076249 .2228283 -9.32 0.000
COL | -1.944423 .2121107 -9.17 0.000
MEX | -1.171794 .1973245 -5.94 0.000
PER | -2.345546 .2431813 -9.65 0.000
VEN | -1.118656 .1590713 -7.03 0.000
-----------------------------------------------------margins, dydx(*) atmeans noci post
estimates store margin_bop_atmeans
estimates restore bop_logit
margins, dydx(*) noci post
estimates store margin_bop
estimates restore bop_logit
Conditional marginal effects Number of obs = 2,773
Model VCE: Robust
Expression: Pr(crisis), predict()
dy/dx wrt: dolar dolarg dom expgr impgr inf mmgr mresg omega rdegr rerres resgr rus
2.pais 3.pais 4.pais 5.pais 6.pais 7.pais
At: dolar = .0911226 (mean)
dolarg = .0190081 (mean)
dom = .0266963 (mean)
expgr = .1372507 (mean)
impgr = .140755 (mean)
inf = .0425624 (mean)
mmgr = .0428351 (mean)
mresg = .0650741 (mean)
omega = -.1289817 (mean)
rdegr = .0813646 (mean)
rerres = .683406 (mean)
resgr = .2681343 (mean)
rus = 6.248104 (mean)
1.pais = .1471331 (mean)
2.pais = .1081861 (mean)
3.pais = .1489362 (mean)
4.pais = .1489362 (mean)
5.pais = .1489362 (mean)
6.pais = .1489362 (mean)
7.pais = .1489362 (mean)
-----------------------------------------------------
| Delta-method
| dy/dx std. err. z P>|z|
-------------+---------------------------------------
dolar | -1.343759 .1231165 -10.91 0.000
dolarg | .2414794 .109476 2.21 0.027
dom | .2380116 .0422329 5.64 0.000
expgr | -.0470219 .0246548 -1.91 0.056
impgr | -.070442 .0262885 -2.68 0.007
inf | .283905 .2636708 1.08 0.282
mmgr | .0493052 .0249174 1.98 0.048
mresg | .1380623 .0220845 6.25 0.000
omega | .0294066 .0237446 1.24 0.216
rdegr | -.1693807 .0342112 -4.95 0.000
rerres | -.0000371 .0003603 -0.10 0.918
resgr | .0293626 .013044 2.25 0.024
rus | .0181604 .0023233 7.82 0.000
|
pais |
BRA | .2545243 .0375927 6.77 0.000
CHI | -.036317 .0264694 -1.37 0.170
COL | -.022652 .0328935 -0.69 0.491
MEX | .0882344 .034709 2.54 0.011
PER | -.0601932 .0254236 -2.37 0.018
VEN | .0979301 .0351546 2.79 0.005
-----------------------------------------------------
Note: dy/dx for factor levels is the discrete change
from the base level.
(results bop_logit are active now)
Average marginal effects Number of obs = 2,773
Model VCE: Robust
Expression: Pr(crisis), predict()
dy/dx wrt: dolar dolarg dom expgr impgr inf mmgr mresg omega rdegr rerres resgr rus
2.pais 3.pais 4.pais 5.pais 6.pais 7.pais
-----------------------------------------------------
| Delta-method
| dy/dx std. err. z P>|z|
-------------+---------------------------------------
dolar | -1.332323 .1220498 -10.92 0.000
dolarg | .2394243 .1080927 2.21 0.027
dom | .235986 .0420344 5.61 0.000
expgr | -.0466217 .0244767 -1.90 0.057
impgr | -.0698425 .0259955 -2.69 0.007
inf | .2814889 .2618264 1.08 0.282
mmgr | .0488856 .0247312 1.98 0.048
mresg | .1368873 .0203387 6.73 0.000
omega | .0291563 .0235405 1.24 0.216
rdegr | -.1679392 .0336569 -4.99 0.000
rerres | -.0000368 .0003574 -0.10 0.918
resgr | .0291127 .0128264 2.27 0.023
rus | .0180059 .0022163 8.12 0.000
|
pais |
BRA | .2237431 .0320667 6.98 0.000
CHI | -.0385945 .0277888 -1.39 0.165
COL | -.0237074 .0344942 -0.69 0.492
MEX | .0841388 .0328992 2.56 0.011
PER | -.0659074 .0272463 -2.42 0.016
VEN | .0928011 .0328409 2.83 0.005
-----------------------------------------------------
Note: dy/dx for factor levels is the discrete change
from the base level.
