DATA FROM SALDRU SURVEY (1994)
This survey contains variables at a household and individual-based level. Before generating any variables, it is important to identify the level of each variable. While household variables are the same for each member of the household, independent variables vary by each respondent. After opening the dataset, we will sort and separate the variables according to household. Therefore, we will create a variable equal to 1 when household id does not equal the previous household id in the dataset (Thus, two consecutive household ids are different).
Sort by household id number:
sort hhid
Generate a variable which separates respondents into households:
gen household =.
replace household= 1 if hhid!=hhid[_n-1]
Remittances Received
The variable for remittances is totm_rec. This is a continuous and household level variable, which shows the amount of remittances received by each household. Using the following command, we will generate a categorical variable labeled remittances, which sets remittances equal to a value of 1 when remittances received are greater than 0. By setting a 0 for remittances when they are equal or less than 0, this variable separates those households which receive remittances out from those who do not. (We will also add household==1 to make sure that this new variable only generates household observations.)
gen remittances =.
replace remittances=1 if totm_rec>0 & household==1
replace remittances=0 if totm_rec<=0 & household==1
In order to display the % of households which receive remittances, we can use the tab command. This shows that nearly 27% of households receive remittances.
tab remittances if household==1
remittances | Freq. Percent Cum.
------------+-----------------------------------
0 | 6,492 73.32 73.32
1 | 2,362 26.68 100.00
------------+-----------------------------------
Total | 8,854 100.00
By using the sum command, we can generate the mean amount of remittances received:
sum totm_rec if household==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
totm_rec | 8848 65.39906 183.1304 0 5301.638
I. Race
One interesting question is whether the amount of remittances vary according to different racial groups. Therefore, it would be important to examine the mean amount of remittances for each race, which can be shown by combining the tab and sum commands.
tab race if household==1, sum (totm_rec)
19 |
:population | Summary of total monthly received
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 78.010302 183.00652 6069
02-colou | 53.170313 203.13704 658
03-india | 41.84966 181.20345 245
04-white | 40.522708 202.3724 1172
------------+------------------------------------
Total | 69.520674 188.09349 8144
If we would like to show this visually, we can graph this data into a bar chart. The following graph shows that Black Africans receive the largest amount of remittances out of all three races.
graph bar (mean) totm_rec if household==1, over(race) blabel(bar, format(%9.1f)) ytitle(Average amount of remittances ) title(Average Amount of Remittances ) subtitle(By Households)
In order to display the % of households which solely receive remittances by race, we can use the cross tab command between race and remittances. By inserting a , row we can display the % each race makes up out of all the racial groups that receive remittances. For example, Black Africans make up nearly 90% of the households that receive remittances while Whites make up less than 5%.
tab race remittances, row
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
19 |
:populatio | remittances
n group | 0 1 | Total
-----------+----------------------+----------
01-afric | 4,005 2,064 | 6,069
| 68.60 89.51 | 74.52
-----------+----------------------+----------
02-colou | 538 120 | 658
| 9.22 5.20 | 8.08
-----------+---------------------+----------
03-india | 221 24 | 245
| 3.79 1.04 | 3.01
-----------+----------------------+----------
04-white | 1,074 98 | 1,172
| 18.40 4.25 | 14.39
-----------+----------------------+----------
Total | 5,838 2,306 | 8,144
| 100.00 100.00 | 100.00
Display a pie chart showing remittances received according to race:
graph pie remittances if household==1, over(race) angle(90) plabel(1 percent, color(white) format(%9.1f)) plabel(2 percent, color(white) format(%9.1f)) plabel(4 percent, color(white) format(%9.1f)) title(Remittances by Race)
The above graph shows that Black Africans receive the largest % of remittances followed by Coloureds, Whites, and then Indians. In order to compare the significance between White and Black Africans, we can perform a t-test to see if there is a significant difference between the amount of remittances received between the two groups. If we would like to restrict race only to Whites and Black Africans, we can generate a new variable called newrace.
gen newrace= 1 if race==1
replace newrace=0 if race==4
ttest totm_rec if household==1, by (newrace)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 1172 40.52271 5.911361 202.3724 28.92467 52.12075
1 | 6069 78.0103 2.349135 183.0065 73.40516 82.61544
---------+--------------------------------------------------------------------
combined | 7241 71.94271 2.194912 186.774 67.64004 76.24537
---------+--------------------------------------------------------------------
diff | -37.48759 5.943376 -49.13835 -25.83684
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -6.3075
Ho: diff = 0 degrees of freedom = 7239
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
This t-test is designed to see whether or not there is a difference between the two selected sample groups. As the value of the t-test statistic grows, the probability decreases that differences between the two means are due to chance. When the value of t exceeds the absolute value of 1.96, the probability that the two groups are statsically similar drops to 0 at the 95% confidence interval. When looking at the above t-test between White and Black Africans, we see that it is statisitcally significant (-6.3075) and there is a significant difference between the average amount of remittances between the two groups.
