***Mixed electoral systems and the proportionality-concentration trade-off
***Johannes Raabe, University of Kiel


***Chapter 7 - international comparison
clear all
sysuse international_electoral_system_data

sort country year election_in_this_year
by country: gen el_no = _n
egen id = group(country)
xtset id el_no


**Generating variables for the analysis
*Electoral system general types
replace mixed_type = 0 if mixed_type == .
gen mmm = 0
replace mmm = 1 if mixed_type == 2 | mixed_type == 5

gen mmp = 0
replace mmp = 1 if mixed_type == 4

gen PR = 0
replace PR = 1 if legislative_type == 2

gen Majoritarian = 0
replace Majoritarian = 1 if legislative_type == 1

gen dm_mean = pr_dm_mean
replace dm_mean = 1 if percent_smds == 1
replace dm_mean = ((percent_smds*1 + (1-percent_smds)*pr_dm_mean)) if mmm == 1

gen PR_high_dm = 0
replace PR_high_dm = 1 if PR == 1 & dm_mean >= 10

gen PR_mod_dm = 0
replace PR_mod_dm = 1 if PR == 1 & dm_mean < 10

*Replacements in order to improve comparability or merge variables to eliminate missing data
replace enpv = enpv_pr if enpv == .

replace lsi = lsi*100
replace lsi = . if country == "Italy" & year == 2001

drop if lsi == . | enps == .

replace pr_topup = 0 if pr_topup == .

*Generating additional variables
gen western = 0
replace western = 1 if region == 5 | region == 10 | region == 8

gen mmpXpercent_smds = mmp*percent_smds

gen dm_mean2 = dm_mean^2

gen legal_threshold = pr_threshold
replace legal_threshold = 0 if legal_threshold == .

gen eff_thresh = 1/dm_mean
replace eff_thresh = legal_threshold/100 if eff_thresh < legal_threshold/100 & legal_threshold > 0

gen strategic_manipulation = 0
replace strategic_manipulation = 1 if country == "Lesotho" & year == 2007
replace strategic_manipulation = 1 if country == "Albania" & year == 2005
replace strategic_manipulation = 1 if country == "Venezuela" & year == 2000

gen compens_pr_tier = 0
replace compens_pr_tier = 1 if mmp == 1

gen compensXperc_smds = compens_pr_tier*percent_smds

*BOBW Benchmarks
*Medians
sum enps, d
gen enps_diff_from_median = enps - r(p50)

sum lsi, d
gen lsi_diff_from_median = lsi - r(p50)

gen good_lsi_med = 0 if lsi != .
replace good_lsi_med = 1 if lsi_diff_from_median < 0

gen good_enps_med = 0 if enps != .
replace good_enps_med = 1 if enps_diff_from_median < 0

gen good_twice_med = 0 if enps_diff_from_median != . & lsi_diff_from_median != .
replace good_twice_med = 1 if enps_diff_from_median < 0 & lsi_diff_from_median < 0

*Means
sum enps
gen enps_diff_from_mean = enps - r(mean)

sum lsi
gen lsi_diff_from_mean = lsi - r(mean)

gen good_lsi = 0 if lsi != .
replace good_lsi = 1 if lsi_diff_from_mean < 0

gen good_enps = 0 if enps != .
replace good_enps = 1 if enps_diff_from_mean < 0

gen good_twice = 0 if enps_diff_from_mean != . & lsi_diff_from_mean != .
replace good_twice = 1 if enps_diff_from_mean < 0 & lsi_diff_from_mean < 0

*Distance from ideal point (ENPS = 2, LSI = 0)
gen dist_from_ideal = sqrt((lsi)^2 + (enps - 2)^2)

sum lsi
gen lsi_normalized = lsi/r(sd)
sum enps
gen enps_normalized = enps/r(sd)

gen dist_from_ideal_norm = sqrt((lsi_normalized)^2 + (enps_normalized - 2)^2)

