------------------------------------------------------------------------------- name: log: /Users/carlosmendez/Documents/GitHub/starter-academic-v501/content > /post/stata_iv_panel/analysis.log log type: text opened on: 27 Apr 2026, 14:53:23 . . // Install required packages (idempotent) . capture ssc install estout, replace . capture ssc install ivreg2, replace . capture ssc install ranktest, replace . capture ssc install xtivreg2, replace . capture ssc install outreg2, replace . capture ssc install schemepack, replace . . // Set graph scheme for clean output . set scheme white_tableau . . // Site color palette . local steel_blue "68 155 204" . local warm_orange "217 119 87" . local near_black "20 20 19" . local teal "0 212 200" . . display _n "=== Setup complete ===" === Setup complete === . . . // ── 1. Data Loading ─────────────────────────────────────────────────────── > ── . . display _n "=== Section 1: Data Loading ===" === Section 1: Data Loading === . . use "reference/EL_regional_conflict_replication.dta", clear . . // Declare panel structure . tsset objectid year Panel variable: objectid (unbalanced) Time variable: year, 1994 to 2010, but with gaps Delta: 1 unit . . // Basic data overview . describe Contains data from reference/EL_regional_conflict_replication.dta Observations: 96,591 Variables: 14 5 Sep 2022 01:06 (_dta has notes) ------------------------------------------------------------------------------- Variable Storage Display Value name type format label Variable label ------------------------------------------------------------------------------- objectid long %12.0g Value year float %9.0g countrycode str3 %9s ISO countryname str32 %32s NAME_0 ucdp_death_du~y float %9.0g Conflict (>1 deaths) ucdp_25death_~y float %9.0g Conflict (>25 deaths) llnlight01 float %9.0g Ln Lights(t-1) l2lnrain01 float %9.0g Ln Rain(t-2) l2meanpdsi float %9.0g (Non) Drought(t-2) ucdp_death_du~t float %9.0g Conflict (>1 deaths) ucdp_25death_~t float %9.0g Conflict (>25 deaths) llnlight01_dt float %9.0g Ln Lights(t-1) l2lnrain01_dt float %9.0g Ln Rain(t-2) l2meanpdsi_dt float %9.0g (Non) Drought(t-2) ------------------------------------------------------------------------------- Sorted by: objectid year . display _n "Panel structure:" Panel structure: . display " Regions (objectid): " _N Regions (objectid): 96591 . xtsum objectid Variable | Mean Std. dev. Min Max | Observations -----------------+--------------------------------------------+---------------- objectid overall | 19459.86 10626.6 329 39162 | N = 96591 between | 10625.27 329 39162 | n = 5689 within | 0 19459.86 19459.86 | T-bar = 16.9786 . . // Export raw data summary . preserve . collapse (mean) ucdp_death_dummy ucdp_25death_dummy llnlight01 l2lnrain01 > l2meanpdsi, by(year) . export delimited using "yearly_averages.csv", replace (file yearly_averages.csv not found) file yearly_averages.csv saved . restore . . display _n "=== Data loading complete ===" === Data loading complete === . . . // ── 2. Descriptive Statistics (Table 1) ─────────────────────────────────── > ── . . display _n "=== Section 2: Descriptive Statistics ===" === Section 2: Descriptive Statistics === . . // Panel summary statistics (between and within variation) . xtsum ucdp_death_dummy ucdp_25death_dummy llnlight01 l2lnrain01 l2meanpdsi Variable | Mean Std. dev. Min Max | Observations -----------------+--------------------------------------------+---------------- ucdp_d~y overall | .0455425 .2084919 0 1 | N = 96591 between | .1176404 0 1 | n = 5689 within | .1721562 -.8956339 .986719 | T-bar = 16.9786 | | ucdp_2~y overall | .0144527 .1193481 0 1 | N = 96591 between | .0575476 0 .9411765 | n = 5689 within | .1045561 -.9267238 .9556292 | T-bar = 16.9786 | | llnli~01 overall | -1.611658 2.619427 -4.60517 4.143293 | N = 96591 between | 2.568635 -4.60517 4.140281 | n = 5689 within | .5277626 -7.699739 2.693339 | T-bar = 16.9786 | | l2lnr~01 overall | 3.8302 1.477743 -4.60517 6.093216 | N = 96591 between | 1.493749 -4.60517 5.514849 | n = 5689 within | .1993702 -2.741027 5.494656 | T-bar = 16.9786 | | l2mean~i overall | -1.215386 2.033711 -12.1292 12.6313 | N = 96591 between | 1.064342 -6.024061 2.538623 | n = 5689 within | 1.733175 -8.565653 10.55658 | T-bar = 16.9786 . . // Standard summary statistics . summarize ucdp_death_dummy ucdp_25death_dummy llnlight01 l2lnrain01 l2meanpds > i Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- ucdp_death~y | 96,591 .0455425 .2084919 0 1 ucdp_25dea~y | 96,591 .0144527 .1193481 0 1 llnlight01 | 96,591 -1.611658 2.619427 -4.60517 4.143293 l2lnrain01 | 96,591 3.8302 1.477743 -4.60517 6.093216 l2meanpdsi | 96,591 -1.215386 2.033711 -12.1292 12.6313 . . // Formatted Table 1 using estout . eststo clear . generate y_temp = uniform() . quietly reg y_temp ucdp_death_dummy ucdp_25death_dummy llnlight01 l2lnrain01 > l2meanpdsi, noconstant . quietly estadd summ . esttab, label /// > cells("mean(fmt(%8.3f)) sd(fmt(%8.3f)) min(fmt(%8.3f)) max(fmt(%8.3f))") > /// > collabels("Mean" "Std.Dev." "Min" "Max") /// > addnote("Notes: Sample period is 1994-2010. N = 96,591 region-year observ > ations.") /// > nogap nomtitle nonumber noobs compress -------------------------------------------------------- Mean Std.Dev. Min Max -------------------------------------------------------- Conflict (>1 d~) 0.046 0.208 0.000 1.000 Conflict (>25 ~) 0.014 0.119 0.000 1.000 Ln Lights(t-1) -1.612 2.619 -4.605 4.143 Ln Rain(t-2) 3.830 1.478 -4.605 6.093 (Non) Drought~2) -1.215 2.034 -12.129 12.631 -------------------------------------------------------- Notes: Sample period is 1994-2010. N = 96,591 region-year observations. . drop y_temp . eststo clear . . // Export Table 1 as CSV . preserve . collapse (mean) mean_conflict01=ucdp_death_dummy mean_conflict25=ucdp_25d > eath_dummy /// > mean_light=llnlight01 mean_rain=l2lnrain01 mean_drought=l2meanpd > si /// > (sd) sd_conflict01=ucdp_death_dummy sd_conflict25=ucdp_25death_d > ummy /// > sd_light=llnlight01 sd_rain=l2lnrain01 sd_drought=l2meanpdsi /// > (min) min_conflict01=ucdp_death_dummy min_conflict25=ucdp_25deat > h_dummy /// > min_light=llnlight01 min_rain=l2lnrain01 min_drought=l2meanpdsi > /// > (max) max_conflict01=ucdp_death_dummy max_conflict25=ucdp_25deat > h_dummy /// > max_light=llnlight01 max_rain=l2lnrain01 max_drought=l2meanpdsi . generate n = _N . export delimited using "table1_summary_stats.csv", replace (file table1_summary_stats.csv not found) file table1_summary_stats.csv saved . restore . . display _n "=== Descriptive statistics complete ===" === Descriptive