II. Income Quartile
Next, we can look at remittances according to income quartiles to examine whether or not households receiving remittances are poorer or perhaps, wealthier than the typical household. In order to separate remittances according to income groups, we will have to generate a new variable. First, we will look at the household variable for income, which is totminc. In order to display the % values of income, we can use the codebook command.
codebook totminc if household==1
totminc total monthly income
type: numeric (float)
range: [-4833.3335,207146.5] units: 1.000e-07
unique values: 5858 missing .: 288/8854
mean: 1956.97
std. dev: 4413.58
percentiles: 10% 25% 50% 75% 90%
204.083 429.355 907 2025 4716.67
Using the percentiles displayed above, we can generate a variable called incquart. The best way to create this variable with the proper labels is to use the recode command. However, it is important to remember that we will want to exclude totminc variables that are equal to missing or negative. Thus, we will want to use the != command in order to exclude missing or negative values.
recode totminc (0/429.555= 1 "first quartile") (429.555/907= 2 "second quartile") (907/2025=3 "third quartile") (2025/max=4 "fourth quartile") if totminc!=. & totminc>0, gen (incquart)
This variable can also be displayed
Now we can display the number of households per each income quartile according to our recode command displayed above.
tab incquart if household==1
RECODE of |
totminc (total |
monthy income) | Freq. Percent Cum.
----------------+-----------------------------------
first quartile | 1,997 23.71 23.71
second quartile | 2,151 25.54 49.26
third quartile | 2,133 25.33 74.59
fourth quartile | 2,140 25.41 100.00
----------------+-----------------------------------
Total | 8,421 100.00
A faster and quicker way to separate by percentiles would be to utilize the xtile command. By displaying the following command, we can quickly generate incquart. Since we classified the quartiles according to the percentiles listed in codebook, the results may be slightly different, but can still be used to accurately report income quartiles.
xtile incquart=totminc, nq(4)
tab incquart if household==1
4 quantiles |
of totminc | Freq. Percent Cum.
------------+-----------------------------------
1 | 2,152 25.12 25.12
2 | 2,055 23.99 49.11
3 | 2,116 24.70 73.82
4 | 2,243 26.18 100.00
------------+-----------------------------------
Total | 8,566 100.00
In order to display the average mean of remittances received per household, we can use the following sum command.
tab incquart if household==1, sum (totm_rec)
RECODE of |
totminc |
(total |
monthy | Summary of total monthly received
income) | Mean Std. Dev. Freq.
------------+------------------------------------
first qua | 71.667794 107.63507 1997
second qu | 84.29892 166.73681 2151
third qua | 69.241219 204.75923 2133
fourth qu | 43.306644 230.09036 2140
------------+------------------------------------
Total | 67.072236 184.75926 8421
Next, we can generate a bar graph of income quartiles for each race. According to the following graph, it appears that the mean amount of remittances is somewhat constant between income quartiles.
graph bar (mean) totm_rec if household==1, over(incquart, relabel(1 "Q1" 2 "Q2" 3 "Q3" 4 "Q4")) by(race) ytitle(Mean of Remittances)
We can also look at income quartiles within each race based on the % who receive remittances rather than solely examining the amount of remittances. According to the graph below, the highest % of those who receive remittances are households whose income falls within the lowest income quartile. This holds true for each race.
graph bar (mean) remittances if household==1, over(incquart, relabel(1 "Q1" 2 "Q2" 3 "Q3" 4 "Q4")) by(race) ytitle(% Receive Remittances) title(% who receive remittances) name(Remittances)
III. Location
Another interesting question is whether remittances are more frequently received by rural households versus households located in urban areas. If we would like to see remittances by location, we can also use the tab command with row.
tab metro remittances if household==1
----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
remittances | metro - urban - rural
| Rural Urban Metro | Total
-----------+---------------------------------+----------
0 | 2,690 1,561 2,241 | 6,492
| 41.44 24.04 34.52 | 100.00
-----------+---------------------------------+----------
1 | 1,681 392 283 | 2,356
| 71.35 16.64 12.01 | 100.00
-----------+---------------------------------+----------
Total | 4,371 1,953 2,524 | 8,848
| 49.40 22.07 28.53 | 100.00
When displaying this in a graph, we can see that the rural area makes up the largest % of all locations which receive remittances.
graph pie remittances if household==1, over(metro) angle(90) plabel(_all percent, orientation(horizontal) color(white) format(%9.1f)) title(Amount of Remittances by Location) subtitle(For Households)
By using the tot_rec command, we can generate the amount of remittances and compare locations for each race.
graph bar (mean) totm_rec if household==1, over(metro, relabel(1 "rural" 2 "urban" 3 "metro")) by(race) ytitle(Mean of Remittances)
The above graph shows that the average amount of remittances for Black Africans is highest within the rural area. In comparison, the average amount of remittances received are significantly higher for Coloureds within the urban areas. (However, coloreds make up a larger % of urban areas compared to other racial groups). By generating binary variables for location, we can perform a t-test to examine the significance of the amount of remittances received by location within each race.
tab metro, gen (newmetro)
bysort race: ttest totm_rec if household==1, by (newmetro1)
This will display a t-test for the amount of remitances between rural and non-rural areas by all races. Displayed below is the t-test for Africans. A t-stat of -12 shows that there is a significant relationship between rurul and non-rural areas who are Black African.
