------------------------------------------------------------------------------- name: log: /Users/carlos/GitHub/starter-academic-v501/content/post/stata_dyna > mic_panel/analysis.log log type: text opened on: 29 Apr 2026, 10:10:24 . . *---------------------------------------------------- . * Section 2: Install dependencies (idempotent) . *---------------------------------------------------- . * xtabond2 - Arellano-Bond / Blundell-Bond GMM (Roodman) . * estout - publication-quality regression tables . * outreg2 - alternative table exporter . * coefplot - plot regression coefficients . capture which xtabond2 . if _rc capture ssc install xtabond2, replace . capture which estout . if _rc capture ssc install estout, replace . capture which outreg2 . if _rc capture ssc install outreg2, replace . capture which coefplot . if _rc capture ssc install coefplot, replace . . *---------------------------------------------------- . * Section 3: Import data . *---------------------------------------------------- . * The CatoJ.dta panel is hosted on GitHub by quarcs-lab. . use "https://github.com/quarcs-lab/data-open/raw/master/panel/CatoJ.dta", cle > ar . . describe Contains data from https://github.com/quarcs-lab/data-open/raw/master/panel/Cat > oJ.dta Observations: 1,663 Variables: 18 26 Oct 2019 15:46 ------------------------------------------------------------------------------- Variable Storage Display Value name type format label Variable label ------------------------------------------------------------------------------- Country str35 %35s Country cty long %35.0g cty Country Year int %ty Year Sample byte %10.0g Sample lnGDPpercapita double %10.0g ln(GDP per capita) ChangeinlnGDP~a double %10.0g Change in ln(GDP per capita) lnGDPpercapit~g double %10.0g ln(GDP per capita) Lag EconFreeLag double %10.0g Econ Free Lag PolitFreeLag double %10.0g Polit Free Lag DemocIndxLag double %10.0g Democ Indx Lag War double %10.0g War WarLag double %10.0g War Lag Coup double %10.0g Coup CoupLag double %10.0g Coup Lag Americas1 byte %10.0g Americas [1] Africa1 byte %10.0g Africa [1] MENA1 byte %10.0g MENA [1] Asia1 byte %10.0g Asia [1] ------------------------------------------------------------------------------- Sorted by: cty Year . sum Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- Country | 0 cty | 1,663 80.82441 45.80199 1 160 Year | 1,663 1988.596 18.06104 1955 2015 Sample | 1,663 1.188815 .3914796 1 2 lnGDPperca~a | 1,663 8.768894 1.204839 5.63479 12.70269 -------------+--------------------------------------------------------- Changeinln~a | 1,663 .0955832 .1752815 -1.189208 1.282702 lnGDPperca~g | 1,663 8.673311 1.19621 5.63479 12.88199 EconFreeLag | 1,663 4.674184 2.548064 0 9.234659 PolitFreeLag | 1,663 40.27603 37.29169 0 100 DemocIndxLag | 1,663 2.135558 6.888007 0 37.5 -------------+--------------------------------------------------------- War | 1,663 .0824843 .1886522 0 1 WarLag | 1,663 .0845975 .1900241 0 1 Coup | 1,663 .0911606 .1912969 0 1 CoupLag | 1,663 .0926037 .1936088 0 1 Americas1 | 1,663 .1858088 .3890693 0 1 -------------+--------------------------------------------------------- Africa1 | 1,663 .2375225 .425693 0 1 MENA1 | 1,663 .1292844 .3356154 0 1 Asia1 | 1,663 .1635598 .3699869 0 1 . sum cty Year DemocIndxLag PolitFreeLag EconFreeLag Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- cty | 1,663 80.82441 45.80199 1 160 Year | 1,663 1988.596 18.06104 1955 2015 DemocIndxLag | 1,663 2.135558 6.888007 0 37.5 PolitFreeLag | 1,663 40.27603 37.29169 0 100 EconFreeLag | 1,663 4.674184 2.548064 0 9.234659 . . * NOTE: It looks like a full balanced panel, but it isn't. Some . * missing observations are coded as 0, so we fix this next. . . *---------------------------------------------------- . * Section 4: Recode missing-as-zero codes to actual missing . *---------------------------------------------------- . * The lag-prefixed variables encode unavailable observations as 0. . * mvdecode replaces those zeros with Stata's missing-value marker . * so they are correctly excluded from the regressions below. . mvdecode DemocIndxLag PolitFreeLag EconFreeLag, mv(0) DemocIndxLag: 1438 missing values generated PolitFreeLag: 495 missing values generated EconFreeLag: 314 missing values generated . sum DemocIndxLag PolitFreeLag EconFreeLag Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- DemocIndxLag | 225 15.78415 11.64599 .1 37.5 PolitFreeLag | 1,168 57.34507 