*WOBW Benchmarks
*Medians
gen bad_lsi_med = 0 if lsi != .
replace bad_lsi_med = 1 if lsi_diff_from_median > 0

gen bad_enps_med = 0 if enps != .
replace bad_enps_med = 1 if enps_diff_from_median > 0

gen bad_twice_med = 0 if enps_diff_from_median != . & lsi_diff_from_median != .
replace bad_twice_med = 1 if enps_diff_from_median > 0 & lsi_diff_from_median > 0

*Means
gen bad_lsi = 0 if lsi != .
replace bad_lsi = 1 if lsi_diff_from_mean > 0

gen bad_enps = 0 if enps != .
replace bad_enps = 1 if enps_diff_from_mean > 0

gen bad_twice = 0 if enps_diff_from_mean != . & lsi_diff_from_mean != .
replace bad_twice = 1 if enps_diff_from_mean > 0 & lsi_diff_from_mean > 0


**Empirical analysis
*Descriptive overview
*Scatterplots (cross hairs based on medians)
twoway (scatter lsi enps if mmp == 1, msymbol(C) mcolor(black) msize(small) legend(label(1 "MMP")) yline(4.68, lpattern(dash)) xline(3.17, lpattern(dash))) (scatter lsi enps if mmm == 1, msymbol(Sh) mcolor(black) msize(small) legend(label(2 "MMM")) graphregion(color(white))) (scatter lsi enps if PR_mod_dm == 1, msymbol(Th) mcolor(black) msize(small) legend(label(3 "PR_mod_dm"))) (scatter lsi enps if PR_high_dm == 1, msymbol(Ch) mcolor(black) msize(small) legend(label(4 "PR_high_dm"))) (scatter lsi enps if Majoritarian == 1, msymbol(X) mcolor(black) msize(small) legend(label(5 "Maj")))
twoway (scatter lsi enps if mmp == 1 & lsi < 20 & enps < 10, msymbol(C) mcolor(black) msize(small) legend(label(1 "MMP")) ylab(,nogrid) yline(4.68, lpattern(dash)) xline(3.17, lpattern(dash))) (scatter lsi enps if mmm == 1 & lsi < 20 & enps < 10, msymbol(Sh) mcolor(black) msize(small) legend(label(2 "MMM")) graphregion(color(white))) (scatter lsi enps if PR_mod_dm == 1 & lsi < 20 & enps < 10, msymbol(Th) mcolor(black) msize(small) legend(label(3 "PR_mod_dm"))) (scatter lsi enps if PR_high_dm == 1 & lsi < 20 & enps < 10, msymbol(T) mcolor(black) msize(small) legend(label(4 "PR_high_dm"))) (scatter lsi enps if Majoritarian == 1 & lsi < 20 & enps < 10, msymbol(X) mcolor(black) msize(small) legend(label(5 "Maj")))

*Linear trade-off for pure systems as benchmark
twoway (lfit lsi enps if mmp == 0 & mmm == 0 & PR_mod_dm == 0, range (0 7.4)) (scatter lsi enps if PR_mod_dm == 1, msymbol(Th) mcolor(black) msize(small) legend(label(2 "PR_mod_dm"))) (scatter lsi enps if mmm == 1, msymbol(Sh) mcolor(black) msize(small) legend(label(3 "MMM")) graphregion(color(white))) (scatter lsi enps if mmp == 1, msymbol(C) mcolor(black) msize(small) legend(label(4 "MMP")))
twoway (lfit lsi enps if mmp == 0 & mmm == 0 & PR_mod_dm == 0 & enps < 10 & lsi < 20, range (0 7.4) ytitle("LSI")) (scatter lsi enps if PR_mod_dm == 1 & enps < 10 & lsi < 20, msymbol(Th) mcolor(black) msize(small) legend(label(2 "PR_mod_dm"))) (scatter lsi enps if mmm == 1 & enps < 10 & lsi < 20, msymbol(Sh) mcolor(black) msize(small) legend(label(3 "MMM")) graphregion(color(white))) (scatter lsi enps if mmp == 1 & enps < 10 & lsi < 20, msymbol(C) mcolor(black) msize(small) legend(label(4 "MMP")))
twoway (fpfit lsi enps if mmp == 0 & mmm == 0 & PR_mod_dm == 0 & enps < 10 & lsi < 20) (scatter lsi enps if PR_mod_dm == 1 & enps < 10 & lsi < 20, msymbol(Th) mcolor(black) msize(small) legend(label(2 "PR_mod_dm"))) (scatter lsi enps if mmm == 1 & enps < 10 & lsi < 20, msymbol(Sh) mcolor(black) msize(small) legend(label(3 "MMM")) graphregion(color(white))) (scatter lsi enps if mmp == 1 & enps < 10 & lsi < 20, msymbol(C) mcolor(black) msize(small) legend(label(4 "MMP")))