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 1976 37.75932 3.68898 163.9834 30.52462 44.99402
1 | 4093 97.44249 2.946108 188.4818 91.66651 103.2185
---------+--------------------------------------------------------------------
combined | 6069 78.0103 2.349135 183.0065 73.40516 82.61544
---------+--------------------------------------------------------------------
diff | -59.68317 4.954665 -69.39607 -49.97026
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -12.0459
Ho: diff = 0 degrees of freedom = 6067
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
bysort race: ttest totm_rec if household==1, by (newmetro2)
This will display a t-test for the amount of remittances between urban and non-urban areas by all races. Displayed below is the t-test for Coloureds. This shows that there is a significant relationship between urban and non-urban households who receive remittances for Coloureds.
race = 02-colou
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 390 30.01112 7.623199 150.5461 15.0233 44.99895
1 | 268 86.87212 15.76357 258.0608 55.83541 117.9088
---------+--------------------------------------------------------------------
combined | 658 53.17031 7.919106 203.137 37.6205 68.72012
---------+--------------------------------------------------------------------
diff | -56.86099 15.97645 -88.23213 -25.48986
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -3.5591
Ho: diff = 0 degrees of freedom = 656
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0002 Pr(|T| > |t|) = 0.0004 Pr(T > t) = 0.9998
IV. Gender
Since remittances are typically known to be received by female headed households, we should also compare the amount of remittances received by the gender of the household head. Because the remittances variable is household level, we will need to generate a variable which separates the heads of households according to gender. This can be created using the egen command and the variable which defines the relationship to head rel_head. ( Rel_head=1 represents household head.) Likewise, gender_n=2 represents all female respondents.
gen femalehead=1 if rel_head==1 & gender_n==2
egen femalehead=sum (femalehead1), by (hhid)
The following command shows that only 25% of households are female headed.
tab femalehead if household==1
femalehead | Freq. Percent Cum.
------------+-----------------------------------
0 | 6,644 75.04 75.04
1 | 2,210 24.96 100.00
------------+-----------------------------------
Total | 8,854 100.00
Now, we can include the remittances variable to see the frequency of femaleheaded households who receive remittances. By including , col we can see how this compares to male-headed households who receive remittances.
tab femalehead remittances if household==1, col
| remittances
femalehead | 0 1 | Total
-----------+----------------------+----------
0 | 5,063 1,581 | 6,644
| 77.99 66.93 | 75.04
-----------+----------------------+----------
1 | 1,429 781 | 2,210
| 22.01 33.07 | 24.96
-----------+----------------------+----------
Total | 6,492 2,362 | 8,854
| 100.00 100.00 | 100.00
Using a bar graph, we can visually display the % of households who receive remittances by the gender of the household head.
graph hbar (mean) remittances if household==1, over(femalehead, relabel(1"male" 2"female")) title(% of Household heads who receive remittances)
The above bar chart shows that 1/3 more of female headed household receive remittances as compared to male headed households. The following displays a t-test for remittances between female headed and male headed households. With a t-stat of -10.69, we can reject the null hypothesis and confirm that there is a statistically significant difference between male and female households.
ttest remittances, by (femalehead)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 6644 .2379591 .0052247 .425866 .227717 .2482011
1 | 2210 .3533937 .0101707 .4781318 .3334485 .3733388
---------+--------------------------------------------------------------------
combined | 8854 .2667721 .0047005 .4422972 .257558 .2759862
---------+--------------------------------------------------------------------
diff | -.1154346 .0107922 -.1365898 -.0942794
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -10.6961
Ho: diff = 0 degrees of freedom = 8852
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
IV. Education Level
Now, we can generate a variable displaying the education level of each household head and apply this to all households. Then, we can test to see if there is a significant difference between education level and those who receive remittances. Before we examine educationlevel, we will need to recode the education level variable in the survey and generate a new education level so the leves are coded from lowest to highest.
generate educ_new = educ_c
label var educ_new "Recoded Education, years"
replace educ_new = . if educ_c < 0
replace educ_new = . if educ_c == 19
replace educ_new = 0 if educ_c == 17
replace educ_new = 0 if educ_c == 18
replace educ_new = 9 if educ_c == 11
replace educ_new = 12 if educ_c == 12
replace educ_new = 12 if educ_c == 13
replace educ_new = 12 if educ_c == 14
replace educ_new = 12 if educ_c == 15
Next, we can generate a variable which displays the education level attained by the household head.
gen educationlevel = educ_new if rel_head==1
egen educationlevelhead= min(educationlevel), by (hhid)
sum educationlevelhead if household==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
educationl~d | 7234 4.994747 4.246839 0 16
tab remittances, sum (educationlevelhead)
| Summary of educationlevelhead
remittances | Mean Std. Dev. Freq.