31.63668 .0201857 100 EconFreeLag | 1,349 5.762171 1.315748 1.820347 9.234659 . . *---------------------------------------------------- . * Section 5: Label key variables . *---------------------------------------------------- . lab var lnGDPpercapita lnGDPpc . lab var EconFreeLag L.EconFreedom . lab var PolitFreeLag L.PolitFreedom . . *---------------------------------------------------- . * Section 6: Exploratory data analysis . *---------------------------------------------------- . . * 6.1 Descriptive statistics -------------------------- . estpost summarize lnGDPpercapita War Coup EconFreeLag PolitFreeLag, detail | e(count) e(sum_w) e(mean) e(Var) e(sd) -------------+------------------------------------------------------- lnGDPperca~a | 1663 1663 8.768894 1.451637 1.204839 War | 1663 1663 .0824843 .0355896 .1886522 Coup | 1663 1663 .0911606 .0365945 .1912969 EconFreeLag | 1349 1349 5.762171 1.731192 1.315748 PolitFreeLag | 1168 1168 57.34507 1000.879 31.63668 | e(skewn~) e(kurto~) e(sum) e(min) e(max) -------------+------------------------------------------------------- lnGDPperca~a | -.0311522 2.249679 14582.67 5.63479 12.70269 War | 2.532639 8.790373 137.1714 0 1 Coup | 2.5981 10.11349 151.6 0 1 EconFreeLag | -.0206745 2.452404 7773.169 1.820347 9.234659 PolitFreeLag | -.0367516 1.571523 66979.05 .0201857 100 | e(p1) e(p5) e(p10) e(p25) e(p50) -------------+------------------------------------------------------- lnGDPperca~a | 6.375025 6.761573 7.134094 7.80751 8.809415 War | 0 0 0 0 0 Coup | 0 0 0 0 0 EconFreeLag | 2.933702 3.61899 4.072251 4.781393 5.78 PolitFreeLag | 8.333333 8.333333 16.66667 25 58.33333 | e(p75) e(p90) e(p95) e(p99) -------------+-------------------------------------------- lnGDPperca~a | 9.704183 10.36151 10.62925 11.23101 War | .0285714 .3428571 .5714286 .8571429 Coup | .2 .4 .4 1 EconFreeLag | 6.783959 7.447417 7.808637 8.712842 PolitFreeLag | 91.66667 100 100 100 . esttab using "summary_stats.csv", replace /// > cells("count(label(N)) mean(label(Mean) fmt(3)) sd(label(SD) fmt(3)) min( > label(Min) fmt(3)) p50(label(Median) fmt(3)) max(label(Max) fmt(3))") /// > nomtitle nonumber noobs /// > title("Descriptive statistics: dynamic panel dataset") (output written to summary_stats.csv) . . * 6.2 Figure: Mean War & Coup intensity over time ---- . preserve . collapse (mean) War Coup, by(Year) . twoway (line War Year, lcolor("106 155 204") lwidth(thick)) /// > (line Coup Year, lcolor("217 119 87") lwidth(thick)), /// > ytitle("Mean intensity (0 = none, 1 = max)") /// > xtitle("Year") /// > title("Mean war & coup intensity by year, 1955-2015") /// > legend(order(1 "War" 2 "Coup") position(6) cols(2)) /// > graphregion(color(white)) plotregion(color(white)) . graph export "stata_dynamic_panel_war_coup_panel.png", replace width(2400 > ) file stata_dynamic_panel_war_coup_panel.png written in PNG format . restore . . * 6.3 Figure: Number of countries with active war by year . preserve . gen byte war_active = War > 0 & !missing(War) . collapse (sum) war_active, by(Year) . twoway bar war_active Year, /// > bcolor("106 155 204") barwidth(3.5) /// > ytitle("Number of countries with War > 0") /// > xtitle("Year") /// > title("Countries experiencing war by year, 1955-2015") /// > subtitle("Reproducing Figure 1 of Thies & Baum (2020)") /// > graphregion(color(white)) plotregion(color(white)) . graph export "stata_dynamic_panel_war_count_by_year.png", replace width(2 > 400) file stata_dynamic_panel_war_count_by_year.png written in PNG format . restore . . * 6.4 Figure: Distribution of log GDP per capita ----- . histogram lnGDPpercapita, /// > fcolor("106 155 204%70") lcolor("20 20 19") bin(40) /// > ytitle("Density") /// > xtitle("log GDP per capita (lnGDPpercapita)") /// > title("Distribution of log GDP per capita") /// > graphregion(color(white)) plotregion(color(white)) (bin=40, start=5.6347896, width=.17669745) . graph export "stata_dynamic_panel_gdp_distribution.png", replace width(2400) file stata_dynamic_panel_gdp_distribution.png written in PNG format . . *---------------------------------------------------- . * Section 7: Long-run effects program (ssta) . *---------------------------------------------------- . * This program (original code by Prof. Baum) is preserved verbatim. . * After each xtabond2 regression we call `ssta` to compute the . * sum of the contemporaneous and lagged War coefficients (and Coup): . * . * SSwar = b[War] + b[L.War] + b[L2.War] . * . * Economically, SSwar measures the *cumulative* impact of war over . * three quinquennia (15 years) following the shock. We also store . * the standard error and t-statistic of the linear combination. . capture program drop ssta . program ssta, rclass 1. qui { 2. nlcom (_b[War]+_b[L.War]+_b[L2.War]) // / (1-_b[L.lnGDPpercapita]) 3. mat b = r(b) 4. mat v = r(V) 5. estadd scalar SSwar = b[1,1] 6. estadd scalar SSwarSE = sqrt(v[1,1]) 7. estadd scalar SSwarT = b[1,1]/sqrt(v[1,1]) 8. nlcom (_b[Coup]+_b[L.Coup]) // / (1-_b[L.lnGDPpercapita]) 9. mat b = r(b) 10. mat v = r(V) 11. estadd scalar SScoup = b[1,1] 12. estadd scalar SScoupSE = sqrt(v[1,1]) 13. estadd scalar SScoupT = b[1,1]/sqrt(v[1,1]) 14. } 15. end . . local addss SSwar SSwarSE SSwarT SScoup SScoupSE SScoupT . mata: mata set matafavor speed, perm (set matafavor preference recorded) . . *---------------------------------------------------- . * Section 8: Declare panel structure (5-year intervals) . *---------------------------------------------------- . xtset cty Year, delta(5) Panel variable: cty (unbalanced) Time variable: Year, 1955 to 2015, but with a gap Delta: 5 years . xtdescribe cty: 1, 2, ..., 160 n = 160 Year: 1955, 1960, ..., 2015 T = 13 Delta(Year) = 5 years Span(Year) = 13 periods (cty*Year uniquely identifies each observation) Distribution of T_i: min 5% 25% 50% 75% 95% max 1 4 8 12 13 13 13 Freq. Percent Cum. | Pattern ---------------------------+--------------- 77 48.12 48.12 | 1111111111111 27 16.88 65.00 | ..11111111111 20 12.50 77.50 | ........11111 7 4.38 81.88 | .....11111111 6 3.75 85.62 | .........1111 6 3.75 89.38 | .111111111111 4 2.50 91.88 | ...........11 4 2.50 94.38 | ....111111111 4 2.50 96.88 | ...1111111111 5 3.12 100.00 | (other patterns) ---------------------------+--------------- 160 100.00 | XXXXXXXXXXXXX . . *---------------------------------------------------- . * Section 9: Dynamic panel regressions (4 models) . *---------------------------------------------------- . * Each model uses xtabond2 with: . * gmm(...) - which variables are endogenous; the lag(2 6) range . * picks lags 2 to 6 of the variable as INTERNAL . * instruments for the differenced equation. . * iv(...) - strictly exogenous instruments (used as-is). . * noleveleq - estimate the difference equation only (Arellano- . * Bond), do not add the level equation (which would . * be Blundell-Bond / system GMM). . * robust - heteroskedasticity-robust (cluster) standard errors. . * twostep - two-step efficient estimator. . * . * The dependent variable is lnGDPpercapita, and L.lnGDPpercapita . * enters the right-hand side, making this a dynamic panel. . * . * Two warnings will appear in the log after each xtabond2 call. . * Both are normal -- not errors: . * . * "Two-step estimated covariance matrix of moments is singular." . * The unadjusted two-step variance is degenerate, so xtabond2 . * falls back to the Windmeijer (2005) finite-sample correction . * for cluster-robust standard errors. This is exactly what . * `robust twostep` is supposed to do; the reported SEs and . * t-statistics use the corrected formula. . * . * "Number of instruments may be large relative to number of . * observations." . * Roodman (2009) calls this "instrument proliferation": . * too many moment conditions relative to N can over-fit the . * endogenous regressors and weaken the Hansen J test. Our . * lag(2 6) range deliberately limits the GMM lag depth to . * contain this risk; we also report Hansen J p-values below . * so the reader can judge instrument validity. . . eststo clear . . * Model 1: War & Coup, no institutional controls ----- . eststo: xtabond2 L(0/1).lnGDPpercapita L(0/2).War L(0/1).Coup i.Year, /// > gmm(lnGDPpercapita War Coup, lag(2 6)) /// > iv(L(0/2).War L(0/1).Coup) /// > iv(i.Year) /// > noleveleq robust twostep Favoring speed over space. To switch, type or click on mata: mata set matafavor > space, perm. 1955b.Year dropped due to collinearity 1960.Year dropped due to collinearity 2010.Year dropped due to collinearity Warning: Two-step estimated covariance matrix of moments is singular. Using a generalized inverse to calculate optimal weighting matrix for two-ste > p estimation. Difference-in-Sargan/Hansen statistics may be negative. Dynamic panel-data estimation, two-step difference GMM ------------------------------------------------------------------------------ Group variable: cty Number