*LABELING MACRO VARIABLES label variable l3_hfgov "Government size, HF (t-3)" label variable l3r_hfgov "Government size, HF (t-3), l3_hfgov reversed" label variable l3_exconst "Constraints on executives (t-3)" label variable l_gdp_pc_ppp "GDP per capita ppp (t-1)" label variable l_dgdp "GDP growth rate (t-1)" label variable l3_corrfree "Freedom from corruption (t-3)" label variable l3_corr "Corruption Index (t-3), HF corrfree reversed" label variable l2_ipp "Intellectual property rights (t-2)" *GENERATING VARIABLES *Generating the dependant variable - the perceived rate of employment growth for new firms <--- gen employees_babyb = OMNOWJOB if OMNOWJOB!= . & BABYBUSO==1 replace employees_babyb = . if employees_babyb >= 999997 gen employm_babyb1= employees_babyb+1 gen ln_employm_babyb1=ln(employm_babyb1) gen BABYYR5JOB=OMYR5JOB+1 if BABYBUSO==1 gen ln_BABYYR5JOB=ln(BABYYR5JOB) gen dbabyb= ln_BABYYR5JOB- ln_employm_babyb1 *Generating individual-level explanatory variables generate byte male=GENDER replace male=0 if GENDER==2 gen knowent_dum = KNOWENT gen educ_post=GEMEDUC replace educ_post=0 if GEMEDUC>0 & GEMEDUC<1720 & GEMEDUC!= . replace educ_post=1 if GEMEDUC==1720 gen omestb_dum=ESTBBUSO gen busang_dum=busang gen age_sq=age^2 *Generating means of inidividual-level variables by country egen memploym_babyb1_c=mean(employm_babyb1), by(country) egen mage_c=mean(age), by(country) egen mage_sq_c=mean(age_sq), by(country) egen mmale_c=mean(male), by(country) egen meduc_post_c=mean(educ_post), by(country) egen momestb_dum_c=mean(omestb_dum), by(country) egen mbusang_dum_c=mean(busang_dum), by(country) egen mknowent_dum_c=mean(knowent_dum), by(country) *Generating GDP quantile dummies xtile pct = l_gdp_pc_ppp, n(5) tab pct, gen(iq) *Generating interaction terms gen l3r_hfgov_x_knowent=l3r_hfgov*knowent_dum gen l3_exconst_x_knowent=l3_exconst*knowent_dum gen l3_corr_x_knowent=l3_corr*knowent_dum gen l2_ipp_x_knowent=l2_ipp *knowent_dum *Labelling variables label variable dbabyb "New businesses' employment growth expectations" label variable employm_babyb1 "Current level of employment" label variable age "Age" label variable male "Male" label variable educ_post "Education: postsecondary" label variable omestb_dum "Owner-manager of exist bus" label variable busang_dum "Bus angel in last 3 years" label variable knowent_dum "Knows other entrepreneurs" label variable memploym_babyb1_c "Current level of employment, country mean" label variable mage_c "Age, country mean" label variable mmale_c "Male, country mean" label variable meduc_post_c "Education: postsecondary, country mean" label variable momestb_dum_c "Owner-manager of exist bus, country mean" label variable mbusang_dum_c "Bus angel in last 3 years, country mean" label variable mknowent_dum_c "Knows other entrepreneurs, country mean" *STANDARDIZING VARIABLES egen zdbabyb=std(dbabyb) egen zemploym_babyb1=std(employm_babyb1) egen zage=std(age) egen zage_sq=std(age_sq) egen zmemploym_babyb1_c=std(memploym_babyb1_c) egen zmage_c=std(mage_c) egen zmmale_c=std(mmale_c) egen zmeduc_post_c=std(meduc_post_c) egen zmomestb_dum_c=std(momestb_dum_c) egen zmbusang_dum_c=std(mbusang_dum_c) egen zmknowent_dum_c=std(mknowent_dum_c) egen zl3r_hfgov=std(l3r_hfgov) egen zl3_exconst=std(l3_exconst) egen zl2_ipp=std(l2_ipp) egen zl3_corr=std(l3_corrfree) egen zl_dgdp=std(l_dgdp) egen zl_gdp_pc_ppp=std(l_gdp_pc_ppp) egen zl3_corr_x_knowent=std(l3_corr_x_knowent) egen zl2_ipp_x_knowent=std(l2_ipp_x_knowent) egen zl3_exconst_x_knowent=std(l3_exconst_x_knowent) egen zl3r_hfgov_x_knowent=std(l3r_hfgov_x_knowent) *IDENTIFYING OUTLIERS IN THE DEPENDENT VARIABLE(keep only variable dbabyb) lv dbabyb *BASE MODEL ESTIMATION FOR CREATING INSAMPM VARIABLE TO MAKE ALL SPECIFICATIONS CONSISTENTLY RUN ON THE SAME OBS **Rev 2 version (with ipp) - just for reference here and to be able to go back to earlier results if needed xi: xtmixed dbabyb employm_babyb1 age age_sq male