*Linear trade-off for pure systems + cross hairs as benchmarks
twoway (lfit lsi enps if mmp == 0 & mmm == 0 & PR_mod_dm == 0 & enps < 10 & lsi < 20, range (0 7.4) ytitle("LSI") yline(4.68, lpattern(dash)) xline(3.17, lpattern(dash))) (scatter lsi enps if PR_mod_dm == 1 & enps < 10 & lsi < 20, msymbol(Th) mcolor(black) msize(small) legend(label(2 "PR_mod_dm"))) (scatter lsi enps if mmm == 1 & enps < 10 & lsi < 20, msymbol(Sh) mcolor(black) msize(small) legend(label(3 "MMM")) graphregion(color(white))) (scatter lsi enps if mmp == 1 & enps < 10 & lsi < 20, msymbol(C) mcolor(black) msize(small) legend(label(4 "MMP")))
twoway (fpfit lsi enps if mmp == 0 & mmm == 0 & PR_mod_dm == 0 & enps < 10 & lsi < 20, ytitle("LSI") yline(4.68, lpattern(dash)) xline(3.17, lpattern(dash))) (scatter lsi enps if PR_mod_dm == 1 & enps < 10 & lsi < 20, msymbol(Th) mcolor(black) msize(small) legend(label(2 "PR_mod_dm"))) (scatter lsi enps if mmm == 1 & enps < 10 & lsi < 20, msymbol(Sh) mcolor(black) msize(small) legend(label(3 "MMM")) graphregion(color(white))) (scatter lsi enps if mmp == 1 & enps < 10 & lsi < 20, msymbol(C) mcolor(black) msize(small) legend(label(4 "MMP")))

*Mean/median comparisons
gen x = _n if _n <= 6
gen type = "."
replace type = "All" if _n == 1
replace type = "PR (high dm)" if _n == 2
replace type = "PR (mod dm)" if _n == 3
replace type = "MMP" if _n == 4
replace type = "MMM" if _n == 5
replace type = "Plurality/Majority" if _n == 6

gen lsi_mean = .
gen lsi_lb = .
gen lsi_ub = .

sum lsi
replace lsi_mean = r(mean) if _n == 1
replace lsi_lb = r(mean) - r(sd) if _n == 1
replace lsi_ub = r(mean) + r(sd) if _n == 1

sum lsi if PR == 1 & dm_mean >= 10
replace lsi_mean = r(mean) if _n == 2
replace lsi_lb = r(mean) - r(sd) if _n == 2
replace lsi_ub = r(mean) + r(sd) if _n == 2

sum lsi if PR == 1 & dm_mean < 10
replace lsi_mean = r(mean) if _n == 3
replace lsi_lb = r(mean) - r(sd) if _n == 3
replace lsi_ub = r(mean) + r(sd) if _n == 3

sum lsi if mmp == 1
replace lsi_mean = r(mean) if _n == 4
replace lsi_lb = r(mean) - r(sd) if _n == 4
replace lsi_ub = r(mean) + r(sd) if _n == 4

