Hello Statalister,
I have been recently dealing with my model for the dissertation. I am an absolute beginner for Stata and newly experimenting with Xtabond2.
My issue is to estimate the "Impact of institutions and governance on FDI inflows to European countries". In literature, such models are formed through dynamic GMM estimators.
I have T=10 and N=37. Difference rather than System GMM is suggested in some papers when the number of countries are relatively small which makes difficult to handle so many instruments in System GMM. I am using Worldwide Governance Indicators (WGI) by Kaufmann et. al..
I have formed the equation in Xtabond2 as:
. xtabond2 logfdi l.logfdi logregula logsize l1.growth logmobile open yr*, gmm(logfdi logregula, laglimit (2 2)) iv(logsize l1.growth logmobile open yr* leg_*) noleveleq nodiffsargan robust small
logfdi : FDI inflows as % GDP (transformed to log by adding the minimum number)
logregula : Regulatory Quality (transformed to log by adding 3 as suggested in literature)
logsize : GDP per capita
l1.growth : lagged GDP growth as suggested in literature
logmobile : Mobile phone subscription per 100 people to proxy Infrastructure
open : Trade Openness
leg_* : Legal origins dummies as English, French, Socialist etc. as Instruments
I want to test 6 individual Governance indicators in seperate models due to multicollinearity. I have 3 fundamental issues. Firstly, I am always getting Negative and Insignificant coeeficients for Governance Indicators. Secondly, my Hansen tests are unsatisfactory and dramatically changes when I experiment with different indicators. Thirdly, I want to get consistent results to show.
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
yr10 dropped due to collinearity
Warning: Two-step estimated covariance matrix of moments is singular.
Using a generalized inverse to calculate robust weighting matrix for Hansen test.
Dynamic panel-data estimation, one-step difference GMM
------------------------------------------------------------------------------
Group variable: c_id Number of obs = 296
Time variable : year Number of groups = 37
Number of instruments = 28 Obs per group: min = 8
F(14, 37) = 63.55 avg = 8.00
Prob > F = 0.000 max = 8
------------------------------------------------------------------------------
| Robust
logfdi | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
logfdi |
L1. | -.4734568 .098903 -4.79 0.000 -.6738533 -.2730603
|
logregula | -8.831733 7.478435 -1.18 0.245 -23.98448 6.321017
logsize | .7407178 .9205677 0.80 0.426 -1.12453 2.605965
|
growth |
L1. | .0077484 .0101572 0.76 0.450 -.012832 .0283287
|
logmobile | -.0877016 .4399102 -0.20 0.843 -.9790443 .8036411
open | .0021909 .0057888 0.38 0.707 -.0095383 .0139201
yr2 | .2736571 .1975537 1.39 0.174 -.1266248 .673939
yr3 | .2724528 .1916885 1.42 0.164 -.1159451 .6608506
yr4 | .035629 .152469 0.23 0.817 -.2733026 .3445606
yr5 | .1370581 .1358924 1.01 0.320 -.1382861 .4124022
yr6 | .1060513 .1272426 0.83 0.410 -.1517668 .3638694
yr7 | .0328566 .0988526 0.33 0.741 -.1674379 .2331511
yr8 | .078686 .0890788 0.88 0.383 -.1018047 .2591768
yr9 | .0719013 .0572243 1.26 0.217 -.0440462 .1878487
------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(logsize L.growth logmobile open yr2 yr3 yr4 yr5 yr6 yr7 yr8 yr9 yr10
leg_eng leg_soc leg_fre leg_ger leg_sca)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L2.(logfdi logregula)
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -1.29 Pr > z = 0.198
Arellano-Bond test for AR(2) in first differences: z = 0.23 Pr > z = 0.819
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(14) = 17.33 Prob > chi2 = 0.239
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(14) = 14.19 Prob > chi2 = 0.436
(Robust, but weakened by many instruments.)
I have been dealing with 4 days and no upgrade yet. Still, I am getting irrelavant results even if I try System GMM and different combinations of instruments. I am open to any advice. THanks in advance.
Kind regards,
Related Posts with GMM, xtabond2: negative and insignificant coefficients
Unexpected behavior (suspected bug) when using Mata views with temporary variables created by fvrevarI encountered a very strange behavior when I use Mata views on temporary variables created by fvreva…
Spatial Autogressive Models for Binary Outcomes?Hi Folks, I was reading through about Stata's spatial autogressive models (https://www.stata.com/fe…
Stata Recoding Variable MacroHi all, I have a numeric variable, range is about from 100-400,000. I would like to recode this var…
Is there an import delimited option to remove "." from variable names?Hi, I just imported a csv using the import delimited command. It imported with no errors or warning…
Pick columns to output using esttab and estpost correlationHello, I was wondering if there was an easy way to pick the columns to output with esttab when buil…
Subscribe to:
Post Comments (Atom)
0 Response to GMM, xtabond2: negative and insignificant coefficients
Post a Comment