(results bop_logit are active now)predict phatbop, pr(251 missing values generated)kdensity phatbop if crisis==1, $kernelplot addplot((kdensity phatbop if crisis==0)) title(Probabilidad pronosticada de crisis de balanza de pagos) name(bop_kdensity, replace)
estat classification
Logistic model for crisis
-------- True --------
Classified | D ~D | Total
-----------+--------------------------+-----------
+ | 179 142 | 321
- | 465 1987 | 2452
-----------+--------------------------+-----------
Total | 644 2129 | 2773
Classified + if predicted Pr(D) >= .5
True D defined as crisis != 0
--------------------------------------------------
Sensitivity Pr( +| D) 27.80%
Specificity Pr( -|~D) 93.33%
Positive predictive value Pr( D| +) 55.76%
Negative predictive value Pr(~D| -) 81.04%
--------------------------------------------------
False + rate for true ~D Pr( +|~D) 6.67%
False - rate for true D Pr( -| D) 72.20%
False + rate for classified + Pr(~D| +) 44.24%
False - rate for classified - Pr( D| -) 18.96%
--------------------------------------------------
Correctly classified 78.11%
--------------------------------------------------generate bopnoise = 0.5*crisis + runiform(-0.1,0.1)
twoway (scatter phatbank bopnoise, $noiseplot1), $noiseplot2 xtitle(¿Crisis de balanza de pagos?) name(bop_clasifica, replace)
lsens, legend(position(3) ring(0) col(1)) name(bop_lsens, replace) xsize(4) ysize(4)
lroc, name(bop_lroc, replace) xsize(4) ysize(4)
Logistic model for crisis
Number of observations = 2773
Area under ROC curve = 0.8095
rename crisis crisisbopetable, column(estimates) estimates(margin_bank_atmeans margin_bank margin_bop_atmeans margin_bop) cstat(_r_b) keep($regresores) title(Efectos marginales) export(crisis-margins.tex, tableonly replace) showstars showstarsnote stars(.10 "*" .05 "**" .01 "***", attach(_r_b))
Efectos marginales
---------------------------------------------------------------------------------------------------
margin_bank_atmeans margin_bank margin_bop_atmeans margin_bop
---------------------------------------------------------------------------------------------------
Dolarización (nivel) 0.388 *** 0.487 *** -1.344 *** -1.332 ***
Dolarización (Δ%) -0.035 -0.043 0.241 ** 0.239 **
Crédito doméstico / PIB 0.141 *** 0.177 *** 0.238 *** 0.236 ***
Exportaciones (Δ%) -0.012 -0.014 -0.047 * -0.047 *
Importaciones (Δ%) -0.105 *** -0.132 *** -0.070 *** -0.070 ***
Inflación (nivel) -0.126 ** -0.158 ** 0.284 0.281
Multiplicador dinero (Δ%) 0.009 0.012 0.049 ** 0.049 **
M2 / Reservas (Δ%) 0.039 *** 0.049 *** 0.138 *** 0.137 ***
Déficit comercial (nivel) -0.080 *** -0.101 *** 0.029 0.029
Depósitos domésticos (Δ%) -0.068 *** -0.085 *** -0.169 *** -0.168 ***
Tipo de cambio real (desv. tendencia) -0.000 -0.000 -0.000 -0.000
Reservas (Δ%) -0.005 -0.007 0.029 ** 0.029 **
Interés EEUU (nivel) 0.007 *** 0.009 *** 0.018 *** 0.018 ***
Number of observations 2773 2773 2773 2773
---------------------------------------------------------------------------------------------------
*** p<.01, ** p<.05, * p<.1
(collection ETable exported to file crisis-margins.tex)format %6.4f phatbank phatbop
summarize crisisbank phatbank crisisbop phatbop, format
label variable phatbank "Prob(crisis bancaria)"
label variable phatbop "Prob(crisis balanza de pagos)"
Variable | Obs Mean Std. dev. Min Max
-------------+---------------------------------------------------------
crisisbank | 3,024 0.1157 0.3200 0.0000 1.0000
phatbank | 2,773 0.1262 0.1516 0.0000 0.9266
crisisbop | 3,024 0.2189 0.4136 0.0000 1.0000
phatbop | 2,773 0.2322 0.1985 0.0008 0.9993graph combine bank_kdensity bop_kdensity, xsize(8) ysize(4)
graph export crisis_kdensity.pdf, replacefile crisis_kdensity.pdf saved as PDF format
graph combine bank_lsens bop_lsens, xsize(8) ysize(4)
graph export crisis_lsens.pdf, replacefile crisis_lsens.pdf saved as PDF format
graph combine bank_lroc bop_lroc, xsize(8) ysize(4)
graph export crisis_lroc.pdf, replacefile crisis_lroc.pdf saved as PDF format
graph combine bank_clasifica bop_clasifica, xsize(8) ysize(4)
graph export crisis_clasifica.pdf, replacefile crisis_clasifica.pdf saved as PDF format