Variables of Interest-LFS (Top of page)
I used the labor force survey data from 2004 to compare credit access data from 1993. However, as this survey was different in scope compared to the earlier saldru survey, some of the variables of interest were missing. For example, the LFS data does not have any household debt data. Nevertheless, the LFS data has information on different sources from where households borrow money. Hence, I was able to make comparisons between the access to credit information between the two time periods.
The main variable of interest here in this dataset is " Q732" that asks households if they have borrowed from any particular source in the past 12 months. The "lookfor" command in STATA tells us the description of this variable.
A listing of different lending sources:
lookfor 732
storage display value
variable name type format label variable label
-------------------------------------------------------------------------------
Q732FamM byte %10.0g Family Member
Q732Neig byte %10.0g Neighbour
Q732LocD byte %10.0g Local dealer/Shop
Q732Co_O byte %10.0g Co-operative
Q732ComB byte %10.0g Commercial bank or building
society
Q732Land byte %10.0g Land bank
Q732GovA byte %10.0g Other government agency
Q732Stok byte %10.0g Stokvel
Q732NGO_ byte %10.0g Non-governmental organization
Q732MnyL byte %10.0g Moneylender/Mashonosa
Q732ComF byte %10.0g Commercial Farmer
Q732Othr byte %10.0g Other Lender
I created a variable labeled "loans" to indicate houses with loans. I created this variable by counting the number of households that answered yes for borrowings from any one or more given sources of credit. In the LFS dataset, the household IDs are represented by the variable "UqNr".
sort UqNr
gen house=1 if UqNr!=UqNr[_n-1]
The above command creates a variable to count each household level variable only once. However, we have to sort by the household identification number before we create the household qualifier.
gen loans=1 if Q732FamM==1|Q732Neig==1|Q732LocD==1|Q732Co_O==1|Q732ComB==1|Q732Land==1|Q732GovA==1|Q732Stok==1|Q732NGO_==1|
Q732MnyL==1|Q732ComF==1|Q732Othr==1
replace loans=0 if loans!=1
I use similar variables to the saldru dataset to do the other comaprisons. I created a house variable to include each household only once in the analysis. To analyze by racial groups I use the variable "popgrp".
tab popgrp if house==1
Race | Freq. Percent Cum.
------------+-----------------------------------
1 | 21,739 76.08 76.08
2 | 3,463 12.12 88.21
3 | 538 1.88 90.09
4 | 2,800 9.80 99.89
9 | 32 0.11 100.00
------------+-----------------------------------
Total | 28,572 100.00
Similar to the Saldru dataset, I created an indictor for informal loans. It takes the value 1 for informal loans and 0 for formal loans.
gen informal=1 if Q732FamM==1|Q732Neig==1|Q732LocD==1|Q732MnyL==1
replace informal=0 if Q732Co_O==1|Q732ComB==1|Q732Land==1|Q732GovA==1|Q732NGO_==1
replace informal=1 if Q732Stok==1
tab informal if house==1
informal | Freq. Percent Cum.
------------+-----------------------------------
0 | 1,693 17.18 17.18
1 | 8,161 82.82 100.00
------------+-----------------------------------
Total | 9,854 100.00
tab loans informal, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| informal
loans | 0 1 | Total
-----------+----------------------+----------
1 | 5,903 34,139 | 40,042
| 14.74 85.26 | 100.00
-----------+----------------------+----------
Total | 5,903 34,139 | 40,042
| 14.74 85.26 | 100.00
This shows that of all houeholds that borrowed, 85% borrowed from informal sources compared to only 15% from formal sources. |
Exploring Data - Descriptive Statistics and Graphs (Top of Page)
This section contains comparisons in data between both the 1993 saldru dataset and the 2004 LFS dataset. The main commands used here are "tab" and "sum". The analysis in each subsection is divided based on the dataset used.