statistics complete === . . . // ── 3. OLS with Fixed Effects — Reduced Form (Tables 2-3, Cols 1-4) ─────── > ─ . . display _n "=== Section 3: OLS with Fixed Effects ===" === Section 3: OLS with Fixed Effects === . . // Generate year dummies for inclusion in regressions . quietly tab year, gen(Iyear) . . // --- Table 2: Conflict with 1+ deaths --- . display _n "--- Table 2: Conflict (1+ deaths) ---" --- Table 2: Conflict (1+ deaths) --- . . eststo clear . . // Col 1: Light -> Conflict . eststo t2m1: quietly xtreg ucdp_death_dummy_dt llnlight01_dt Iyear*, fe robus > t cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 2: Rain -> Conflict (reduced form) . eststo t2m2: quietly xtreg ucdp_death_dummy_dt l2lnrain01_dt Iyear*, fe robus > t cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 3: Drought -> Conflict (reduced form) . eststo t2m3: quietly xtreg ucdp_death_dummy_dt l2meanpdsi_dt Iyear*, fe robus > t cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 4: Rain + Drought -> Conflict (reduced form) . eststo t2m4: quietly xtreg ucdp_death_dummy_dt l2meanpdsi_dt l2lnrain01_dt Iy > ear*, fe robust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . display _n "OLS estimates for Table 2 (cols 1-4) stored." OLS estimates for Table 2 (cols 1-4) stored. . . // --- Table 3: Conflict with 25+ deaths --- . display _n "--- Table 3: Conflict (25+ deaths) ---" --- Table 3: Conflict (25+ deaths) --- . . // Col 1: Light -> Conflict25 . eststo t3m1: quietly xtreg ucdp_25death_dummy_dt llnlight01_dt Iyear*, fe rob > ust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 2: Rain -> Conflict25 (reduced form) . eststo t3m2: quietly xtreg ucdp_25death_dummy_dt l2lnrain01_dt Iyear*, fe rob > ust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 3: Drought -> Conflict25 (reduced form) . eststo t3m3: quietly xtreg ucdp_25death_dummy_dt l2meanpdsi_dt Iyear*, fe rob > ust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . // Col 4: Rain + Drought -> Conflict25 (reduced form) . eststo t3m4: quietly xtreg ucdp_25death_dummy_dt l2meanpdsi_dt l2lnrain01_dt > Iyear*, fe robust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "None", replace . . display _n "OLS estimates for Table 3 (cols 1-4) stored." OLS estimates for Table 3 (cols 1-4) stored. . . . // ── 4. 2SLS/IV with Fixed Effects (Tables 2-3, Cols 5-7) ────────────────── > ─ . . display _n "=== Section 4: 2SLS/IV Estimation ===" === Section 4: 2SLS/IV Estimation === . display _n "Estimand: Causal effect of economic shocks on conflict probabilit > y" Estimand: Causal effect of economic shocks on conflict probability . display _n "Instruments: Lagged rainfall (t-2) and drought intensity (t-2)" Instruments: Lagged rainfall (t-2) and drought intensity (t-2) . display _n "Exclusion restriction: Weather(t-2) -> Light(t-1) -> Conflict(t)" Exclusion restriction: Weather(t-2) -> Light(t-1) -> Conflict(t) . . // --- Table 2 (continued): 2SLS for Conflict 1+ --- . display _n "--- Table 2: 2SLS Estimates (cols 5-7) ---" --- Table 2: 2SLS Estimates (cols 5-7) --- . . // Col 5: IV with Rain . eststo t2m5: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2lnrain01_d > t) Iyear*, fe robust cluster(objectid) first . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Rain(t-2)", replace . . // Col 6: IV with Drought . eststo t2m6: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_d > t) Iyear*, fe robust cluster(objectid) first . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Drought(t-2)", replace . . // Col 7: IV with Both instruments . eststo t2m7: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_d > t l2lnrain01_dt) Iyear*, fe robust cluster(objectid) first . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Both", replace . . // Display Table 2 (full) . display _n "=== TABLE 2: Effects on regional conflicts (1+ deaths) ===" === TABLE 2: Effects on regional conflicts (1+ deaths) === . #delimit ; delimiter now ; . esttab t2m1 t2m2 t2m3 t2m4 t2m5 t2m6 t2m7, > keep(llnlight01_dt l2lnrain01_dt l2meanpdsi_dt) > se > label > stats(N N_g r2 FE_region Trend_region FE_year Instrument, > fmt(0 0 2) > label("Observations" "N Regions" "R-squared" "Region FE" "Region tren > d" "Year FE" "Instrument")) > mtitles("OLS" "OLS" "OLS" "OLS" "2SLS" "2SLS" "2SLS") > nonotes > addnote("Notes: Sample period is 1994-2010. Standard errors adjusted for > clustering at the regional level." > "Dependent variable: Conflict with 1+ deaths. Standard errors in > parentheses." > "* p<0.10, ** p<0.05, *** p<0.01") > star(* 0.10 ** 0.05 *** 0.01) > b(%7.3f) > compress > replace ; ------------------------------------------------------------------------------- > ---------------------------- (1) (2) (3) (4) (5) > (6) (7) OLS OLS OLS OLS 2SLS > 2SLS 2SLS ------------------------------------------------------------------------------- > ---------------------------- Ln Lights(t-1) 0.001 -0.303* > ** -0.293*** -0.296*** (0.001) (0.111) > (0.085) (0.076) Ln Rain(t-2) -0.011*** -0.007* > (0.003) (0.004) > (Non) Drought~2) -0.002*** -0.001*** > (0.000) (0.000) > ------------------------------------------------------------------------------- > ---------------------------- Observations 96591 96591 96591 96591 96591 > 96591 96591 N Regions 5689 5689 5689 5689 5689 > 5689 5689 R-squared 0.00 0.00 0.00 0.00 -0.54 > -0.51 -0.52 Region FE Yes Yes Yes Yes Yes > Yes Yes Region trend Yes Yes Yes Yes Yes > Yes Yes Year FE Yes Yes Yes Yes Yes > Yes Yes Instrument None None None None Rain(t-2) > Drought(t-2) Both ------------------------------------------------------------------------------- > ---------------------------- Notes: Sample period is 1994-2010. Standard errors adjusted for clustering at t > he regional level. Dependent variable: Conflict with 1+ deaths. Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01 . #delimit cr delimiter now cr . . // --- Table 3 (continued): 2SLS for Conflict 25+ --- . display _n "--- Table 3: 2SLS Estimates (cols 5-7) ---" --- Table 3: 2SLS Estimates (cols 5-7) --- . . // Col 5: IV with Rain . eststo t3m5: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2lnrain01 > _dt) Iyear*, fe robust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Rain(t-2)", replace . . // Col 6: IV with Drought . eststo t3m6: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2meanpdsi > _dt) Iyear*, fe robust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Drought(t-2)", replace . . // Col 7: IV with Both instruments . eststo t3m7: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2meanpdsi > _dt l2lnrain01_dt) Iyear*, fe robust cluster(objectid) . quietly estadd local FE_region "Yes", replace . quietly estadd local Trend_region "Yes", replace . quietly estadd local FE_year "Yes", replace . quietly estadd local Instrument "Both", replace . . // Display Table 3 (full) . display _n "=== TABLE 3: Effects on regional conflicts (25+ deaths) ===" === TABLE 3: Effects on regional conflicts (25+ deaths) === . #delimit ; delimiter now ; . esttab t3m1 t3m2 t3m3 t3m4 t3m5 t3m6 t3m7, > keep(llnlight01_dt l2lnrain01_dt l2meanpdsi_dt) > se > label > stats(N N_g r2 FE_region Trend_region FE_year Instrument, > fmt(0 0 2) > label("Observations" "N Regions" "R-squared" "Region FE" "Region tren > d" "Year FE" "Instrument")) > mtitles("OLS" "OLS" "OLS" "OLS" "2SLS" "2SLS" "2SLS") > nonotes > addnote("Notes: Sample period is 1994-2010. Standard errors adjusted for > clustering at the regional level." > "Dependent variable: Conflict with 25+ deaths. Standard errors in > parentheses." > "* p<0.10, ** p<0.05, *** p<0.01") > star(* 0.10 ** 0.05 *** 0.01) > b(%7.3f) > compress > replace ; ------------------------------------------------------------------------------- > ---------------------------- (1) (2) (3) (4) (5) > (6) (7) OLS OLS OLS OLS 2SLS > 2SLS 2SLS ------------------------------------------------------------------------------- > ---------------------------- Ln Lights(t-1) 0.001 -0.092 > -0.093** -0.093** (0.001) (0.057) > (0.046) (0.040) Ln Rain(t-2) -0.003* -0.002 > (0.002) (0.002) > (Non) Drought~2) -0.001** -0.000* > (0.000) (0.000) > ------------------------------------------------------------------------------- > ---------------------------- Observations 96591 96591 96591 96591 96591 > 96591 96591 N Regions 5689 5689 5689 5689 5689 > 5689 5689 R-squared 0.00 0.00 0.00 0.00 -0.14 > -0.14 -0.14 Region FE Yes Yes Yes Yes Yes > Yes Yes Region trend Yes Yes Yes Yes Yes > Yes Yes Year FE Yes Yes Yes Yes Yes > Yes Yes Instrument None None None None Rain(t-2) > Drought(t-2) Both ------------------------------------------------------------------------------- > ---------------------------- Notes: Sample period is 1994-2010. Standard errors adjusted for clustering at t > he regional level. Dependent variable: Conflict with 25+ deaths. Standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01 . #delimit cr delimiter now cr . . eststo clear . . . // ── 5. First-Stage Results and Diagnostics (Table 4) ────────────────────── > ── . . display _n "=== Section 5: First-Stage Results and IV Diagnostics ===" === Section 5: First-Stage Results and IV Diagnostics === . . // First-stage with Rain as instrument . display _n "--- First Stage: Rain(t-2) -> Light(t-1) ---" --- First Stage: Rain(t-2) -> Light(t-1) --- . eststo fs1: xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2lnrain01_dt) Iyear* > , fe robust cluster(objectid) first Warning - collinearities detected Vars dropped: Iyear17 FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 Warning - collinearities detected Vars dropped: Iyear17 First-stage regressions ----------------------- FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 First-stage regression of llnlight01_dt: Statistics robust to heteroskedasticity and clustering on objectid Number of obs = 96591 Number of clusters (objectid) = 5689 ------------------------------------------------------------------------------ | Robust llnlight01~t | Coefficient std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- l2lnrain01~t | .0360693 .0072692 4.96 0.000 .0218217 .0503169 Iyear1 | .1621198 .0071988 22.52 0.000 .1480103 .1762293 Iyear2 | -.2300151 .0096504 -23.83 0.000 -.2489299 -.2111003 Iyear3 | .2088411 .0066666 31.33 0.000 .1957745 .2219076 Iyear4 | .2011741 .0068272 29.47 0.000 .1877929 .2145554 Iyear5 | -.007674 .0074362 -1.03 0.302 -.0222488 .0069009 Iyear6 | .0517506 .0078119 6.62 0.000 .0364394 .0670619 Iyear7 | .0510439 .0082413 6.19 0.000 .0348911 .0671967 Iyear8 | -.0048014 .0080271 -0.60 0.550 -.0205345 .0109316 Iyear9 | .2272877 .0090691 25.06 0.000 .2095123 .245063 Iyear10 | .2289018 .0087508 26.16 0.000 .2117504 .2460533 Iyear11 | -.0930477 .0080567 -11.55 0.000 -.1088388 -.0772566 Iyear12 | -.0007095 .008474 -0.08 0.933 -.0173184 .0158994 Iyear13 | -.0077135 .0081401 -0.95 0.343 -.0236681 .008241 Iyear14 | .0743706 .008446 8.81 0.000 .0578165 .0909247 Iyear15 | .1775274 .0073881 24.03 0.000 .1630468 .1920081 Iyear16 | .1039426 .006106 17.02 0.000 .0919749 .1159103 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ F test of excluded instruments: F( 1, 5688) = 24.62 Prob > F = 0.0000 Sanderson-Windmeijer multivariate F test of excluded instruments: F( 1, 5688) = 24.62 Prob > F = 0.0000 Summary results for first-stage regressions ------------------------------------------- (Underid) (Weak id) Variable | F( 1, 5688) P-val | SW Chi-sq( 1) P-val | SW F( 1, 5688) llnlight01_d | 24.62 0.0000 | 24.63 0.0000 | 24.62 NB: first-stage test statistics cluster-robust Stock-Yogo weak ID F test critical values for single endogenous regressor: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for i.i.d. errors only. Underidentification test Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified) Ha: matrix has rank=K1 (identified) Kleibergen-Paap rk LM statistic Chi-sq(1)=27.44 P-val=0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 24.17 Kleibergen-Paap Wald rk F statistic 24.62 Stock-Yogo weak ID test critical values for K1=1 and L1=1: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. Weak-instrument-robust inference Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(1,5688)= 10.54 P-val=0.0012 Anderson-Rubin Wald test Chi-sq(1)= 10.54 P-val=0.0012 Stock-Wright LM S statistic Chi-sq(1)= 10.37 P-val=0.0013 NB: Underidentification, weak identification and weak-identification-robust test statistics cluster-robust Number of clusters N_clust = 5689 Number of observations N = 96591 Number of regressors K = 17 Number of endogenous regressors K1 = 1 Number of instruments L = 17 Number of excluded instruments L1 = 1 IV (2SLS) estimation -------------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on objectid Number of clusters (objectid) = 5689 Number of obs = 96591 F( 17, 5688) = 10.17 Prob > F = 0.0000 Total (centered) SS = 2512.033049 Centered R2 = -0.5442 Total (uncentered) SS = 2512.033049 Uncentered R2 = -0.5442 Residual SS = 3879.177476 Root MSE = .2066 ------------------------------------------------------------------------------ | Robust ucdp_death~t | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- llnlight01~t | -.3033282 .1107854 -2.74 0.006 -.5204636 -.0861928 Iyear1 | .0587702 .0182029 3.23 0.001 .0230932 .0944472 Iyear2 | -.0743488 .0259151 -2.87 0.004 -.1251416 -.0235561 Iyear3 | .0518044 .0230236 2.25 0.024 .006679 .0969297 Iyear4 | .0604805 .0225695 2.68 0.007 .0162452 .1047159 Iyear5 | .0257732 .0037826 6.81 0.000 .0183594 .0331871 Iyear6 | .0345351 .0072224 4.78 0.000 .0203795 .0486907 Iyear7 | .0347076 .0066864 5.19 0.000 .0216025 .0478127 Iyear8 | .010068 .0038102 2.64 0.008 .0026002 .0175358 Iyear9 | .0911123 .0250906 3.63 0.000 .0419357 .1402889 Iyear10 | .089894 .02545 3.53 0.000 .0400129 .1397751 Iyear11 | -.0176401 .0111655 -1.58 0.114 -.039524 .0042439 Iyear12 | -.0001902 .003797 -0.05 0.960 -.0076322 .0072518 Iyear13 | -.0020389 .0037135 -0.55 0.583 -.0093173 .0052394 Iyear14 | .0280824 .008687 3.23 0.001 .0110561 .0451086 Iyear15 | .0636768 .020099 3.17 0.002 .0242834 .1030702 Iyear16 | .0401534 .0119012 3.37 0.001 .0168274 .0634793 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ Underidentification test (Kleibergen-Paap rk LM statistic): 27.443 Chi-sq(1) P-val = 0.0000 ------------------------------------------------------------------------------ Weak identification test (Cragg-Donald Wald F statistic): 24.173 (Kleibergen-Paap rk Wald F statistic): 24.621 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. ------------------------------------------------------------------------------ Hansen J statistic (overidentification test of all instruments): 0.000 (equation exactly identified) ------------------------------------------------------------------------------ Instrumented: llnlight01_dt Included instruments: Iyear1 Iyear2 Iyear3 Iyear4 Iyear5 Iyear6 Iyear7 Iyear8 Iyear9 Iyear10 Iyear11 Iyear12 Iyear13 Iyear14 Iyear15 Iyear16 Excluded instruments: l2lnrain01_dt Dropped collinear: Iyear17 ------------------------------------------------------------------------------ . display _n "First-stage F-stat (Rain): " e(widstat) First-stage F-stat (Rain): 24.620605 . . // First-stage with Drought as instrument . display _n "--- First Stage: Drought(t-2) -> Light(t-1) ---" --- First Stage: Drought(t-2) -> Light(t-1) --- . eststo fs2: xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_dt) Iyear* > , fe robust cluster(objectid) first Warning - collinearities detected Vars dropped: Iyear17 FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 Warning - collinearities detected Vars dropped: Iyear17 First-stage regressions ----------------------- FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 First-stage regression of llnlight01_dt: Statistics robust to heteroskedasticity and clustering on objectid Number of obs = 96591 Number of clusters (objectid) = 5689 ------------------------------------------------------------------------------ | Robust llnlight01~t | Coefficient std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- l2meanpdsi~t | .0055157 .0008685 6.35 0.000 .0038134 .007218 Iyear1 | .1617902 .0072031 22.46 0.000 .1476722 .1759083 Iyear2 | -.2299293 .009624 -23.89 0.000 -.2487921 -.2110664 Iyear3 | .211015 .0066875 31.55 0.000 .1979074 .2241225 Iyear4 | .2012739 .0068281 29.48 0.000 .187891 .2146569 Iyear5 | -.012051 .0075541 -1.60 0.111 -.0268568 .0027549 Iyear6 | .0516968 .0078389 6.59 0.000 .0363325 .067061 Iyear7 | .0463421 .0082258 5.63 0.000 .0302195 .0624647 Iyear8 | -.0040847 .0080478 -0.51 0.612 -.0198583 .011689 Iyear9 | .2242962 .0089647 25.02 0.000 .2067255 .2418668 Iyear10 | .2291772 .0087421 26.22 0.000 .2120429 .2463116 Iyear11 | -.0895078 .0081125 -11.03 0.000 -.1054081 -.0736075 Iyear12 | -.0018213 .0084927 -0.21 0.830 -.0184669 .0148244 Iyear13 | -.006848 .0081407 -0.84 0.400 -.0228037 .0091077 Iyear14 | .0719372 .0083867 8.58 0.000 .0554994 .088375 Iyear15 | .1775356 .0073867 24.03 0.000 .1630577 .1920135 Iyear16 | .1025331 .0061098 16.78 0.000 .0905579 .1145083 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ F test of excluded instruments: F( 1, 5688) = 40.33 Prob > F = 0.0000 Sanderson-Windmeijer multivariate F test of excluded instruments: F( 1, 5688) = 40.33 Prob > F = 0.0000 Summary results for first-stage regressions ------------------------------------------- (Underid) (Weak id) Variable | F( 1, 5688) P-val | SW Chi-sq( 1) P-val | SW F( 1, 5688) llnlight01_d | 40.33 0.0000 | 40.35 0.0000 | 40.33 NB: first-stage test statistics cluster-robust Stock-Yogo weak ID F test critical values for single endogenous regressor: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for i.i.d. errors only. Underidentification test Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified) Ha: matrix has rank=K1 (identified) Kleibergen-Paap rk LM statistic Chi-sq(1)=39.98 P-val=0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 40.65 Kleibergen-Paap Wald rk F statistic 40.33 Stock-Yogo weak ID test critical values for K1=1 and L1=1: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. Weak-instrument-robust inference Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(1,5688)= 17.24 P-val=0.0000 Anderson-Rubin Wald test Chi-sq(1)= 17.24 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(1)= 18.32 P-val=0.0000 NB: Underidentification, weak identification and weak-identification-robust test statistics cluster-robust Number of clusters N_clust = 5689 Number of observations N = 96591 Number of regressors K = 17 Number of endogenous regressors K1 = 1 Number of instruments L = 17 Number of excluded instruments L1 = 1 IV (2SLS) estimation -------------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on objectid Number of clusters (objectid) = 5689 Number of obs = 96591 F( 17, 5688) = 10.87 Prob > F = 0.0000 Total (centered) SS = 2512.033049 Centered R2 = -0.5081 Total (uncentered) SS = 2512.033049 Uncentered R2 = -0.5081 Residual SS = 3788.362167 Root MSE = .2041 ------------------------------------------------------------------------------ | Robust ucdp_death~t | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- llnlight01~t | -.2931316 .0848165 -3.46 0.001 -.459369 -.1268943 Iyear1 | .0571301 .0139031 4.11 0.000 .0298805 .0843796 Iyear2 | -.0719871 .0199655 -3.61 0.000 -.1111187 -.0328554 Iyear3 | .0496888 .0178446 2.78 0.005 .0147139 .0846636 Iyear4 | .0584275 .0171679 3.40 0.001 .024779 .092076 Iyear5 | .0258154 .0037327 6.92 0.000 .0184995 .0331314 Iyear6 | .0339741 .0060435 5.62 0.000 .022129 .0458192 Iyear7 | .0341996 .005533 6.18 0.000 .0233551 .0450441 Iyear8 | .0100926 .0037465 2.69 0.007 .0027495 .0174356 Iyear9 | .0888318 .0193818 4.58 0.000 .0508441 .1268194 Iyear10 | .0875773 .0195214 