sum lsi if mmm == 1
replace lsi_mean = r(mean) if _n == 5
replace lsi_lb = r(mean) - r(sd) if _n == 5
replace lsi_ub = r(mean) + r(sd) if _n == 5

sum lsi if Majoritarian == 1
replace lsi_mean = r(mean) if _n == 6
replace lsi_lb = r(mean) - r(sd) if _n == 6
replace lsi_ub = r(mean) + r(sd) if _n == 6

twoway (scatter lsi_mean x) ///
(rcap lsi_lb lsi_ub x)

gen general_type = "."
replace general_type = "PR (high dm)" if PR == 1 & dm_mean >= 10
replace general_type = "PR (moderate dm)" if PR == 1 & dm_mean < 10
replace general_type = "MMP" if mmp == 1
replace general_type = "MMM" if mmm == 1
replace general_type = "Plurality/Majority" if Majoritarian == 1

gen order = 1 if general_type == "PR (high dm)"
replace order = 2 if general_type == "PR (moderate dm)"
replace order = 3 if general_type == "MMP"
replace order = 4 if general_type == "MMM"
replace order = 5 if general_type == "Plurality/Majority"

graph box lsi, yline(4.67805) over(general_type, sort(order))
graph save lsi_box, replace

gen enps_mean = .
gen enps_lb = .
gen enps_ub = .

sum enps
replace enps_mean = r(mean) if _n == 1
replace enps_lb = r(mean) - r(sd) if _n == 1
replace enps_ub = r(mean) + r(sd) if _n == 1

sum enps if PR == 1 & dm_mean >= 10
replace enps_mean = r(mean) if _n == 2
replace enps_lb = r(mean) - r(sd) if _n == 2
replace enps_ub = r(mean) + r(sd) if _n == 2

sum enps if PR == 1 & dm_mean < 10
replace enps_mean = r(mean) if _n == 3
replace enps_lb = r(mean) - r(sd) if _n == 3
replace enps_ub = r(mean) + r(sd) if _n == 3

sum enps if mmp == 1
replace enps_mean = r(mean) if _n == 4
replace enps_lb = r(mean) - r(sd) if _n == 4
replace enps_ub = r(mean) + r(sd) if _n == 4

sum enps if mmm == 1
replace enps_mean = r(mean) if _n == 5
replace enps_lb = r(mean) - r(sd) if _n == 5
replace enps_ub = r(mean) + r(sd) if _n == 5

sum enps if Majoritarian == 1
replace enps_mean = r(mean) if _n == 6
replace enps_lb = r(mean) - r(sd) if _n == 6
replace enps_ub = r(mean) + r(sd) if _n == 6

twoway (scatter enps_mean x) ///
(rcap enps_lb enps_ub x)

graph box enps, yline(3.168894) over(general_type, sort(order))
graph save enps_box, replace

graph combine lsi_box.gph enps_box.gph

*Four classes of performances
*good, good (lower-left quadrant)
tab PR_high_dm if good_twice_med == 1
tab PR_mod_dm if good_twice_med == 1
tab mmp if good_twice_med == 1
tab mmm if good_twice_med == 1
tab Majoritarian if good_twice_med == 1
*bad,bad (upper-right quadrant)
tab PR_high_dm if bad_twice_med == 1
tab PR_mod_dm if bad_twice_med == 1
tab mmp if bad_twice_med == 1
tab mmm if bad_twice_med == 1
tab Majoritarian if bad_twice_med == 1
*good prop, bad conc (lower-right quadrant)
tab PR_high_dm if good_lsi_med == 1 & good_enps_med == 0
tab PR_mod_dm if good_lsi_med == 1 & good_enps_med == 0
tab mmp if good_lsi_med == 1 & good_enps_med == 0
tab mmm if good_lsi_med == 1 & good_enps_med == 0
tab Majoritarian if good_lsi_med == 1 & good_enps_med == 0
*bad prop, good conc (upper-left quadrant)
tab PR_high_dm if good_lsi_med == 0 & good_enps_med == 1
tab PR_mod_dm if good_lsi_med == 0 & good_enps_med == 1
tab mmp if good_lsi_med == 0 & good_enps_med == 1
tab mmm if good_lsi_med == 0 & good_enps_med == 1
tab Majoritarian if good_lsi_med == 0 & good_enps_med == 1