Saldru Data
sum haveloans if house==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
haveloans | 8854 .4571945 .4981925 0 1
The above command gives the basic summary statistics of the variable haveloans. Since the variable is a binary variable the mean here denotes the proportion of households that reported having borrowed from one of the lending sources. In 1993, 45% of houses reported having a loan. The "sum" command also gives the number of observation, standard deviation, minimum and maximum observed value for the variable of interest.
sum bond_owe if house==1, det
7 :amt owed on bond or loan
-------------------------------------------------------------
Percentiles Smallest
1% -4 -4
5% -2 -4
10% -2 -4 Obs 8835
25% -2 -4 Sum of Wgt. 8835
50% -2 Mean 7604.177
Largest Std. Dev. 28842.23
75% -2 300000
90% 18000 320000 Variance 8.32e+08
95% 62000 350000 Skewness 9.255908
99% 135000 1058000 Kurtosis 222.7534
The "sum" command can also be used to give a more detailed descriptive statistics by adding the option "detail" to it. So the above command shows a detailed descriptive statistics of the variable "bond_owe", the amount of household debt. The result above shows not only the mean and the std deviation of the variable but also gives other summaries such as percetile values, the four smallest observations and the four largest observations.
LFS Data
tab loans
loans | Freq. Percent Cum.
------------+-----------------------------------
0 | 69,403 63.16 63.16
1 | 40,485 36.84 100.00
------------+-----------------------------------
Total | 109,888 100.00
About 37% of the sampled population reported having an outstanding loan in 2004 compared to about 48% in 1993. So there has been a reduction in the number of people reporting having a loan.
Due to the non availability of data on amount of debt for each household I could not comapre this between 1993 and 2004.
Saldru Data
To check if there are any differences in percentage of people borrowing money by race, we can run the following command.
tab race if house==1, sum(haveloans)
19 |
:population | Summary of haveloans
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | .42165101 .49386397 6069
02-colou | .69300912 .46159647 658
03-india | .62040816 .48627877 245
04-white | .46075085 .49866991 1172
------------+------------------------------------
Total | .45518173 .49801785 8144
The above command calculates the mean of haveloans for each racial group. Since haveloans is a binary variable, the means in this case are giving us the proportion of households that have a loan for each racial group.
To find out the mean amount owed by hosueholds in loans for different racial groups, we will follow the same procedure above but use the bond_owe variable in place of haveloans.
tab race if house==1, sum (bond_owe)
19 | Summary of 7 :amt owed on bond or
:population | loan
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 950.48185 7560.7481 6060
02-colou | 8757.0091 21560.315 658
03-india | 18525.155 29697.22 245
04-white | 34708.091 61129.303 1170
------------+------------------------------------
Total | 6967.8055 28031.545 8133
The above table shows that the amount of money owed on average by households of different racial groups differs widely. Black households owe 950 Rand compared to 34, 708 Rand by White households.
LFS Data
tab popgrp loans if house==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| loans
Race | 0 1 | Total
-----------+----------------------+----------
1 | 14,073 7,666 | 21,739
| 64.74 35.26 | 100.00
-----------+----------------------+----------
2 | 2,264 1,199 | 3,463
| 65.38 34.62 | 100.00
-----------+----------------------+----------
3 | 393 145 | 538
| 73.05 26.95 | 100.00
-----------+----------------------+----------
4 | 1,854 946 | 2,800
| 66.21 33.79 | 100.00
-----------+----------------------+----------
9 | 23 9 | 32
| 71.88 28.13 | 100.00
-----------+----------------------+----------
Total | 18,607 9,965 | 28,572
| 65.12 34.88 | 100.00
The command "tab var..., row" gives the percentages for families with loans by racial group.
The distribution of percentage of households reporting having loans is similar for each racial group in 2004. 35% of black households reported having loans while 34% of white houeholds reported having loans. This is a small reduction from 1993 when 42% of black households reported taking loans and 46% of white households reported taking loans.
tab loans popgrp if house==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| Race
loans | 1 2 3 4 9 | Total
-----------+-------------------------------------------------------+----------
0 | 14,073 2,264 393 1,854 23 | 18,607
| 75.63 12.17 2.11 9.96 0.12 | 100.00
-----------+-------------------------------------------------------+----------
1 | 7,666 1,199 145 946 9 | 9,965
| 76.93 12.03 1.46 9.49 0.09 | 100.00
-----------+-------------------------------------------------------+----------
Total | 21,739 3,463 538 2,800 32 | 28,572
| 76.08 12.12 1.88 9.80 0.11 | 100.00
This table above gives a breakdown by racial group for families that have loans and that do not have loans. As we can see from the above table, of all households with loans 76% of the loans were borrowed by black families compared to only 9% by white families. So in other words it gives the conditional probability of belonging to a particual racial group given a family has a loan.