4.49 0.000 .0493161 .1258386 Iyear11 | -.0166849 .0088815 -1.88 0.060 -.0340924 .0007225 Iyear12 | -.000206 .0037316 -0.06 0.956 -.0075197 .0071077 Iyear13 | -.0019693 .0036307 -0.54 0.588 -.0090853 .0051467 Iyear14 | .0273638 .0070506 3.88 0.000 .0135448 .0411827 Iyear15 | .0618618 .0153748 4.02 0.000 .0317278 .0919958 Iyear16 | .0390875 .0092247 4.24 0.000 .0210074 .0571676 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ Underidentification test (Kleibergen-Paap rk LM statistic): 39.982 Chi-sq(1) P-val = 0.0000 ------------------------------------------------------------------------------ Weak identification test (Cragg-Donald Wald F statistic): 40.648 (Kleibergen-Paap rk Wald F statistic): 40.332 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 15% maximal IV size 8.96 20% maximal IV size 6.66 25% maximal IV size 5.53 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. ------------------------------------------------------------------------------ Hansen J statistic (overidentification test of all instruments): 0.000 (equation exactly identified) ------------------------------------------------------------------------------ Instrumented: llnlight01_dt Included instruments: Iyear1 Iyear2 Iyear3 Iyear4 Iyear5 Iyear6 Iyear7 Iyear8 Iyear9 Iyear10 Iyear11 Iyear12 Iyear13 Iyear14 Iyear15 Iyear16 Excluded instruments: l2meanpdsi_dt Dropped collinear: Iyear17 ------------------------------------------------------------------------------ . display _n "First-stage F-stat (Drought): " e(widstat) First-stage F-stat (Drought): 40.332128 . . // First-stage with Both instruments . display _n "--- First Stage: Both instruments -> Light(t-1) ---" --- First Stage: Both instruments -> Light(t-1) --- . eststo fs3: xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_dt l2lnrai > n01_dt) Iyear*, fe robust cluster(objectid) first Warning - collinearities detected Vars dropped: Iyear17 FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 Warning - collinearities detected Vars dropped: Iyear17 First-stage regressions ----------------------- FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 5689 Obs per group: min = 16 avg = 17.0 max = 17 First-stage regression of llnlight01_dt: Statistics robust to heteroskedasticity and clustering on objectid Number of obs = 96591 Number of clusters (objectid) = 5689 ------------------------------------------------------------------------------ | Robust llnlight01~t | Coefficient std. err. t P>|t| [95% conf. interval] -------------+---------------------------------------------------------------- l2meanpdsi~t | .0046142 .0009215 5.01 0.000 .0028081 .0064203 l2lnrain01~t | .0231511 .0075003 3.09 0.002 .0084506 .0378515 Iyear1 | .1624512 .0072044 22.55 0.000 .1483307 .1765717 Iyear2 | -.2291742 .0096265 -23.81 0.000 -.248042 -.2103063 Iyear3 | .2113099 .0066896 31.59 0.000 .1981984 .2244214 Iyear4 | .2011767 .0068264 29.47 0.000 .187797 .2145565 Iyear5 | -.0130269 .0075358 -1.73 0.084 -.027797 .0017432 Iyear6 | .0501433 .0078196 6.41 0.000 .0348168 .0654697 Iyear7 | .0476971 .0082704 5.77 0.000 .0314873 .0639069 Iyear8 | -.005346 .0080194 -0.67 0.505 -.0210639 .0103719 Iyear9 | .2265218 .0090565 25.01 0.000 .2087711 .2442726 Iyear10 | .2299452 .0087585 26.25 0.000 .2127786 .2471117 Iyear11 | -.0897874 .0081099 -11.07 0.000 -.1056827 -.0738921 Iyear12 | -.0027209 .0084898 -0.32 0.749 -.0193609 .0139191 Iyear13 | -.0074102 .008139 -0.91 0.363 -.0233626 .0085422 Iyear14 | .0741982 .0084486 8.78 0.000 .0576389 .0907574 Iyear15 | .1773088 .0073862 24.01 0.000 .162832 .1917856 Iyear16 | .1024814 .006114 16.76 0.000 .0904981 .1144647 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ F test of excluded instruments: F( 2, 5688) = 25.32 Prob > F = 0.0000 Sanderson-Windmeijer multivariate F test of excluded instruments: F( 2, 5688) = 25.32 Prob > F = 0.0000 Summary results for first-stage regressions ------------------------------------------- (Underid) (Weak id) Variable | F( 2, 5688) P-val | SW Chi-sq( 2) P-val | SW F( 2, 5688) llnlight01_d | 25.32 0.0000 | 50.66 0.0000 | 25.32 NB: first-stage test statistics cluster-robust Stock-Yogo weak ID F test critical values for single endogenous regressor: 10% maximal IV size 19.93 15% maximal IV size 11.59 20% maximal IV size 8.75 25% maximal IV size 7.25 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for i.i.d. errors only. Underidentification test Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified) Ha: matrix has rank=K1 (identified) Kleibergen-Paap rk LM statistic Chi-sq(2)=50.68 P-val=0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 24.76 Kleibergen-Paap Wald rk F statistic 25.32 Stock-Yogo weak ID test critical values for K1=1 and L1=2: 10% maximal IV size 19.93 15% maximal IV size 11.59 20% maximal IV size 8.75 25% maximal IV size 7.25 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. Weak-instrument-robust inference Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(2,5688)= 11.23 P-val=0.0000 Anderson-Rubin Wald test Chi-sq(2)= 22.46 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(2)= 23.38 P-val=0.0000 NB: Underidentification, weak identification and weak-identification-robust test statistics cluster-robust Number of clusters N_clust = 5689 Number of observations N = 96591 Number of regressors K = 17 Number of endogenous regressors K1 = 1 Number of instruments L = 18 Number of excluded instruments L1 = 2 IV (2SLS) estimation -------------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on objectid Number of clusters (objectid) = 5689 Number of obs = 96591 F( 17, 5688) = 10.88 Prob > F = 0.0000 Total (centered) SS = 2512.033049 Centered R2 = -0.5193 Total (uncentered) SS = 2512.033049 Uncentered R2 = -0.5193 Residual SS = 3816.486116 Root MSE = .2049 ------------------------------------------------------------------------------ | Robust ucdp_death~t | Coefficient std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- llnlight01~t | -.2963267 .0755672 -3.92 0.000 -.4444357 -.1482177 Iyear1 | .057644 .0125288 4.60 0.000 .0330879 .0822001 Iyear2 | -.0727271 .0178633 -4.07 0.000 -.1077386 -.0377156 Iyear3 | .0503517 .0158858 3.17 0.002 .019216 .0814873 Iyear4 | .0590708 .0154049 3.83 0.000 .0288777 .089264 Iyear5 | .0258022 .0037411 6.90 0.000 .0184698 .0331347 Iyear6 | .0341499 .0056678 6.03 0.000 .0230413 .0452585 Iyear7 | .0343588 .005235 6.56 0.000 .0240983 .0446193 Iyear8 | .0100849 .003764 2.68 0.007 .0027076 .0174622 Iyear9 | .0895464 .0173613 5.16 0.000 .0555188 .1235739 Iyear10 | .0883033 .0174895 5.05 0.000 .0540245 .1225821 