of obs = 1187 Time variable : Year Number of groups = 155 Number of instruments = 146 Obs per group: min = 1 Wald chi2(15) = 3169.64 avg = 7.66 Prob > chi2 = 0.000 max = 10 ------------------------------------------------------------------------------ | Corrected lnGDPperca~a | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- lnGDPperca~a | L1. | .6787863 .051373 13.21 0.000 .5780972 .7794755 | War | --. | -.2186432 .0569308 -3.84 0.000 -.3302255 -.1070609 L1. | -.0655071 .0468636 -1.40 0.162 -.1573581 .0263438 L2. | -.0687009 .0470814 -1.46 0.145 -.1609787 .0235769 | Coup | --. | -.090843 .0284565 -3.19 0.001 -.1466167 -.0350692 L1. | .0387093 .029151 1.33 0.184 -.0184257 .0958443 | Year | 1965 | -.2611315 .0550686 -4.74 0.000 -.369064 -.153199 1970 | -.1940239 .0470275 -4.13 0.000 -.2861962 -.1018517 1975 | -.1497198 .0375196 -3.99 0.000 -.223257 -.0761826 1980 | -.1201203 .0302131 -3.98 0.000 -.1793369 -.0609038 1985 | -.1633711 .0215068 -7.60 0.000 -.2055236 -.1212187 1990 | -.142305 .0210282 -6.77 0.000 -.1835196 -.1010904 1995 | -.1520839 .0229545 -6.63 0.000 -.1970739 -.1070939 2000 | -.0853571 .0184544 -4.63 0.000 -.121527 -.0491872 2005 | -.0216631 .0110329 -1.96 0.050 -.0432872 -.000039 2015 | .0122262 .0122692 1.00 0.319 -.011821 .0362733 ------------------------------------------------------------------------------ Instruments for first differences equation Standard D.(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 1990.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) D.(War L.War L2.War Coup L.Coup) GMM-type (missing=0, separate instruments for each period unless collapsed) L(2/6).(lnGDPpercapita War Coup) ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z = -4.37 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = -1.69 Pr > z = 0.091 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(130) = 280.36 Prob > chi2 = 0.000 (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(130) = 144.32 Prob > chi2 = 0.184 (Robust, but weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: iv(War L.War L2.War Coup L.Coup) Hansen test excluding group: chi2(125) = 142.50 Prob > chi2 = 0.135 Difference (null H = exogenous): chi2(5) = 1.82 Prob > chi2 = 0.874 iv(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 199 > 0.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) Hansen test excluding group: chi2(120) = 128.49 Prob > chi2 = 0.281 Difference (null H = exogenous): chi2(10) = 15.83 Prob > chi2 = 0.105 (est1 stored) . ssta . estimates store m1 . . * Model 2: + lagged Economic Freedom ----------------- . eststo: xtabond2 L(0/1).lnGDPpercapita EconFreeLag L(0/2).War /// > L(0/1).Coup i.Year, /// > gmm(lnGDPpercapita War Coup, lag(2 6)) /// > iv(L(0/2).War L(0/1).Coup) /// > iv(i.Year) /// > iv(EconFreeLag) /// > noleveleq robust twostep Favoring speed over space. To switch, type or click on mata: mata set matafavor > space, perm. 1955b.Year dropped due to collinearity 1960.Year dropped due to collinearity 2015.Year dropped due to collinearity Warning: Number of instruments may be large relative to number of observations. Warning: Two-step estimated covariance matrix of moments is singular. Using a generalized inverse to calculate optimal weighting matrix for two-ste > p estimation. Difference-in-Sargan/Hansen statistics may be negative. Dynamic panel-data estimation, two-step difference GMM ------------------------------------------------------------------------------ Group variable: cty Number of obs = 987 Time variable : Year Number of groups = 137 Number of instruments = 147 Obs per group: min = 0 Wald chi2(16) = 2615.84 avg = 7.20 Prob > chi2 = 0.000 max = 10 ------------------------------------------------------------------------------ | Corrected lnGDPperca~a | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- lnGDPperca~a | L1. | .6660685 .0561399 11.86 0.000 .5560363 .7761006 | EconFreeLag | .0200732 .0077316 2.60 0.009 .0049196 .0352268 | War | --. | -.2385981 .0459219 -5.20 0.000 -.3286033 -.1485929 L1. | -.0197117 .0525133 -0.38 0.707 -.1226358 .0832124 L2. | -.0123415 .0480008 -0.26 0.797 -.1064214 .0817385 | Coup | --. | -.0757088 .0247314 -3.06 0.002 -.1241814 -.0272361 L1. | .0143737 .0261366 0.55 0.582 -.0368531 .0656005 | Year | 1965 | -.2626063 .0647487 -4.06 0.000 -.3895115 -.1357012 1970 | -.2000976 .0554223 -3.61 0.000 -.3087233 -.091472 1975 | -.1637295 .0458253 -3.57 0.000 -.2535455 -.0739135 1980 | -.1137174 .0408697 -2.78 0.005 -.1938205 -.0336143 1985 | -.1655646 .0331111 -5.00 0.000 -.2304612 -.1006679 1990 | -.1265364 .0374968 -3.37 0.001 -.2000287 -.0530441 1995 | -.1101329 .0304156 -3.62 0.000 -.1697464 -.0505194 2000 | -.0840618 .0279849 -3.00 0.003 -.1389111 -.0292125 2005 | -.0353139 .0173954 -2.03 0.042 -.0694082 -.0012195 2010 | -.0198248 .0140098 -1.42 0.157 -.0472835 .007634 ------------------------------------------------------------------------------ Instruments for first differences equation Standard D.EconFreeLag D.