Acess to credit from different sources (Top of Page):
To understand sources of borrowings better, we have to analyze borrowings from different sources by racial groups. We first should try to find the popular sources of borrowing. To do this we create indicator variables for each source as the variable "owe" tells us how much each house owes to the different sources.
Saldru Data
Coding for creating indicator variables for the different loan sources: The above code creates binary variables for the four common sources of lending in South Africa in 1993, i.e. banks, family members/friends, moneylenders, stokvels.
gen bank=1 if owe4>0 & owe4!=.
(42582 missing values generated)
replace bank=0 if bank!=1
(42582 real changes made)
gen mlender=1 if owe6>0 & owe6!=.
(43607 missing values generated)
replace mlender=0 if mlender!=1
(43607 real changes made)
gen friend=1 if owe1>0 & owe1!=.
(40256 missing values generated)
replace friend=0 if friend!=1
(40256 real changes made)
gen stokvel=1 if owe7>0 & owe7!=.
(43706 missing values generated)
replace stokvel=0 if stokvel!=1
(43706 real changes made) |
Once we have the indicator variables for the different sources, we can graph them to compare the borrowing from different sources by race.
The graph above gives a clear indication that acess to formal sources of credit is available only to white households. Only 1.6% of black households reported having borrowed from a bank or a formal financial institution. Of the four sources, borrowing from family/friends was the most common form of credit. 20.5% of blackhouseholds reported having borrowed from a family member/ friend.
Family members
To find out more about individual borrowing patterns, we can use tab commands and pie charts to illustrate the differences in racial groups. The table shows that the amount owed to family members and friends by households of different races. Blacks and Coloured households seem to borrow a much smaller amount in comparison to Indian and White houses.
tab race if house==1, sum(owe1)
19 |
:population | Summary of 1 owe
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 362.68251 2139.0795 526
02-colou | 800.29412 1610.3835 51
03-india | 10772 19797.549 25
04-white | 8408.8636 13464.191 66
------------+------------------------------------
Total | 1580.6452 6675.4485 668
The pie chart above tells us that of all people borrowing from friends and family members, 79.2% were black households. So black households seem to rely heavily on informal credit acess through friends and family whereas only a small percentage of the other racial groups borrow money from friends or family. This clearly shows that the black houses depended heavily on informal credit.
Banks
On the other hand we can look at the percentage of houses by race that owed money to banks. The following table shows that the amount owed to banks for all racial groups is high compared to money owed by households using family and friends to borrow money from.
tab race if house==1, sum(owe4) (*money owed to banks)
19 |
:population | Summary of 4 owe
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 16907.773 21073.067 44
02-colou | 12023.786 17126.321 42
03-india | 8956.1111 10193.728 18
04-white | 29516.921 58553.471 229
------------+------------------------------------
Total | 24533.111 50113.538 333
The pie chart above gives us information on the percentage of households borrowing from banks by race given that a household owed money to banks. So we can see that it was predominantly White households that borrowed from banks compared to other races in 1993.
We next move on to find the association between races and formal and informal loans.The table below tells us that of all households who had an informal loan, 74.22% were balck households, 11.99% were colored households, 4.36% were white households and 9.43% were white households.
tab race informal if house==1, col
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
19 |
:populatio | informal
n group | Formal lo Informal | Total
-----------+----------------------+----------
01-afric | 74 2,297 | 2,371
| 24.50 74.22 | 69.80
-----------+----------------------+----------
02-colou | 42 371 | 413
| 13.91 11.99 | 12.16
-----------+----------------------+----------
03-india | 13 135 | 148
| 4.30 4.36 | 4.36
-----------+----------------------+----------
04-white | 173 292 | 465
| 57.28 9.43 | 13.69
-----------+----------------------+----------
Total | 302 3,095 | 3,397
| 100.00 100.00 | 100.00
LFS Data
Banks
tab loans Q732ComB if house==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| Commercial bank or building
| society
loans | 1 2 9 | Total
-----------+---------------------------------+----------
0 | 0 18,598 30 | 18,628
| 0.00 99.84 0.16 | 100.00
-----------+---------------------------------+----------
1 | 1,356 8,608 2 | 9,966
| 13.61 86.37 0.02 | 100.00
-----------+---------------------------------+----------
Total | 1,356 27,206 32 | 28,594
| 4.74 95.15 0.11 | 100.00
This table above shows that of all families reporting borrowing money, only 13.6% were able to get loans from commerical banks and building societies.