Iyear11 | -.0169842 .0081401 -2.09 0.037 -.0329384 -.00103 Iyear12 | -.0002011 .003751 -0.05 0.957 -.0075529 .0071508 Iyear13 | -.0019911 .0036369 -0.55 0.584 -.0091193 .0051371 Iyear14 | .0275889 .0065209 4.23 0.000 .0148082 .0403697 Iyear15 | .0624306 .013838 4.51 0.000 .0353086 .0895525 Iyear16 | .0394215 .0083195 4.74 0.000 .0231156 .0557275 Iyear17 | 0 (omitted) ------------------------------------------------------------------------------ Underidentification test (Kleibergen-Paap rk LM statistic): 50.681 Chi-sq(2) P-val = 0.0000 ------------------------------------------------------------------------------ Weak identification test (Cragg-Donald Wald F statistic): 24.763 (Kleibergen-Paap rk Wald F statistic): 25.319 Stock-Yogo weak ID test critical values: 10% maximal IV size 19.93 15% maximal IV size 11.59 20% maximal IV size 8.75 25% maximal IV size 7.25 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. ------------------------------------------------------------------------------ Hansen J statistic (overidentification test of all instruments): 0.007 Chi-sq(1) P-val = 0.9319 ------------------------------------------------------------------------------ Instrumented: llnlight01_dt Included instruments: Iyear1 Iyear2 Iyear3 Iyear4 Iyear5 Iyear6 Iyear7 Iyear8 Iyear9 Iyear10 Iyear11 Iyear12 Iyear13 Iyear14 Iyear15 Iyear16 Excluded instruments: l2meanpdsi_dt l2lnrain01_dt Dropped collinear: Iyear17 ------------------------------------------------------------------------------ . display _n "First-stage F-stat (Both): " e(widstat) First-stage F-stat (Both): 25.318906 . display _n "Hansen J statistic: " e(j) Hansen J statistic: .0073071 . display _n "Hansen J p-value: " e(jp) Hansen J p-value: .93187856 . . // Display IV diagnostics summary . display _n "=== IV DIAGNOSTICS SUMMARY ===" === IV DIAGNOSTICS SUMMARY === . display "Stock-Yogo 10% critical value (1 endogenous, 1 instrument): 16.38" Stock-Yogo 10% critical value (1 endogenous, 1 instrument): 16.38 . display "Stock-Yogo 10% critical value (1 endogenous, 2 instruments): 19.93" Stock-Yogo 10% critical value (1 endogenous, 2 instruments): 19.93 . . eststo clear . . . // ── 6. Visualizations ───────────────────────────────────────────────────── > ── . . display _n "=== Section 6: Visualizations ===" === Section 6: Visualizations === . . // --- Figure 1: OLS vs 2SLS Coefficient Comparison (Conflict 1+) --- . display _n "--- Figure 1: OLS vs 2SLS Coefficient Comparison ---" --- Figure 1: OLS vs 2SLS Coefficient Comparison --- . . // Re-estimate to capture scalars . quietly xtreg ucdp_death_dummy_dt llnlight01_dt Iyear*, fe robust cluster(obj > ectid) . local ols_b1 = _b[llnlight01_dt] . local ols_se1 = _se[llnlight01_dt] . . quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2lnrain01_dt) Iyear*, fe > robust cluster(objectid) . local iv_rain_b = _b[llnlight01_dt] . local iv_rain_se = _se[llnlight01_dt] . . quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_dt) Iyear*, fe > robust cluster(objectid) . local iv_drought_b = _b[llnlight01_dt] . local iv_drought_se = _se[llnlight01_dt] . . quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_dt l2lnrain01_ > dt) Iyear*, fe robust cluster(objectid) . local iv_both_b = _b[llnlight01_dt] . local iv_both_se = _se[llnlight01_dt] . . // Build coefficient plot data . preserve . clear . set obs 4 Number of observations (_N) was 0, now 4. . generate method = "" (4 missing values generated) . generate coef = . (4 missing values generated) . generate se = . (4 missing values generated) . generate ci_lo = . (4 missing values generated) . generate ci_hi = . (4 missing values generated) . generate xpos = _n . . replace method = "OLS" in 1 variable method was str1 now str3 (1 real change made) . replace coef = `ols_b1' in 1 (1 real change made) . replace se = `ols_se1' in 1 (1 real change made) . . replace method = "2SLS: Rain" in 2 variable method was str3 now str10 (1 real change made) . replace coef = `iv_rain_b' in 2 (1 real change made) . replace se = `iv_rain_se' in 2 (1 real change made) . . replace method = "2SLS: Drought" in 3 variable method was str10 now str13 (1 real change made) . replace coef = `iv_drought_b' in 3 (1 real change made) . replace se = `iv_drought_se' in 3 (1 real change made) . . replace method = "2SLS: Both" in 4 (1 real change made) . replace coef = `iv_both_b' in 4 (1 real change made) . replace se = `iv_both_se' in 4 (1 real change made) . . replace ci_lo = coef - 1.96 * se (4 real changes made) . replace ci_hi = coef + 1.96 * se (4 real changes made) . . // Export for CSV . export delimited using "coef_comparison_conflict01.csv", replace (file coef_comparison_conflict01.csv not found) file coef_comparison_conflict01.csv saved . . // Plot . twoway (bar coef xpos if xpos == 1, barwidth(0.6) color("`steel_blue'") f > intensity(80)) /// > (bar coef xpos if xpos > 1, barwidth(0.6) color("`warm_orange'") f > intensity(80)) /// > (rcap ci_hi ci_lo xpos, lcolor("`near_black'") lwidth(medthick)) / > // > , xlabel(1 "OLS" 2 "2SLS: Rain" 3 "2SLS: Drought" 4 "2SLS: Both", > labsize(small)) /// > ylabel(, format(%5.2f)) /// > ytitle("Coefficient on Light(t-1)") /// > title("OLS vs 2SLS: Effect of Economic Activity on Conflict (1+ > deaths)") /// > subtitle("95% confidence intervals shown") /// > yline(0, lpattern(dash) lcolor(gs10)) /// > legend(off) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Source: Hodler & Raschky (2014). 96,591 region-year observ > ations, 5,689 regions.") . graph export "stata_iv_panel_coef_comparison.png", replace width(2400) (file stata_iv_panel_coef_comparison.png not found) file stata_iv_panel_coef_comparison.png written in PNG format . restore . . display _n "Figure 1 exported: stata_iv_panel_coef_comparison.png" Figure 1 exported: stata_iv_panel_coef_comparison.png . . . // --- Figure 2: Reduced-Form Evidence (Weather -> Conflict) --- . display _n "--- Figure 2: Reduced-Form Coefficients ---" --- Figure 2: Reduced-Form Coefficients --- . . // Reduced-form coefficients for both conflict measures . quietly xtreg ucdp_death_dummy_dt l2lnrain01_dt Iyear*, fe robust cluster(obj > ectid) . local rf_rain_c01 = _b[l2lnrain01_dt] . local rf_rain_se_c01 = _se[l2lnrain01_dt] . . quietly xtreg ucdp_death_dummy_dt l2meanpdsi_dt Iyear*, fe robust cluster(obj > ectid) . local rf_drought_c01 = _b[l2meanpdsi_dt] . local rf_drought_se_c01 = _se[l2meanpdsi_dt] . . quietly xtreg ucdp_25death_dummy_dt l2lnrain01_dt Iyear*, fe robust cluster(o > bjectid) . local rf_rain_c25 = _b[l2lnrain01_dt] . local