Dear all,
It is my first post on Statalist, I hope I am doing everything fine, I read all the FAQs but I am still uncertain.
I am trying to assess the cost of default on the GDP growth of a defaulter country. I have panel data, more in detail 201 countries and 41 years (1970-2011).
I choose, following "Furceri, D., & Zdzienicka, A. (2012). How costly are debt crises?. Journal of International Money and Finance, 31(4), 726-742." to adopt the GMM estimator, in particular, the code xtabond2. I read on "Roodman, D. (2009). How to do xtabond2: An introduction to difference and system GMM in Stata. The stata journal, 9(1), 86-136." that this estimator should be applied only in a situation in which T is small, and probably this is my first issue in my analysis. But I cannot understand when T is small, according to you, 41 years is too large?
Consequently my model is a dynamic model in which the dependent variable is GDP growth rate and the independent variables, in the baseline regression, are: the first lag of GDP growth rate, the second lag of GDP growth rate, population growth, credit growth, first lag of the log of real GDP level, real exchange rate, measure of openness and a dummy of sovereign default. The latter is my variable of interest, take the value of 1 for every year that the country is in default.
This is the code to estimate my model with xtabond2. Polity_index is a measure of polity score,
(If someone could be interested in this analysis, I took the data from the paper of: Kuvshinov, D., & Zimmermann, K. (2019). Sovereigns going bust: estimating the cost of default. European Economic Review.
The paper and the data are available at: https://doi.org/10.1016/j.euroecorev.2019.04.009
Here it is an example of this dataset
My question is, if according to you my T is too large, and it could be the reason for which I am obtaining a lot of non-significant coefficients for my independent variables.
Thank you so much in advance.
Giampaolo.
It is my first post on Statalist, I hope I am doing everything fine, I read all the FAQs but I am still uncertain.
I am trying to assess the cost of default on the GDP growth of a defaulter country. I have panel data, more in detail 201 countries and 41 years (1970-2011).
I choose, following "Furceri, D., & Zdzienicka, A. (2012). How costly are debt crises?. Journal of International Money and Finance, 31(4), 726-742." to adopt the GMM estimator, in particular, the code xtabond2. I read on "Roodman, D. (2009). How to do xtabond2: An introduction to difference and system GMM in Stata. The stata journal, 9(1), 86-136." that this estimator should be applied only in a situation in which T is small, and probably this is my first issue in my analysis. But I cannot understand when T is small, according to you, 41 years is too large?
Consequently my model is a dynamic model in which the dependent variable is GDP growth rate and the independent variables, in the baseline regression, are: the first lag of GDP growth rate, the second lag of GDP growth rate, population growth, credit growth, first lag of the log of real GDP level, real exchange rate, measure of openness and a dummy of sovereign default. The latter is my variable of interest, take the value of 1 for every year that the country is in default.
Code:
xtabond2 rgdp_growth l.rgdp_growth l2.rgdp_growth popgrowth loans_gdp_growth l.ln_rgdp_level r_exrate_growth openness in_dummy_s_and_p, gmm (l.rgdp_growth l.in_dummy_s_and_p, laglimits (1 40) collapse) iv(polity_index inflation coup,equation(level)) nodiffsargan twostep robust orthogonal small
Code:
xtabond2 rgdp_growth l.rgdp_growth l2.rgdp_growth popgrowth loans_gdp_growth l.ln_rgdp_level r_exrate_growth openness in_dummy_s_and_p, gmm (l.rgdp_growth l.in_dummy_s_and_p, laglimit
> s (1 40) collapse) iv(polity_index inflation coup,equation(level)) nodiffsargan twostep robust orthogonal small
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
Warning: Two-step estimated covariance matrix of moments is singular.
Using a generalized inverse to calculate optimal weighting matrix for two-step estimation.
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: i Number of obs = 4747
Time variable : year Number of groups = 144
Number of instruments = 86 Obs per group: min = 0
F(8, 143) = 12.41 avg = 32.97
Prob > F = 0.000 max = 41
----------------------------------------------------------------------------------
| Corrected
rgdp_growth | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
rgdp_growth |
L1. | .1734639 .0400013 4.34 0.000 .0943937 .2525341
L2. | .0163024 .0299815 0.54 0.587 -.0429618 .0755666
|
popgrowth | .4590989 5.71886 0.08 0.936 -10.84533 11.76352
loans_gdp_growth | .4131774 .2608927 1.58 0.115 -.1025271 .9288819
|
ln_rgdp_level |
L1. | -.7471318 1.795971 -0.42 0.678 -4.297215 2.802951
|
r_exrate_growth | -3.879862 2.763294 -1.40 0.162 -9.342044 1.582321
openness | .0332902 .0520659 0.64 0.524 -.0696281 .1362084
in_dummy_s_and_p | -1.342024 .7757352 -1.73 0.086 -2.875414 .1913657
_cons | 3.394403 10.78057 0.31 0.753 -17.91546 24.70427
----------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(1/40).(L.rgdp_growth L.in_dummy_s_and_p) collapsed
Instruments for levels equation
Standard
polity_index inflation coup
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.(L.rgdp_growth L.in_dummy_s_and_p) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -5.20 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = -1.72 Pr > z = 0.085
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(77) = 83.74 Prob > chi2 = 0.280
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(77) = 85.84 Prob > chi2 = 0.230
(Robust, but weakened by many instruments.)
(If someone could be interested in this analysis, I took the data from the paper of: Kuvshinov, D., & Zimmermann, K. (2019). Sovereigns going bust: estimating the cost of default. European Economic Review.
The paper and the data are available at: https://doi.org/10.1016/j.euroecorev.2019.04.009
Here it is an example of this dataset
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input str46 country double year float(rgdp_growth r_exrate_growth in_dummy_s_and_p popgrowth ln_rgdp_level) "Aruba" 1970 . .04352283 0 . 8.138505 "Aruba" 1971 8.40391 .04343969 0 . 8.219199 "Aruba" 1972 8.34116 .05448103 0 . 8.299314 "Aruba" 1973 8.384026 .08301789 0 . 8.379825 "Aruba" 1974 8.572991 .09668136 0 . 8.462077 "Aruba" 1975 8.854675 .07771558 0 . 8.546921 "Aruba" 1976 9.205509 .09121424 0 . 8.634982 "Aruba" 1977 9.491458 .10704023 0 . 8.725658 "Aruba" 1978 9.556724 .1337465 0 . 8.816931 "Aruba" 1979 9.321277 .1599741 0 . 8.906052 "Aruba" 1980 8.872477 .18441904 0 . 8.991058 "Aruba" 1981 8.240195 .15544665 0 . 9.070241 "Aruba" 1982 7.706792 .13252604 0 . 9.144484 "Aruba" 1983 7.578631 .13882983 0 . 9.217535 "Aruba" 1984 8.003371 .13464308 0 . 9.294527 "Aruba" 1985 8.747523 .12561679 0 . 9.3783865 "Aruba" 1986 9.758915 -.014589548 0 . 9.471502 "Aruba" 1987 17.608486 .007916689 0 . 9.633694 "Aruba" 1988 20.12718 -.003917694 0 . 9.817074 "Aruba" 1989 12.208845 -.04144764 0 . 9.932265 end format %ty year
Thank you so much in advance.
Giampaolo.
Comment