*Regressions with general design types
*General performance
reg lsi Majoritarian mmm mmp PR_mod_dm PR_high_dm, nocons vce(robust)
reg enps Majoritarian mmm mmp PR_mod_dm PR_high_dm, nocons vce(robust)

logit good_twice_med mmm mmp PR_mod_dm PR_high_dm, vce(robust)

*Predicted probabilities
gen prob_bobw = .
gen prob_bobw_lb = .
gen prob_bobw_ub = .

margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw = a[1,1] if _n == 6
replace prob_bobw_lb = a[5,1]  if _n == 6
replace prob_bobw_ub = a[6,1] if _n == 6
margins, at(mmm=1 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw = a[1,1] if _n == 5
replace prob_bobw_lb = a[5,1] if _n == 5
replace prob_bobw_ub = a[6,1] if _n == 5
margins, at(mmm=0 mmp=1 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw = a[1,1] if _n == 4
replace prob_bobw_lb = a[5,1] if _n == 4
replace prob_bobw_ub = a[6,1] if _n == 4
margins, at(mmm=0 mmp=0 PR_mod_dm=1 PR_high_dm=0)
mat a = r(table)
replace prob_bobw = a[1,1] if _n == 3
replace prob_bobw_lb = a[5,1] if _n == 3
replace prob_bobw_ub = a[6,1] if _n == 3
margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=1)
mat a = r(table)
replace prob_bobw = a[1,1] if _n == 2
replace prob_bobw_lb = a[5,1] if _n == 2
replace prob_bobw_ub = a[6,1] if _n == 2

twoway (scatter prob_bobw x) ///
(rcap prob_bobw_lb prob_bobw_ub x)

*Adding context vars
reg lsi Majoritarian mmm mmp PR_mod_dm PR_high_dm polity tensys prespow1 pres_concurrent pol_dcen, nocons vce(robust)

*CONTROL FOR MANIPUL. OF MMP: reg lsi Majoritarian mmm mmp PR_mod_dm PR_high_dm polity tensys prespow1 pres_concurrent pol_dcen strategic_m, nocons vce(robust)

reg enps Majoritarian mmm mmp PR_mod_dm PR_high_dm polity tensys prespow1 pres_concurrent pol_dcen, nocons vce(robust)

logit good_twice_med PR_high_dm PR_mod_dm mmp mmm polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc, replace
lroc, nograph
predict bobw_type, xb

logit good_twice_med PR_high_dm PR_mod_dm mmp mmm polity tensys prespow1 pol_dcen ef, vce(robust)
outreg2 using myfile.doc, replace

*Predicted probabilities
gen prob_bobw2 = .
gen prob_bobw_lb2 = .
gen prob_bobw_ub2 = .

margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw2 = a[1,1] if _n == 6
replace prob_bobw_lb2 = a[5,1] if _n == 6
replace prob_bobw_ub2 = a[6,1] if _n == 6
margins, at(mmm=1 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw2 = a[1,1] if _n == 5
replace prob_bobw_lb2 = a[5,1] if _n == 5
replace prob_bobw_ub2 = a[6,1] if _n == 5
margins, at(mmm=0 mmp=1 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_bobw2 = a[1,1] if _n == 4
replace prob_bobw_lb2 = a[5,1] if _n == 4
replace prob_bobw_ub2 = a[6,1] if _n == 4
margins, at(mmm=0 mmp=0 PR_mod_dm=1 PR_high_dm=0)
mat a = r(table)
replace prob_bobw2 = a[1,1] if _n == 3
replace prob_bobw_lb2 = a[5,1] if _n == 3
replace prob_bobw_ub2 = a[6,1] if _n == 3
margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=1)
mat a = r(table)
replace prob_bobw2 = a[1,1] if _n == 2
replace prob_bobw_lb2 = a[5,1] if _n == 2
replace prob_bobw_ub2 = a[6,1] if _n == 2