tab popgrp Q732ComB if house==1 & loans==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| Commercial bank or building
| society
Race | 1 2 9 | Total
-----------+---------------------------------+----------
1 | 456 7,208 2 | 7,666
| 5.95 94.03 0.03 | 100.00
-----------+---------------------------------+----------
2 | 231 968 0 | 1,199
| 19.27 80.73 0.00 | 100.00
-----------+---------------------------------+----------
3 | 63 82 0 | 145
| 43.45 56.55 0.00 | 100.00
-----------+---------------------------------+----------
4 | 603 343 0 | 946
| 63.74 36.26 0.00 | 100.00
-----------+---------------------------------+----------
9 | 3 6 0 | 9
| 33.33 66.67 0.00 | 100.00
-----------+---------------------------------+----------
Total | 1,356 8,607 2 | 9,965
| 13.61 86.37 0.02 | 100.00
When we further break it apart by race, we find that of all households having access to credit from banks, only 6% were black households compared to 63% white houses.
graph pie famloan if house==1 & famloan==100, over(popgrp) angle(90) pie( 1, explode color(emerald)) plabel(1 percent, size(medsmall) orientation(horizontal) color(white) format(%9.2f)) plabel(2 percent, size(medsmall) orientation(vertical) color(white) format(%9.2f)) plabel(4 percent, size(medsmall) orientation(vertical) color(white) format(%9.2f)) title(Percentage of all loans from family members by racial groups, size(medsmall)) note(LFS 2004) legend(order(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others"))
The above bar chart gives a graphical description of percetage of families by race borrowing from banks in 2004.
Family members
tab popgrp Q732FamM if house==1 & loans==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| Family Member
Race | 1 2 9 | Total
-----------+---------------------------------+----------
1 | 2,909 4,756 1 | 7,666
| 37.95 62.04 0.01 | 100.00
-----------+---------------------------------+----------
2 | 301 898 0 | 1,199
| 25.10 74.90 0.00 | 100.00
-----------+---------------------------------+----------
3 | 47 98 0 | 145
| 32.41 67.59 0.00 | 100.00
-----------+---------------------------------+----------
4 | 132 814 0 | 946
| 13.95 86.05 0.00 | 100.00
-----------+---------------------------------+----------
9 | 4 5 0 | 9
| 44.44 55.56 0.00 | 100.00
-----------+---------------------------------+----------
Total | 3,393 6,571 1 | 9,965
| 34.05 65.94 0.01 | 100.00
We now look at borrowings from family members and friends by race. As it is evident from the above table, more black and coloured households tend to borrow from famliy members than white households. Approximately 38% of black households and 25% coloured households borrow from family members while only 14% of white families borrow from family members or friends.
A graphhical representation of the table is given below:
graph bar (mean) famloan if house==1 & loans==1, over(popgrp, relabel(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others")) bar(1, bfcolor(bluishgray) blcolor(black)) bar(2, bfcolor(bluishgray) blcolor(black)) bar(3, bfcolor(bluishgray) blcolor(black)) bar(4, bfcolor(bluishgray) blcolor(black)) bar(5, bfcolor(bluishgray) blcolor(black)) blabel(bar, color(black) format(%9.2f)) ytitle(Percentage borrowing loans, margin(medsmall)) title(Percentage of households borrowing loans from family by race, size(medium)) note(LFS 2004)
We can analyze this further by calculating what precetage of each racial group borrows from family members. So we calculate the probability of being in each racial group given a house has a loan from a family member or friend.
tab popgrp Q732FamM if house==1 & loans==1, col
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
| Family Member
Race | 1 2 9 | Total
-----------+---------------------------------+----------
1 | 2,909 4,756 1 | 7,666
| 85.74 72.38 100.00 | 76.93
-----------+---------------------------------+----------
2 | 301 898 0 | 1,199
| 8.87 13.67 0.00 | 12.03
-----------+---------------------------------+----------
3 | 47 98 0 | 145
| 1.39 1.49 0.00 | 1.46
-----------+---------------------------------+----------
4 | 132 814 0 | 946
| 3.89 12.39 0.00 | 9.49
-----------+---------------------------------+----------
9 | 4 5 0 | 9
| 0.12 0.08 0.00 | 0.09
-----------+---------------------------------+----------
Total | 3,393 6,571 1 | 9,965
| 100.00 100.00 100.00 | 100.00
So we can see from the above table that 85% of family member loans are given to black households. So black families depend more on borrowing from family and friends as they have limited access to other forms of credit. In comaprison, only 4% of white households were taking money from their family members or friends.
