Do file to check consistency of figures/statistics presented in Dupas & Robinson (2013) against what appears in the data

gistfile1.txt

***************************************************************************************************
/*  Do file to check consistency of figures/statistics presented in Dupas & Robinson (2013) against 
	what appears in the data */
***************************************************************************************************

clear 
set mem 500m

* set directory
gl DRIVE	"/Users/katiacovarrubias/Documents/Academic/IHEID/Courses/Advanced Econometrics/Exam/Dupas_data/"

* log file
capture log close
log using "$DRIVE/dupas_consistencycheck.log",replace

* load data and run variable creation from their do file
use "$DRIVE/dataset_savingsAEJ.dta",clear

 
*********************************************************
*Variable creation from dupas do file
  * Active if had more than one bank transaction in first 6 months
  gen active=(first6_num_trans_savings>1) if first6_num_trans_savings!=.
  * Treatment interacted with works as boda-boda dummy
  gen treatment_bg_boda=treatment*bg_boda
  * Active bank account interacted with works as boda-boda dummy
  gen active_bg_boda=active*bg_boda
  * Works as boda-boda and sampled at wave 2
  gen bg_boda_wave2=bg_boda & wave2
  * Works as male vendor and sampled at wave 2
  gen bg_malevendor_wave2=bg_malevendor  & wave2
  * Works as male vendor and sampled at wave 3
  gen bg_malevendor_wave3=bg_malevendor  & wave3
  * Treatment interacted with works as male vendor
  gen treatment_bg_malevendor=treatment*bg_malevendor
  * Active bank account interacted with works as male vendor
  gen active_bg_malevendor=active*bg_malevendor
  * literacy var
  gen literate_swahili=1 if bg_kis_read==1 & bg_kis_write==1
  replace literate_swahili=0 if bg_kis_read==0 | bg_kis_write==0

 
*****************************************************************************
* PAGE 166
	sum investment
		* 353 Ksh
	sum investment if investment!=0 
		* 365 Ksh
	sum investment if !(wave1==. & wave2==. & wave3==.)
		* 353 Ksh
	sum investment if !(wave1==. & wave2==. & wave3==.) & investment !=0
		* 365 Ksh


	tab bg_loan_bank if !(wave1==. & wave2==. & wave3==.)
		* 2.8 percent
	tab bg_loan_bank if !(wave1==. & wave2==. & wave3==.) & not_trac==0
		* 3.26 percent if only consider individuals that were able to be tracked down.

*****************************************************************************
* page 167
	tab per_hard_save if !(wave1==. & wave2==. & wave3==.)
		* 85.62% (125 obs)
	tab per_hard_save 
		* 85.71% (126)
	
*****************************************************************************
* PAGE 168
* explore number of observations in male bicycle taxi drivers	 (92 OR 96?)

*****************************************************************************
* PAGE 169
	g illiterate=literate==0
	table bg_gender if inlog==1 , c(mean illiterate)
		* 7.5% males; 34.1% females
		
	sum exp_total if inlog==1
	display 50/r(mean)
		* 28.97%

*****************************************************************************	
* PAGE 170
	g logrefuse=inlog==0
	table  not_traced_acco, c(mean inlog mean logref) row
		* 28.9% overall

	count if per_hard_save!=. & per_invest_choice2!=. & per_somewhat_patient!=. & per_time_consistent!=. & per_hyperbolic!=. & per_pat_now_impat_later!=. & per_maximpat!=. 
		* 143 obs
		
*****************************************************************************
* PAGE 173
	tab num_dep_sav if (treatment==1 & not_traced_account_opening==0) ,m
		* 10 missing obs (6.4%), 72 zeroes (46%)
		
*****************************************************************************	
* PAGE 174
	* median deposits for both genders
	sum first6_dep_savings if treatment==1 &  wave1!=. & wave2!=. & wave3!=. , det
		display r(p50) /*0*/
			
	*75th and 90th percentile of total deposits- males	
	sum first6_dep_savings if treatment==1 & bg_gender==1  & wave1!=. & wave2!=. & wave3!=. , det
		display r(p75) /*300*/
		display r(p90) /*1200*/
			
	*75th and 90th percentile of total deposits- females	
	sum first6_dep_savings  if treatment==1 & bg_gender==0 & wave1!=. & wave2!=. & wave3!=. , det
		display r(p75) /*500*/
		display r(p90) /*5000*/

	* Mean deposits across gender
	table bg_gender if treatment==1 & wave1!=. & wave2!=. & wave3!=. , c(mean first6_dep_savings)


*****************************************************************************	
* PAGE 176
	tab active if treatment==1
	tab active if treatment==1 & filled_log  !=. & filled_log  !=0 

*****************************************************************************
* PAGE 187
	* check median of average deposit size
		* From D&R do file: Average deposit size is total of all deposits ever made/total number of deposits ever made
			gen size_deposit=total_dep_savings/num_dep_savings
			sum size_deposit, detail
			sum size_deposit if active==1, detail
 				display r(p50)
 				* median= 300 Ksh
 		* for active women:	
			sum size_deposit if active==1 & bg_gender==0, detail
 				display r(p50)
 				* median=588.8889 
		* this median should be equivalent to 1.6 days of mean expenditures for women in the sample
			sum exp_total if active==1 & bg_gender==0, detail
				display 588.8889/r(mean)	/*2.41 times average daily expenditure for active women*/
				display 300/r(mean)			/*1.23 times average daily expenditure for active women*/
			sum exp_total if bg_gender==0, detail
				display 588.8889/r(mean)	/*3.25 times average daily expenditure for women*/
				display 300/r(mean)			/*1.65 times average daily expenditure for women*/

*****************************************************************************
* EXTRA: Check trimming
	count if filled_log!=0 & filled_log!=.
	
	list investment* bg_boda bg_malevendor bg_femalevendor if filled_log!=0 & filled_log!=. & (investment!=investment_t5), separator(0)
	count if filled_log!=0 & filled_log!=. & (investment!=investment_t5)
	 	* 36 obs (14%)
	list revenues* bg_boda bg_malevendor bg_femalevendor if filled_log!=0 & filled_log!=. & (revenues!=revenues_t5), separator(0)
	count if filled_log!=0 & filled_log!=. & (revenues!=revenues_t5)
		* 46 obs (18.4%)
		
		* it seems that trimming was not on the overall sample, but for each individual's records



*****************************************************************************
log close