(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 1990.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) D.(War L.War L2.War Coup L.Coup) GMM-type (missing=0, separate instruments for each period unless collapsed) L(2/6).(lnGDPpercapita War Coup) ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z = -4.19 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = -0.15 Pr > z = 0.881 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(130) = 315.06 Prob > chi2 = 0.000 (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(130) = 125.02 Prob > chi2 = 0.607 (Robust, but weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: iv(War L.War L2.War Coup L.Coup) Hansen test excluding group: chi2(125) = 120.31 Prob > chi2 = 0.602 Difference (null H = exogenous): chi2(5) = 4.72 Prob > chi2 = 0.452 iv(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 199 > 0.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) Hansen test excluding group: chi2(120) = 122.77 Prob > chi2 = 0.413 Difference (null H = exogenous): chi2(10) = 2.25 Prob > chi2 = 0.994 iv(EconFreeLag) Hansen test excluding group: chi2(129) = 124.45 Prob > chi2 = 0.597 Difference (null H = exogenous): chi2(1) = 0.57 Prob > chi2 = 0.448 (est2 stored) . ssta . estimates store m2 . . * Model 3: + lagged Political Freedom ---------------- . eststo: xtabond2 L(0/1).lnGDPpercapita PolitFreeLag L(0/2).War /// > L(0/1).Coup i.Year, /// > gmm(lnGDPpercapita War Coup, lag(2 6)) /// > iv(L(0/2).War L(0/1).Coup) /// > iv(i.Year) /// > iv(PolitFreeLag) /// > noleveleq robust twostep Favoring speed over space. To switch, type or click on mata: mata set matafavor > space, perm. 1955b.Year dropped due to collinearity 1960.Year dropped due to collinearity 1965.Year dropped due to collinearity 1970.Year dropped due to collinearity 2010.Year dropped due to collinearity Warning: Two-step estimated covariance matrix of moments is singular. Using a generalized inverse to calculate optimal weighting matrix for two-ste > p estimation. Difference-in-Sargan/Hansen statistics may be negative. Dynamic panel-data estimation, two-step difference GMM ------------------------------------------------------------------------------ Group variable: cty Number of obs = 918 Time variable : Year Number of groups = 151 Number of instruments = 130 Obs per group: min = 0 Wald chi2(14) = 1808.34 avg = 6.08 Prob > chi2 = 0.000 max = 8 ------------------------------------------------------------------------------ | Corrected lnGDPperca~a | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- lnGDPperca~a | L1. | .631893 .0524498 12.05 0.000 .5290933 .7346927 | PolitFreeLag | .0002759 .000351 0.79 0.432 -.000412 .0009639 | War | --. | -.1593212 .0368103 -4.33 0.000 -.231468 -.0871744 L1. | -.0763529 .053819 -1.42 0.156 -.1818361 .0291303 L2. | .0113921 .0484824 0.23 0.814 -.0836316 .1064158 | Coup | --. | -.0951722 .0316872 -3.00 0.003 -.1572779 -.0330665 L1. | -.0018327 .0262472 -0.07 0.944 -.0532763 .0496109 | Year | 1975 | -.1717179 .0408648 -4.20 0.000 -.2518115 -.0916244 1980 | -.1353085 .033384 -4.05 0.000 -.20074 -.0698771 1985 | -.1859607 .0263806 -7.05 0.000 -.2376657 -.1342556 1990 | -.1483654 .0272167 -5.45 0.000 -.201709 -.0950217 1995 | -.139652 .024633 -5.67 0.000 -.1879318 -.0913723 2000 | -.0824814 .019337 -4.27 0.000 -.1203813 -.0445816 2005 | -.0289773 .011535 -2.51 0.012 -.0515855 -.0063691 2015 | .0239487 .0117777 2.03 0.042 .0008648 .0470326 ------------------------------------------------------------------------------ Instruments for first differences equation Standard D.PolitFreeLag D.(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 1990.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) D.(War L.War L2.War Coup L.Coup) GMM-type (missing=0, separate instruments for each period unless collapsed) L(2/6).