The following pie chart shows this in graphical form:
graph pie famloan if house==1 & famloan==100, over(popgrp) angle(90) pie( 1, explode color(emerald)) plabel(1 percent, size(medsmall) orientation(horizontal) color(white) format(%9.2f)) plabel(2 percent, size(medsmall) orientation(vertical) color(white) format(%9.2f)) plabel(4 percent, size(medsmall) orientation(vertical) color(white) format(%9.2f)) title(Percentage of all loans from family members by racial groups, size(medsmall)) note(LFS 2004) legend(order(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others"))
Money lenders
tab popgrp Q732MnyL if house==1 & loans==1, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
| Moneylender/Mashonosa
Race | 1 2 9 | Total
-----------+---------------------------------+----------
1 | 798 6,867 1 | 7,666
| 10.41 89.58 0.01 | 100.00
-----------+---------------------------------+----------
2 | 98 1,101 0 | 1,199
| 8.17 91.83 0.00 | 100.00
-----------+---------------------------------+----------
3 | 2 143 0 | 145
| 1.38 98.62 0.00 | 100.00
-----------+---------------------------------+----------
4 | 5 941 0 | 946
| 0.53 99.47 0.00 | 100.00
-----------+---------------------------------+----------
9 | 0 9 0 | 9
| 0.00 100.00 0.00 | 100.00
-----------+---------------------------------+----------
Total | 903 9,061 1 | 9,965
| 9.06 90.93 0.01 | 100.00
The above table shows that the amount of borrowing from money lenders is relatively low compared to the borrowing from family members or banks for most racial groups. only 10.41% of black households reported borrowing money from money lenders whereas a minute percentage of white families borrowed from money lenders.
graph bar (mean) mlenderloan if house==1 & loans==1, over(popgrp, relabel(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others")) bar(1, bfcolor(bluishgray) blcolor(black)) bar(2, bfcolor(bluishgray) blcolor(black)) bar(3, bfcolor(bluishgray) blcolor(black)) bar(4, bfcolor(bluishgray) blcolor(black)) bar(5, bfcolor(bluishgray) blcolor(black)) blabel(bar, color(black) format(%9.2f)) ytitle(Percentage borrowing loans, margin(medsmall)) title(Percentage of households borrowing loans from moneylender by race, size(medium)) note(LFS 2004)
Looking at the percentage of households borrowing from money lenders by race given a house has a loan from a moneylender, we see that 88% of all moneylender loans go to black households. In this case, we basically shrink our sample size. We only look at households that have borrowed from a money lender.
tab popgrp Q732MnyL if house==1 & loans==1, col
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
| Moneylender/Mashonosa
Race | 1 2 9 | Total
-----------+---------------------------------+----------
1 | 798 6,867 1 | 7,666
| 88.37 75.79 100.00 | 76.93
-----------+---------------------------------+----------
2 | 98 1,101 0 | 1,199
| 10.85 12.15 0.00 | 12.03
-----------+---------------------------------+----------
3 | 2 143 0 | 145
| 0.22 1.58 0.00 | 1.46
-----------+---------------------------------+----------
4 | 5 941 0 | 946
| 0.55 10.39 0.00 | 9.49
-----------+---------------------------------+----------
9 | 0 9 0 | 9
| 0.00 0.10 0.00 | 0.09
-----------+---------------------------------+----------
Total | 903 9,061 1 | 9,965
| 100.00 100.00 100.00 | 100.00
graph pie mlenderloan if house==1 & mlenderloan==100, over(popgrp) angle(90) plabel(1 percent, color(white)format(%9.2f)) plabel(2 percent, color(white)) title(Percentage of all loans borrowed from moneylenders byrace, size(medium)) note(LFS 2004) legend(order(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites"))
I then created a series of indicator variables for borrowings from different sources and compared it by race.