educ_post omestb_dum busang_dum knowent_dum memploym_babyb1_c mage_c mage_sq_c mmale_c meduc_post_c momestb_dum_c mbusang_dum_c mknowent_dum_c l3r_hfgov l3_exconst l2_ipp l3_corrfree l_dgdp l_gdp_pc_ppp i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 || country: || year:, mle gen insampm=0 replace insampm=1 if e(sample) *PAIRWISE CORRELATION MATRIX pwcorr zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmomestb_dum_c zl3r_hfgov zl3_exconst zl2_ipp zl3_corr zl_dgdp zl_gdp_pc_ppp iq1 iq2 iq3 iq4 iq5 if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 *VIF xi: reg zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmomestb_dum_c zl3r_hfgov zl3_exconst zl2_ipp zl3_corr zl3_exconst_x_knowent zl_dgdp zl_gdp_pc_ppp i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1, r cluster(country_year) estat vif *MODEL ESTIMATION * Please note that in reported results the sign of 'freedom from corruption" was reversed to facilitate interpretation set more off *Spec 1 (without country averages) xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zl3r_hfgov zl3_exconst zl3_corr zl_dgdp zl_gdp_pc_ppp i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model1 estimates stats model1 *Spec 2 (with all country averages) xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmemploym_babyb1_c zmage_c zmeduc_post_c zmomestb_dum_c zmbusang_dum_c zmknowent_dum_c zl3r_hfgov zl3_exconst zl3_corr zl_dgdp zl_gdp_pc_ppp i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model2 estimates stats model2 lrtest model2 model1 *Spec 3 (with only business ownership at a country level added) xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmomestb_dum_c zl3r_hfgov zl3_exconst zl3_corr zl_dgdp zl_gdp_pc_ppp i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model3 estimates stats model3 lrtest model3 model1 *Spec 4 (same as (3) but replacing gdp_pc with quantile dummies) xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmomestb_dum_c zl3r_hfgov zl3_exconst zl3_corr zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model4 estimates stats model4 *Spec 5 (with ipp) xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum knowent_dum zmomestb_dum_c zl3r_hfgov zl2_ipp zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model5 estimates stats model5 *Specifications 6-8 (with interaction term) *Spec 6 xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum zmomestb_dum_c zl3_exconst zl3_corr c.zl3r_hfgov##i.knowent zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model6 estimates stats model6 *Spec 7 xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum zmomestb_dum_c zl3r_hfgov zl3_exconst c.zl3_corr##i.knowent zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model7 estimates stats model7 *Spec 8 xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum zmomestb_dum_c zl3r_hfgov zl3_corr c.zl3_exconst##i.knowent zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle estimates store model8 estimates stats model8 *PRODUCING GOODNESS OF FIT (AIC +BIC) estimates stats model1 model2 model3 model4 model5 model6 model7 model8 *PRODUCING FIGURES *Figure 1 *Estimating random intercepts-countries (without outliers) and producing a caterpillar plot to show the country effects in rank order with 95% confidence intervalls xtmixed dbabyb if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 ||country:, mle variance *please note the options below 'reffects' and 'reses' are only allowed with xtmixed and cannot be run with xtreg predict u01, reffects predict u01se, reses egen pickone1=tag(country) egen u01rank = rank(u01) if pickone1==1 label define cat1 1 "Greece" 2 "Norway" 3 "Italy" 4 "France" 5 "Jordan" 6 "Hungary" 7 "Poland" 8 "Spain" 9 "Finland" 10 "Austria" 11 "South Africa" 12 "Australia" 13 "Netherlands" 14 "Switzerland" 15 "Denmark" 16 "Thailand" 17 "Germany" 18 "Latvia" 19 "Japan" 20 "Argentina" 21 "Sweden" 22 "Portugal" 23 "Mexico" 24 "Belgium" 25 "China" 26 "Russia" 27 "Israel" 28 "Canada" 29 "India" 30 "South Korea" 31 "Jamaica" 32 "United States" 33 "United Kingdom" 34 "New Zealand" 35 "Slovenia" 36 "Croatia" 37 "Venezuela" 38 "Singapore" 39 "Brazil" 40 "Peru" 41 "Ireland" 42 "Chile" label values u01rank cat1 serrbar u01 u01se u01rank if pickone1==1, scheme(sj) scale(1.96) yline(0) ytitle("Predicted random intercept") xtitle(Rank) xlabel(1 (1) 42, valuelabel labsize(2) angle(vertical) g) **Figure 2 **Scatterplots between HGA and babybusinesses' entry egen mdbaby_c=mean(dbaby), by(country) egen mbabybus_c=mean(BABYBUSO), by(country) label variable mbabybus_c "Young firm activity, country averages" gen insampm_c=0 replace insampm_c=1 if country==1 replace insampm_c=1 if country==7 replace insampm_c=1 if country==27 replace insampm_c=1 if country==30 replace insampm_c=1 if country==31 replace insampm_c=1 if country==32 replace insampm_c=1 if country==33 replace insampm_c=1 if country==34 replace insampm_c=1 if country==36 replace insampm_c=1 if country==39 replace insampm_c=1 if country==41 replace insampm_c=1 if country==43 replace insampm_c=1 if country==44 replace insampm_c=1 if country==45 replace insampm_c=1 if country==46 replace insampm_c=1 if country==47 replace insampm_c=1 if country==48 replace insampm_c=1 if country==49 replace insampm_c=1 if country==51 replace insampm_c=1 if country==52 replace insampm_c=1 if country==54 replace insampm_c=1 if country==55 replace insampm_c=1 if country==56 replace insampm_c=1 if country==61 replace insampm_c=1 if country==64 replace insampm_c=1 if country==65 replace insampm_c=1 if country==66 replace insampm_c=1 if country==81 replace insampm_c=1 if country==82 replace insampm_c=1 if country==86 replace insampm_c=1 if country==91 replace insampm_c=1 if country==101 replace insampm_c=1 if country==351 replace insampm_c=1 if country==353 replace insampm_c=1 if country==358 replace insampm_c=1 if country==371 replace insampm_c=1 if country==385 replace insampm_c=1 if country==386 replace insampm_c=1 if country==582 replace insampm_c=1 if country==876 replace insampm_c=1 if country==962 replace insampm_c=1 if country==972 gen pos=9 replace pos=12 if CTRYALP=="BR" replace pos=12 if CTRYALP=="RU" replace pos=12 if CTRYALP=="DK" replace pos=6 if CTRYALP=="SE" replace pos=6 if CTRYALP=="NO" replace pos=6 if CTRYALP=="NL" replace pos=6 if CTRYALP=="CN" replace pos=6 if CTRYALP=="SG" scatter mdbaby_c mbabybus_c if insampm_c==1, scheme(sj) mlabel(CTRYALP) mlabv(pos) legend(off) xtitle("Young firm activity rate, country means") ytitle("Young businesses' growth expectations, country means") *PLOTTING INTERACTION TERMS EFFECTS *Predictive margins of social capital by executive constraints xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum zmomestb_dum_c zl3r_hfgov zl3_corr c.zl3_exconst##i.knowent zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle margins knowent_dum, at(zl3_exconst=(-3.845274 -2.797254 -1.749234 -.7012143 .3468057)) marginsplot, scheme(sj) xtitle("Constraints on executives (t-3)") ytitle("Predicted growth aspirations") title("Predictive margins of knows other entrepreneurs with 95 % CIs") marginsplot, scheme(sj) xtitle("Constraints on executives (t-3)") ytitle("Predicted growth aspirations") title(" ") xi: xtmixed zdbabyb zemploym_babyb1 zage male educ_post omestb_dum busang_dum zmomestb_dum_c zl3r_hfgov zl3_exconst c.zl3_corr##i.knowent zl_dgdp iq2 iq3 iq4 iq5 i.OMTYPE4C if dbabyb>=-2.365372 & dbabyb<=3.15383 & insampm==1 || country: || year:, mle margins knowent_dum, at(zl3_corr=(-1.40541(0.1)2.49248)) marginsplot, scheme(sj) xtitle("Corruption (t-3)") ytitle("Predicted growth aspirations") title("Predictive margins of knows other entrepreneurs with 95 % CIs") marginsplot, scheme(sj) xtitle("Corruption (t-3)") ytitle("Predicted growth aspirations") title(" ") marginsplot, scheme(sj) xtitle("Corruption (t-3)") ytitle("Predicted growth aspirations") title(" ")xscale(reverse range(-2,2.5))