gen base = 1 if _n == 2
replace base = 2 if _n == 3
replace base = 3 if _n == 4
replace base = 4 if _n == 5
replace base = 5 if _n == 6

twoway (scatter prob_bobw2 base, xlabel(1 "PR (high dm)" 2 "PR (mod dm)" 3 "MMP" 4 "MMM" 5 "Plurality/Majority") graphregion(color(white))) ///
(rcap prob_bobw_lb2 prob_bobw_ub2 base)

*WOBW
logit bad_twice_med mmm mmp PR_mod_dm PR_high_dm, vce(robust)

logit bad_twice_med PR_high_dm PR_mod_dm mmp mmm polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc, replace
lroc, nograph
predict wobw_type, xb

logit bad_twice_med mmm mmp PR_mod_dm PR_high_dm polity tensys prespow1 pol_dcen ef, vce(robust)
outreg2 using myfile.doc, replace

*Predicted probabilities
gen prob_wobw = .
gen prob_wobw_lb = .
gen prob_wobw_ub = .

margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_wobw = a[1,1] if _n == 6
replace prob_wobw_lb = a[5,1]  if _n == 6
replace prob_wobw_ub = a[6,1] if _n == 6
margins, at(mmm=1 mmp=0 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_wobw = a[1,1] if _n == 5
replace prob_wobw_lb = a[5,1] if _n == 5
replace prob_wobw_ub = a[6,1] if _n == 5
margins, at(mmm=0 mmp=1 PR_mod_dm=0 PR_high_dm=0)
mat a = r(table)
replace prob_wobw = a[1,1] if _n == 4
replace prob_wobw_lb = a[5,1] if _n == 4
replace prob_wobw_ub = a[6,1] if _n == 4
margins, at(mmm=0 mmp=0 PR_mod_dm=1 PR_high_dm=0)
mat a = r(table)
replace prob_wobw = a[1,1] if _n == 3
replace prob_wobw_lb = a[5,1] if _n == 3
replace prob_wobw_ub = a[6,1] if _n == 3
margins, at(mmm=0 mmp=0 PR_mod_dm=0 PR_high_dm=1)
mat a = r(table)
replace prob_wobw = a[1,1] if _n == 2
replace prob_wobw_lb = a[5,1] if _n == 2
replace prob_wobw_ub = a[6,1] if _n == 2

twoway (scatter prob_wobw base, xlabel(1 "PR (high dm)" 2 "PR (mod dm)" 3 "MMP" 4 "MMM" 5 "Plurality/Majority") graphregion(color(white))) ///
(rcap prob_wobw_lb prob_wobw_ub base)


*Regressions with technical details
*Technical variables only
reg lsi compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold, vce(robust)
outreg2 using myfile.doc, replace
reg enps compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold, vce(robust)
outreg2 using myfile.doc

logit good_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold, vce(robust)
outreg2 using myfile.doc, replace

*Adding context (interaction between compensXperc_smds)
reg lsi compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc
reg enps compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc

logit good_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc
lroc, nograph
predict bobw_det, xb
roccomp good_twice_med bobw_type bobw_det

logit good_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen ef, vce(robust)

*WOBW
logit bad_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold, vce(robust)
outreg2 using myfile.doc, replace

*Adding context
logit bad_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen, vce(robust)
outreg2 using myfile.doc
lroc, nograph
predict wobw_det, xb
roccomp bad_twice_med wobw_type wobw_det

logit bad_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen ef, vce(robust)

logit good_twice_med compens_pr_tier percent_smds dm_mean dm_mean2 pr_topup legal_threshold polity tensys prespow1 pol_dcen ef, vce(robust)