recode Q732ComB (1=1) (2=0), gen(bankloan)
(27206 differences between Q732ComB and bankloan)
recode Q732FamM (1=1) (2=0), gen(famloan)
(25171 differences between Q732FamM and famloan)
recode Q732MnyL (1=1) (2=0), gen(mlenderloan)
(27657 differences between Q732MnyL and mlenderloan)
recode Q732Stok (1=1) (2=0), gen(stokvelloan)
(28244 differences between Q732Stok and stokvelloan)
In this case, I only coded the four prominent forms of borrowing in South Africa. Instead of having the values 1 and 2 for a "yes" and "no" answer, I coded a "yes" as 1 and a "no" as 0.
graph bar (mean) bankloan (mean) famloan (mean) mlenderloan (mean) stokvelloan if house==1 & loans==1, over(popgrp, relabel(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others")) bar(1, blcolor(black)) bar(2, blcolor(black)) bar(3, blcolor(black)) bar(4, blcolor(black)) bar(5, blcolor(black)) blabel(bar, color(black) format(%9.1f)) ytitle(Percentage borrowing loans, margin(medsmall)) title(Percentage of households borrowing loans from different sources by race, size(medium)) note(LFS 2004) legend(order(1 "Bank" 2 "Family" 3 "Moneylender" 4 "Stokvel"))
Above graph compares borrowing reported from different credit sources by race. Family member borrowing is most common form of credit access to most families. 38% of balck families said they borrowed from a family member compared to 14% for whites. The percentage of families getting bank loans by race were also different with only 6% of black households receiving loans from banks compared to 63.7% of white households.
I also divided the lending sources into two categories - formal and informal sources. I then was able to compare the difference in credit access by race for formal and informal loans.
gen informal=1 if Q732FamM==1|Q732Neig==1|Q732LocD==1|Q732MnyL==1
replace informal=0 if Q732Co_O==1|Q732ComB==1|Q732Land==1|Q732GovA==1|Q732NGO_==1
replace informal=1 if Q732Stok==1
label def informal 0 "Formal" 1 "Informal"
label val informal informal
tab informal if house==1
informal | Freq. Percent Cum.
------------+-----------------------------------
0 | 1,693 17.18 17.18
1 | 8,161 82.82 100.00
------------+-----------------------------------
Total | 9,854 100.00
Of all house that reported having a loan, 83% borrowed from informal sources compared to only 17% who borrowed from formal sources.
graph pie loans if house==1 & loans==1, over(informal) angle(90) plabel(_all percent, color(white) format(%9.1f)) title(Percentage of households with fromal and informal loans, size(medlarge)) note(LFS 2004) legend(order(1 "Formal loan" 2 "Informal loan"))
Dividing the formal and informal lending by racial groups, shows us that more black households borrow from informal sources compared to the other racial groups. 84% of black households borrow from informal sources compared to 3% of whites.
tab popgrp informal if house==1, col
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
| informal
Race | 0 1 | Total
-----------+----------------------+----------
1 | 702 6,881 | 7,583
| 41.46 84.33 | 76.96
-----------+----------------------+----------
2 | 256 920 | 1,176
| 15.12 11.27 | 11.94
-----------+----------------------+----------
3 | 73 71 | 144
| 4.31 0.87 | 1.46
-----------+----------------------+----------
4 | 658 283 | 941
| 38.87 3.47 | 9.55
-----------+----------------------+----------
9 | 4 5 | 9
| 0.24 0.06 | 0.09
-----------+----------------------+----------
Total | 1,693 8,160 | 9,853
| 100.00 100.00 | 100.00
graph pie if house==1 & loans==1, over(popgrp) angle(90) by(informal, title(Percentage borrowing from formal and infornal sources by race, size(medsmall)) note(LFS 2004)) plabel(1 percent, color(white) format(%9.2f)) plabel(2 percent, color(white) format(%9.2f)) plabel(4 percent, orientation(vertical) color(white) format(%9.2f)) legend(order(1 "Africans" 2 "Coloured" 3 "Indians" 4 "Whites" 5 "Others"))
Comparing the data from 1993 and 2004, we find that the percentage of balck households that have access to formal credit has increased in the 11 years from 24.5% to 41.5%. However, on the other hand the percentage of balck households borrowing money from informal sources has also gone up in this time period from 74% to 84%.