rf_rain_se_c25 = _se[l2lnrain01_dt] . . quietly xtreg ucdp_25death_dummy_dt l2meanpdsi_dt Iyear*, fe robust cluster(o > bjectid) . local rf_drought_c25 = _b[l2meanpdsi_dt] . local rf_drought_se_c25 = _se[l2meanpdsi_dt] . . preserve . clear . set obs 4 Number of observations (_N) was 0, now 4. . generate str30 instrument = "" (4 missing values generated) . generate str30 outcome = "" (4 missing values generated) . generate coef = . (4 missing values generated) . generate se = . (4 missing values generated) . generate ci_lo = . (4 missing values generated) . generate ci_hi = . (4 missing values generated) . generate xpos = . (4 missing values generated) . . replace instrument = "Rain(t-2)" in 1 (1 real change made) . replace outcome = "Conflict 1+" in 1 (1 real change made) . replace coef = `rf_rain_c01' in 1 (1 real change made) . replace xpos = 1 in 1 (1 real change made) . replace se = `rf_rain_se_c01' in 1 (1 real change made) . . replace instrument = "Drought(t-2)" in 2 (1 real change made) . replace outcome = "Conflict 1+" in 2 (1 real change made) . replace coef = `rf_drought_c01' in 2 (1 real change made) . replace xpos = 2 in 2 (1 real change made) . replace se = `rf_drought_se_c01' in 2 (1 real change made) . . replace instrument = "Rain(t-2)" in 3 (1 real change made) . replace outcome = "Conflict 25+" in 3 (1 real change made) . replace coef = `rf_rain_c25' in 3 (1 real change made) . replace xpos = 3.5 in 3 (1 real change made) . replace se = `rf_rain_se_c25' in 3 (1 real change made) . . replace instrument = "Drought(t-2)" in 4 (1 real change made) . replace outcome = "Conflict 25+" in 4 (1 real change made) . replace coef = `rf_drought_c25' in 4 (1 real change made) . replace xpos = 4.5 in 4 (1 real change made) . replace se = `rf_drought_se_c25' in 4 (1 real change made) . . replace ci_lo = coef - 1.96 * se (4 real changes made) . replace ci_hi = coef + 1.96 * se (4 real changes made) . . export delimited using "reduced_form_coefficients.csv", replace (file reduced_form_coefficients.csv not found) file reduced_form_coefficients.csv saved . . twoway (bar coef xpos if instrument == "Rain(t-2)", barwidth(0.6) color(" > `steel_blue'") fintensity(80)) /// > (bar coef xpos if instrument == "Drought(t-2)", barwidth(0.6) colo > r("`warm_orange'") fintensity(80)) /// > (rcap ci_hi ci_lo xpos, lcolor("`near_black'") lwidth(medthick)) / > // > , xlabel(1 `" "Rain" "(1+ deaths)" "' 2 `" "Drought" "(1+ deaths)" > "' 3.5 `" "Rain" "(25+ deaths)" "' 4.5 `" "Drought" "(25+ deaths)" "', labsi > ze(small)) /// > ylabel(, format(%6.4f)) /// > ytitle("Reduced-Form Coefficient") /// > title("Reduced-Form Evidence: Weather Shocks and Conflict") /// > subtitle("Direct effect of instruments on conflict outcomes") // > / > yline(0, lpattern(dash) lcolor(gs10)) /// > legend(order(1 "Rain(t-2)" 2 "Drought(t-2)") rows(1) position(6) > ) /// > graphregion(color(white)) plotregion(color(white)) /// > note("All coefficients are negative: higher rainfall / lower dro > ught -> less conflict." /// > "95% confidence intervals. Clustered standard errors at reg > ional level.") . graph export "stata_iv_panel_reduced_form.png", replace width(2400) (file stata_iv_panel_reduced_form.png not found) file stata_iv_panel_reduced_form.png written in PNG format . restore . . display _n "Figure 2 exported: stata_iv_panel_reduced_form.png" Figure 2 exported: stata_iv_panel_reduced_form.png . . . // --- Figure 3: First-Stage Relationship (Weather -> Light) --- . display _n "--- Figure 3: First-Stage Binned Scatter ---" --- Figure 3: First-Stage Binned Scatter --- . . // Binned scatter: rainfall residuals vs light residuals . // First partial out year FEs . quietly reg llnlight01_dt Iyear* . predict light_resid, residuals . quietly reg l2lnrain01_dt Iyear* . predict rain_resid, residuals . quietly reg l2meanpdsi_dt Iyear* . predict drought_resid, residuals . . // Create binned scatter for rain -> light . preserve . xtile rain_bin = rain_resid, nq(50) . collapse (mean) light_resid rain_resid, by(rain_bin) . . twoway (scatter light_resid rain_resid, mcolor("`steel_blue'") msize(medi > um) msymbol(circle)) /// > (lfit light_resid rain_resid, lcolor("`warm_orange'") lwidth(thick > )) /// > , ytitle("Nighttime Light Intensity (residual)") /// > xtitle("Rainfall (residual)") /// > title("First Stage: Rainfall Predicts Economic Activity") /// > subtitle("Binned scatter plot (50 bins), partialing out year fix > ed effects") /// > legend(off) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Positive slope: higher rainfall -> higher economic activit > y." /// > "Data: 96,591 region-year obs from 5,689 African regions, 1 > 994-2010.") . graph export "stata_iv_panel_first_stage_rain.png", replace width(2400) (file stata_iv_panel_first_stage_rain.png not found) file stata_iv_panel_first_stage_rain.png written in PNG format . restore . . // Create binned scatter for drought -> light . preserve . xtile drought_bin = drought_resid, nq(50) . collapse (mean) light_resid drought_resid, by(drought_bin) . . twoway (scatter light_resid drought_resid, mcolor("`steel_blue'") msize(m > edium) msymbol(circle)) /// > (lfit light_resid drought_resid, lcolor("`warm_orange'") lwidth(th > ick)) /// > , ytitle("Nighttime Light Intensity (residual)") /// > xtitle("Palmer Drought Severity Index (residual)") /// > title("First Stage: Drought Intensity Predicts Economic Activity > ") /// > subtitle("Binned scatter plot (50 bins), partialing out year fix > ed effects") /// > legend(off) /// > graphregion(color(white)) plotregion(color(white)) /// > note("Positive slope: less drought (higher PDSI) -> higher econo > mic activity." /// > "Data: 96,591 region-year obs from 5,689 African regions, 1 > 994-2010.") . graph export "stata_iv_panel_first_stage_drought.png", replace width(2400 > ) (file stata_iv_panel_first_stage_drought.png not found) file stata_iv_panel_first_stage_drought.png written in PNG format . restore . . drop light_resid rain_resid drought_resid . . display _n "Figures 3a-3b exported: first-stage binned scatter plots" Figures 3a-3b exported: first-stage binned scatter plots . . . // --- Figure 4: Conflict Prevalence Over Time --- . display _n "--- Figure 4: Conflict Prevalence Over Time ---" --- Figure 4: Conflict Prevalence Over Time --- . . preserve . collapse (mean) conflict01=ucdp_death_dummy conflict25=ucdp_25death_dummy > , by(year) . // Scale to percentage . replace conflict01 = conflict01 * 100 (17 real changes made) . replace conflict25 = conflict25 * 100 (17 real changes made) . . export delimited using "conflict_prevalence_by_year.csv", replace (file conflict_prevalence_by_year.csv not found) file conflict_prevalence_by_year.csv saved . . twoway (connected conflict01 year, lcolor("`steel_blue'") mcolor("`steel_ > blue'") lwidth(medthick) msymbol(circle)) /// > (connected conflict25 year, lcolor("`warm_orange'") mcolor("`warm_ > orange'") lwidth(medthick) msymbol(triangle)) /// > , ytitle("Share of Regions with Conflict (%)") /// > xtitle("Year") /// > title("Prevalence of Regional Conflict in Africa, 1992-2010") // > / > legend(order(1 "1+ deaths" 2 "25+ deaths") rows(1) position(6)) > /// > graphregion(color(white)) plotregion(color(white)) /// > ylabel(0(2)10, format(%3.1f)) /// > xlabel(1992(2)2010) /// > note("Source: UCDP Georeferenced Event Dataset. 