(lnGDPpercapita War Coup) ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z = -3.81 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = -0.24 Pr > z = 0.810 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(115) = 250.08 Prob > chi2 = 0.000 (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(115) = 132.38 Prob > chi2 = 0.128 (Robust, but weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: iv(War L.War L2.War Coup L.Coup) Hansen test excluding group: chi2(110) = 127.38 Prob > chi2 = 0.123 Difference (null H = exogenous): chi2(5) = 5.00 Prob > chi2 = 0.415 iv(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 199 > 0.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) Hansen test excluding group: chi2(107) = 123.93 Prob > chi2 = 0.126 Difference (null H = exogenous): chi2(8) = 8.45 Prob > chi2 = 0.391 iv(PolitFreeLag) Hansen test excluding group: chi2(114) = 131.95 Prob > chi2 = 0.120 Difference (null H = exogenous): chi2(1) = 0.43 Prob > chi2 = 0.510 (est3 stored) . ssta . estimates store m3 . . * Model 4: full specification (both institutions) ---- . eststo: xtabond2 L(0/1).lnGDPpercapita EconFreeLag PolitFreeLag /// > L(0/2).War L(0/1).Coup i.Year, /// > gmm(lnGDPpercapita War Coup, lag(2 6)) /// > iv(L(0/2).War L(0/1).Coup) /// > iv(i.Year) /// > iv(EconFreeLag PolitFreeLag) /// > noleveleq robust twostep Favoring speed over space. To switch, type or click on mata: mata set matafavor > space, perm. 1955b.Year dropped due to collinearity 1960.Year dropped due to collinearity 1965.Year dropped due to collinearity 1970.Year dropped due to collinearity 1980.Year dropped due to collinearity Warning: Two-step estimated covariance matrix of moments is singular. Using a generalized inverse to calculate optimal weighting matrix for two-ste > p estimation. Difference-in-Sargan/Hansen statistics may be negative. Dynamic panel-data estimation, two-step difference GMM ------------------------------------------------------------------------------ Group variable: cty Number of obs = 821 Time variable : Year Number of groups = 137 Number of instruments = 131 Obs per group: min = 0 Wald chi2(15) = 1998.13 avg = 5.99 Prob > chi2 = 0.000 max = 8 ------------------------------------------------------------------------------ | Corrected lnGDPperca~a | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- lnGDPperca~a | L1. | .6192125 .0542553 11.41 0.000 .512874 .725551 | EconFreeLag | .0282808 .0085535 3.31 0.001 .0115162 .0450455 PolitFreeLag | .0001734 .000342 0.51 0.612 -.0004968 .0008437 | War | --. | -.1602315 .041899 -3.82 0.000 -.2423521 -.0781109 L1. | -.0110606 .0586253 -0.19 0.850 -.125964 .1038429 L2. | .005423 .0529713 0.10 0.918 -.0983989 .1092449 | Coup | --. | -.0902476 .0289706 -3.12 0.002 -.1470289 -.0334662 L1. | -.0054947 .0278824 -0.20 0.844 -.0601432 .0491538 | Year | 1975 | -.0598282 .0212708 -2.81 0.005 -.1015183 -.0181381 1985 | -.0584756 .0185438 -3.15 0.002 -.0948208 -.0221303 1990 | -.0113368 .0195435 -0.58 0.562 -.0496413 .0269678 1995 | -.0019037 .0271393 -0.07 0.944 -.0550958 .0512885 2000 | .0219811 .0287636 0.76 0.445 -.0343946 .0783568 2005 | .0617407 .0359388 1.72 0.086 -.008698 .1321794 2010 | .0830273 .0386551 2.15 0.032 .0072647 .1587899 2015 | .1118697 .0436962 2.56 0.010 .0262267 .1975128 ------------------------------------------------------------------------------ Instruments for first differences equation Standard D.(EconFreeLag PolitFreeLag) D.(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 1990.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) D.(War L.War L2.War Coup L.Coup) GMM-type (missing=0, separate instruments for each period unless collapsed) L(2/6).(lnGDPpercapita War Coup) ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z = -3.82 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = 0.49 Pr > z = 0.625 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(115) = 254.14 Prob > chi2 = 0.000 (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(115) = 128.77 Prob > chi2 = 0.179 (Robust, but weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: iv(War L.War L2.War Coup L.Coup) Hansen test excluding group: chi2(110) = 123.93 Prob > chi2 = 0.172 Difference (null H = exogenous): chi2(5) = 4.83 Prob > chi2 = 0.436 iv(1955b.Year 1960.Year 1965.Year 1970.Year 1975.Year 1980.Year 1985.Year 199 > 0.Year 1995.Year 2000.Year 2005.Year 2010.Year 2015.Year) Hansen test excluding group: chi2(107) = 118.09 Prob > chi2 = 0.218 Difference (null H = exogenous): chi2(8) = 