2. Income and household debt (Top of page)
Saldru Data
Let us now look at the relationship between household total monthly income and the debt they have. The total monthly household income is coded as "totminc ".
First let us find out what the mean total monthly income of households in debt is:
sum totminc if house==1 & haveloans==1,d
total monthy income
-------------------------------------------------------------
Percentiles Smallest
1% 0 -4833.333
5% 118.4833 0
10% 259.25 0 Obs 3921
25% 524.1642 0 Sum of Wgt. 3921
50% 1132.457 Mean 2256.977
Largest Std. Dev. 4787.539
75% 2537 39700
90% 5616.667 45640 Variance 2.29e+07
95% 7711.667 107863.1 Skewness 23.87753
99% 13702.5 207146.5 Kurtosis 922.6451
tab race if house==1 & haveloans==1, sum(totminc)
19 |
:population | Summary of total monthy income
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 1117.9766 1295.604 2513
02-colou | 2039.6831 1599.2119 426
03-india | 3422.5004 2501.7789 146
04-white | 6741.7541 11207.721 506
------------+------------------------------------
Total | 2113.4483 4814.8908 3591
The above table gives the total monthly income of hosueholds that have loans. The table shows us that the mean monthly household income is much smaller for blacks compared to White households. However, it would be better to divide income into quartiles and then check the source of credit for different income quartiles by race. The command we use for recoding of a variable is "recode".
recode totminc (0/430=1 "1st quartile") (430/907=2 "2nd quartile") (907/2025=3 "3rd quartile") (2025/max=4 "4th quartile") if totminc!=. & totminc>=0, gen(incomequartile)
The above command recodes the total monthly household income variable and puts them into 4 categories. Households with total income between 0-430 Rand are classified as 1st quartile, 430-907 are classified as 2nd quartile and so on. This command categorizes the variables and then stores them under a new variable named "incomequartile"
tab race incomequartile if house==1, sum(owe1)
Means, Standard Deviations and Frequencies of 1 owe
19 |
:populatio | RECODE of totminc (total monthy income)
n group | 1st quart 2nd quart 3rd quart 4th quart | Total
-----------+--------------------------------------------+----------
01-afric | 232.03349 210.89759 319.73394 1711.1613 | 332.81748
| 485.18928 333.81776 412.70368 8042.7096 | 2016.3755
| 209 166 109 31 | 515
-----------+--------------------------------------------+----------
02-colou | 188.375 290.28571 688.26087 1333 | 681.22917
| 349.13073 158.01025 1052.2699 2196.4533 | 1271.9978
| 8 7 23 10 | 48
-----------+--------------------------------------------+----------
03-india | 5000 10000 2870 18844.444 | 10150
| 0 0 4340.0077 29611.278 | 20025.026
| 1 2 10 9 | 22
-----------+--------------------------------------------+----------
04-white | 6550 1943.3333 2610.6 11644.769 | 8612.0492
| 4634.2925 2653.6076 3911.0168 16331.53 | 13830.737
| 4 3 15 39 | 61
-----------+--------------------------------------------+----------
Total | 365.77477 353.20787 755.03185 7754.1798 | 1474.822
| 1150.8269 1139.2237 1866.6351 15945.777 | 6533.0591
| 222 178 157 89 | 646
The above table is a cross tab of race and income quartiles with the amount owed by households to family/friends. It is clear that black households borrow more from family members as the observations for black households is much larger when compared to the other racial groups.
Similarly, we can graph the proportion of families that have access to credit from bank and compare by income quartiles. As we would expect, a higher proportion of families from the upper quartile report owing money to banks compared to the lower quartiles.
The above graph once again illustrates that more rich white and Indian households have access to bank credit than black households. Only 4% of black houses who belong to the upper quartile said they owed money to banks compared to 20% of white households. However, this graph also shows that black and colored houses belonging to both the first and second income quartiles do not have access to bank credit.
We can also classify the households into poor and nonpoor. I classified households with total monthly income less than half of the median income as poor. We must first find the median income total monthly household income. We can use the "codebook" command to find the median and then generate a new variable to categorize households as poor or nonpoor.
codebook totminc if house==1
-----------------------------------------------------------------------------------------------------------
totminc total monthy income
-----------------------------------------------------------------------------------------------------------
type: numeric (float)
&nbs