5,689 subnationa > l regions, 53 African countries.") . graph export "stata_iv_panel_conflict_prevalence.png", replace width(2400 > ) (file stata_iv_panel_conflict_prevalence.png not found) file stata_iv_panel_conflict_prevalence.png written in PNG format . restore . . display _n "Figure 4 exported: stata_iv_panel_conflict_prevalence.png" Figure 4 exported: stata_iv_panel_conflict_prevalence.png . . . // ── 7. Export Regression Tables as CSV ──────────────────────────────────── > ── . . display _n "=== Section 7: CSV Table Exports ===" === Section 7: CSV Table Exports === . . // Re-estimate all models for Table 2 export . eststo clear . . eststo t2m1: quietly xtreg ucdp_death_dummy_dt llnlight01_dt Iyear*, fe robus > t cluster(objectid) . eststo t2m2: quietly xtreg ucdp_death_dummy_dt l2lnrain01_dt Iyear*, fe robus > t cluster(objectid) . eststo t2m3: quietly xtreg ucdp_death_dummy_dt l2meanpdsi_dt Iyear*, fe robus > t cluster(objectid) . eststo t2m4: quietly xtreg ucdp_death_dummy_dt l2meanpdsi_dt l2lnrain01_dt Iy > ear*, fe robust cluster(objectid) . eststo t2m5: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2lnrain01_d > t) Iyear*, fe robust cluster(objectid) . eststo t2m6: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_d > t) Iyear*, fe robust cluster(objectid) . eststo t2m7: quietly xtivreg2 ucdp_death_dummy_dt (llnlight01_dt=l2meanpdsi_d > t l2lnrain01_dt) Iyear*, fe robust cluster(objectid) . . #delimit ; delimiter now ; . esttab t2m1 t2m2 t2m3 t2m4 t2m5 t2m6 t2m7 using "table2_conflict01.csv", > keep(llnlight01_dt l2lnrain01_dt l2meanpdsi_dt) > se > label > stats(N N_g r2, fmt(0 0 3) label("Observations" "N Regions" "R-squared")) > mtitles("OLS" "OLS" "OLS" "OLS" "2SLS-Rain" "2SLS-Drought" "2SLS-Both") > star(* 0.10 ** 0.05 *** 0.01) > b(%9.4f) > csv > replace ; (file table2_conflict01.csv not found) (output written to table2_conflict01.csv) . #delimit cr delimiter now cr . . eststo clear . . // Re-estimate all models for Table 3 export . eststo t3m1: quietly xtreg ucdp_25death_dummy_dt llnlight01_dt Iyear*, fe rob > ust cluster(objectid) . eststo t3m2: quietly xtreg ucdp_25death_dummy_dt l2lnrain01_dt Iyear*, fe rob > ust cluster(objectid) . eststo t3m3: quietly xtreg ucdp_25death_dummy_dt l2meanpdsi_dt Iyear*, fe rob > ust cluster(objectid) . eststo t3m4: quietly xtreg ucdp_25death_dummy_dt l2meanpdsi_dt l2lnrain01_dt > Iyear*, fe robust cluster(objectid) . eststo t3m5: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2lnrain01 > _dt) Iyear*, fe robust cluster(objectid) . eststo t3m6: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2meanpdsi > _dt) Iyear*, fe robust cluster(objectid) . eststo t3m7: quietly xtivreg2 ucdp_25death_dummy_dt (llnlight01_dt=l2meanpdsi > _dt l2lnrain01_dt) Iyear*, fe robust cluster(objectid) . . #delimit ; delimiter now ; . esttab t3m1 t3m2 t3m3 t3m4 t3m5 t3m6 t3m7 using "table3_conflict25.csv", > keep(llnlight01_dt l2lnrain01_dt l2meanpdsi_dt) > se > label > stats(N N_g r2, fmt(0 0 3) label("Observations" "N Regions" "R-squared")) > mtitles("OLS" "OLS" "OLS" "OLS" "2SLS-Rain" "2SLS-Drought" "2SLS-Both") > star(* 0.10 ** 0.05 *** 0.01) > b(%9.4f) > csv > replace ; (file table3_conflict25.csv not found) (output written to table3_conflict25.csv) . #delimit cr delimiter now cr . . eststo clear . . display _n "=== CSV table exports complete ===" === CSV table exports complete === . . . // ── 8. Summary ──────────────────────────────────────────────────────────── > ── . . display _n "=============================================" ============================================= . display "=== ANALYSIS SUMMARY ===" === ANALYSIS SUMMARY === . display "=============================================" ============================================= . display _n "Paper: Hodler & Raschky (2014, Economics Letters)" Paper: Hodler & Raschky (2014, Economics Letters) . display "Topic: Economic shocks and civil conflict at the regional level" Topic: Economic shocks and civil conflict at the regional level . display "Method: Panel FE with 2SLS/IV estimation" Method: Panel FE with 2SLS/IV estimation . display _n "Data: 96,591 region-year observations" Data: 96,591 region-year observations . display " 5,689 subnational regions" 5,689 subnational regions . display " 53 African countries" 53 African countries . display " Period: 1994-2010" Period: 1994-2010 . display _n "Key findings:" Key findings: . display " - OLS: Light(t-1) has near-zero, insignificant effect on confli > ct" - OLS: Light(t-1) has near-zero, insignificant effect on conflict . display " - 2SLS: Light(t-1) coefficient ~ -0.30 for Conflict 1+ (p<0.01) > " - 2SLS: Light(t-1) coefficient ~ -0.30 for Conflict 1+ (p<0.01) . display " - 2SLS: Light(t-1) coefficient ~ -0.09 for Conflict 25+ (p<0.05 > )" - 2SLS: Light(t-1) coefficient ~ -0.09 for Conflict 25+ (p<0.05) . display " - 10% drop in nightlights -> ~3pp increase in conflict risk" - 10% drop in nightlights -> ~3pp increase in conflict risk . display " - First-stage F-stats exceed Stock-Yogo 10% critical value (16. > 38)" - First-stage F-stats exceed Stock-Yogo 10% critical value (16.38) . display " - OLS-2SLS gap consistent with attenuation bias from measuremen > t error" - OLS-2SLS gap consistent with attenuation bias from measurement error . display _n "Figures generated: 5 PNG files" Figures generated: 5 PNG files . display "Tables generated: 5 CSV files" Tables generated: 5 CSV files . . display _n(2) "=== Script completed successfully ===" === Script completed successfully === . . log close name: log: /Users/carlosmendez/Documents/GitHub/starter-academic-v501/content > /post/stata_iv_panel/analysis.log log type: text closed on: 27 Apr 2026, 14:55:16 -------------------------------------------------------------------------------