10.68 Prob > chi2 = 0.221 iv(EconFreeLag PolitFreeLag) Hansen test excluding group: chi2(113) = 127.79 Prob > chi2 = 0.162 Difference (null H = exogenous): chi2(2) = 0.97 Prob > chi2 = 0.614 (est4 stored) . ssta . estimates store m4 . . *---------------------------------------------------- . * Section 10: Publication-quality regression table . *---------------------------------------------------- . esttab, /// > lab star(* 0.1 ** 0.05 *** 0.01) /// > indicate(Quinquennia effects = *.Year) /// > stat(N N_g `addss' hansen hansen_df hansenp, /// > labels("N" "N. Countries" "Sum War coeff." "s.e. War" "t War" /// > "Sum Coup coeff." "s.e. Coup" "t Coup" /// > "Hansen J" "J d.f." "J pvalue")) /// > ti("Dynamic panel data estimates of log GDP per capita") nomti (tabulating estimates stored by eststo; specify "." to tabulate the active resu > lts) Dynamic panel data estimates of log GDP per capita ------------------------------------------------------------------------------- > ----- (1) (2) (3) ( > 4) ------------------------------------------------------------------------------- > ----- L.lnGDPpc 0.679*** 0.666*** 0.632*** 0.6 > 19*** (13.21) (11.86) (12.05) (11.4 > 1) War -0.219*** -0.239*** -0.159*** -0.1 > 60*** (-3.84) (-5.20) (-4.33) (-3.8 > 2) L.War -0.0655 -0.0197 -0.0764 -0.01 > 11 (-1.40) (-0.38) (-1.42) (-0.1 > 9) L2.War -0.0687 -0.0123 0.0114 0.005 > 42 (-1.46) (-0.26) (0.23) (0.1 > 0) Coup -0.0908*** -0.0757*** -0.0952*** -0.09 > 02*** (-3.19) (-3.06) (-3.00) (-3.1 > 2) L.Coup 0.0387 0.0144 -0.00183 -0.005 > 49 (1.33) (0.55) (-0.07) (-0.2 > 0) L.EconFreedom 0.0201*** 0.02 > 83*** (2.60) (3.3 > 1) L.PolitFreedom 0.000276 0.0001 > 73 (0.79) (0.5 > 1) Quinquennia effects Yes Yes Yes Y > es ------------------------------------------------------------------------------- > ----- N 1187 987 918 8 > 21 N. Countries 155 137 151 1 > 37 Sum War coeff. -0.353 -0.271 -0.224 -0.1 > 66 s.e. War 0.0787 0.0741 0.0751 0.07 > 59 t War -4.482 -3.650 -2.988 -2.1 > 85 Sum Coup coeff. -0.0521 -0.0613 -0.0970 -0.09 > 57 s.e. Coup 0.0356 0.0366 0.0458 0.04 > 40 t Coup -1.463 -1.678 -2.120 -2.1 > 74 Hansen J 144.3 125.0 132.4 128 > .8 J d.f. 130 130 115 1 > 15 J pvalue 0.184 0.607 0.128 0.1 > 79 ------------------------------------------------------------------------------- > ----- t statistics in parentheses * p<0.1, ** p<0.05, *** p<0.01 . . * Rich-text file (preserved from original code) ------ . esttab using catoj2.rtf, replace /// > lab star(* 0.1 ** 0.05 *** 0.01) /// > indicate(Quinquennia effects = *.Year) /// > stat(N N_g `addss' hansen hansen_df hansenp, /// > labels("N" "N. Countries" "Sum War coeff." "s.e. War" "t War" /// > "Sum Coup coeff." "s.e. Coup" "t Coup" /// > "Hansen J" "J d.f." "J pvalue")) /// > ti("Dynamic panel data estimates of log GDP per capita") nomti (tabulating estimates stored by eststo; specify "." to tabulate the active resu > lts) (output written to catoj2.rtf) . . * CSV version of the regression results -------------- . esttab using "regression_results.csv", replace /// > lab star(* 0.1 ** 0.05 *** 0.01) /// > indicate(Quinquennia effects = *.Year) /// > stat(N N_g `addss' hansen hansen_df hansenp, /// > labels("N" "N. Countries" "Sum War coeff." "s.e. War" "t War" /// > "Sum Coup coeff." "s.e. Coup" "t Coup" /// > "Hansen J" "J d.f." "J pvalue")) /// > ti("Dynamic panel data estimates of log GDP per capita") nomti (tabulating estimates stored by eststo; specify "." to tabulate the active resu > lts) (output written to regression_results.csv) . . *---------------------------------------------------- . * Section 11: Coefficient plot for War (across models) . *---------------------------------------------------- . coefplot (m1, label("Model 1")) /// > (m2, label("Model 2")) /// > (m3, label("Model 3")) /// > (m4, label("Model 4")), /// > keep(War L.War L2.War) /// > xline(0, lcolor("217 119 87") lpattern(dash)) /// > xtitle("Coefficient on log GDP per capita") /// > title("War coefficients across the four dynamic-panel models") /// > subtitle("Contemporaneous, lag-1, and lag-2 effects of War") /// > graphregion(color(white)) plotregion(color(white)) /// > ciopts(lcolor("106 155 204") recast(rcap)) /// > mcolor("106 155 204") . graph export "stata_dynamic_panel_war_coef_plot.png", replace width(2400) file stata_dynamic_panel_war_coef_plot.png written in PNG format . . *---------------------------------------------------- . * Section 12: Long-run sum-of-coefficients plot . *---------------------------------------------------- . * Build a small dataset from the stored ssta scalars. . preserve . clear . set obs 4 Number of observations (_N) was 0, now 4. . gen model = _n . gen str20 mname = "" (4 missing values generated) . gen sswar = . (4 missing values generated) . gen sswar_se = . (4 missing values generated) . gen sscoup = . (4 missing values generated) . gen sscoup_se = . (4 missing values generated) . . forvalues i = 1/4 { 2. estimates restore m`i' 3. replace mname = "Model `i'" in `i' 4. replace sswar = e(SSwar) in `i' 5. replace sswar_se = e(SSwarSE) in `i' 6. replace sscoup = e(SScoup) in `i' 7. replace sscoup_se = e(SScoupSE) in `i' 8. } (results m1 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m2 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m3 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m4 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) . . gen sswar_lo = sswar - 1.96*sswar_se . gen sswar_hi = sswar + 1.96*sswar_se . gen sscoup_lo = sscoup - 1.96*sscoup_se . gen sscoup_hi = sscoup + 1.96*sscoup_se . . export delimited using "longrun_effects.csv", replace file longrun_effects.csv saved . . twoway /// > (bar sswar model, fcolor("106 155 204%70") lcolor("106 155 204") bar > width(0.6)) /// > (rcap sswar_lo sswar_hi model, lcolor("20 20 19")), /// > yline(0, lcolor("217 119 87") lpattern(dash)) /// > xlabel(1 "Model 1" 2 "Model 2" 3 "Model 3" 4 "Model 4") /// > ytitle("Sum of War coefficients (long-run effect)") /// > xtitle("") /// > title("Long-run effect of War on log GDP per capita") /// > subtitle("Sum of contemporaneous + L1 + L2 War coefficients, 95% CI") > /// > legend(off) /// > graphregion(color(white)) plotregion(color(white)) . graph export "stata_dynamic_panel_longrun_effects.png", replace width(240 > 0) file stata_dynamic_panel_longrun_effects.png written in PNG format . restore . . *---------------------------------------------------- . * Section 13: Diagnostic test summary . *---------------------------------------------------- . preserve . clear . set obs 4 Number of observations (_N) was 0, now 4. . gen model = _n . gen str20 mname = "" (4 missing values generated) . gen ar2_p = . (4 missing values generated) . gen hansen_p = . (4 missing values generated) . gen hansen = . (4 missing values generated) . gen hansen_df = . (4 missing values generated) . . forvalues i = 1/4 { 2. estimates restore m`i' 3. replace mname = "Model `i'" in `i' 4. replace ar2_p = e(ar2p) in `i' 5. replace hansen_p = e(hansenp) in `i' 6. replace hansen = e(hansen) in `i' 7. replace hansen_df = e(hansen_df) in `i' 8. } (results m1 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m2 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m3 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (results m4 are active now) (1 real change made) (1 real change made) (1 real change made) (1 real change made) (1 real change made) . . export delimited using "diagnostics.csv", replace file diagnostics.csv saved . . * Dodge the two bar series so AR(2) and Hansen J appear side-by-side . * rather than stacked at the same x-coordinate. . gen ar2_x = model - 0.2 . gen hansen_x = model + 0.2 . . twoway /// > (bar ar2_p ar2_x, fcolor("106 155 204%70") lcolor("106 155 204" > ) barwidth(0.35)) /// > (bar hansen_p hansen_x, fcolor("217 119 87%70") lcolor("217 119 87") > barwidth(0.35)), /// > yline(0.05, lcolor("20 20 19") lpattern(dash)) /// > xlabel(1 "Model 1" 2 "Model 2" 3 "Model 3" 4 "Model 4") /// > ytitle("p-value") /// > xtitle("") /// > title("Diagnostic tests: p-values by model") /// > subtitle("AR(2) test (blue) and Hansen J test (orange); want p > 0.05 > ") /// > legend(order(1 "AR(2) p" 2 "Hansen J p") position(6) cols(2)) /// > graphregion(color(white)) plotregion(color(white)) . graph export "stata_dynamic_panel_diagnostics.png", replace width(2400) file stata_dynamic_panel_diagnostics.png written in PNG format . restore . . *---------------------------------------------------- . * End . *---------------------------------------------------- . display _newline "=== Script completed successfully ===" === Script completed successfully === . log close name: log: /Users/carlos/GitHub/starter-academic-v501/content/post/stata_dyna > mic_panel/analysis.log log type: text closed on: 29 Apr 2026, 10:10:35 -------------------------------------------------------------------------------