------------+------------------------------------
0 | 5.4546093 4.2927639 5695
1 | 3.2930474 3.5937053 1539
------------+------------------------------------
Total | 4.994747 4.2468391 7234
According to the above table, it is apparent that those who receive remittances have a lower education level for the household head as compared to those who do not receive remittances. We can also generate a graph to see this within each racial group.
graph bar (mean) educationlevelhead if household==1, over(remittances, relabel(1 "no remittances" 2 "remittances")) by(race) bar(1, bfcolor(black)) bar(2, bfcolor(forest_green)) ytitle(Mean Education Level for Households)
For all races, education level is lower for the household heads which receive remittances as compared to those who do not receive remittances. The below t-test (-13.4750) shows that there is a statistically significant difference for the education level of their household head between those who receive remittances and those who do not.
ttest educationlevelhead, by (remittances)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 5695 .2129939 .0054258 .4094593 .2023572 .2236305
1 | 1539 .3781676 .0123652 .4850874 .3539132 .402422
---------+--------------------------------------------------------------------
combined | 7234 .2481338 .0050787 .4319597 .238178 .2580896
---------+--------------------------------------------------------------------
diff | -.1651738 .0122578 -.1892026 -.141145
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -13.4750
Ho: diff = 0 degrees of freedom = 7232
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
V. Age
By using the egen and min command, we can generate a household level variable which designates the age of the household head. Combining the tab and sum commands, we can display the mean age for households that receive and do not receive remittances.
gen age1= age if rel_head==1
egen agehead= min(age1), by (hhid)
tab agehead, sum (remittances)
| Summary of agehead
remittances | Mean Std. Dev. Freq.
------------+------------------------------------
0 | 45.596973 15.037705 5682
1 | 52.32239 16.011878 1523
------------+------------------------------------
Total | 47.018598 15.492988 7205
The average age for household head appears to be higher. We can confirm the significance of this relationship through a t-test.
ttest agehead, by (remittances)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 5682 45.59697 .1994945 15.0377 45.20589 45.98806
1 | 1523 52.32239 .4102913 16.01188 51.51759 53.12719
---------+--------------------------------------------------------------------
combined | 7205 47.0186 .1825233 15.49299 46.6608 47.3764
---------+--------------------------------------------------------------------
diff | -6.725417 .4399975 -7.587941 -5.862893
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -15.2851
Ho: diff = 0 degrees of freedom = 7203
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
Now, we can use the recode command with the variable agehead to create a new variable agegroup for household heads.
recode agehead (20/34= 1 "20-34") (35/49= 2 "35-49") (50/65=3 "50-65") if agehead!=. & agehead>=20 & agehead<=65, gen (agegrouphead)
tab remittances agegrouphead, row
----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
remittance | RECODE of agehead
s | 20-34 35-49 50-65 | Total
-----------+---------------------------------+----------
0 | 1,510 2,079 1,389 | 4,978
| 30.33 41.76 27.90 | 100.00
-----------+---------------------------------+----------
1 | 243 386 524 | 1,153
| 21.08 33.48 45.45 | 100.00
-----------+---------------------------------+----------
Total | 1,753 2,465 1,913 | 6,131
| 28.59 40.21 31.20 | 100.00
Based on the above table and below pie chart, we see that the household heads between 50-65 represent the largest % of age groups receiving remittances.
graph pie remittances, over(agegrouphead) angle(90) plabel(1 percent, color(white) format(%9.1f)) plabel(2 percent, color(white) format(%9.1f)) plabel(3 percent, color(white) format(%9.1f)) title(Remittances by Age Group)
VI. Family Size
Household size is another characteristic which may significantly affect the amount of remittances received. By displaying the below tables, we see that the mean hh size is slightly bigger for those who receive remittances as compared to those who do not. With a t-test for hhsizem, we see that there is a significant relationship for hhsize for those who receive remittances and those who do not receive remittances.
sum hhsizem if household==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
hhsizem | 8848 4.552893 2.969938 1 30
tab remittances, sum (hhsizem)
| Summary of hh size members
remittances | Mean Std. Dev. Freq.
------------+------------------------------------
0 | 4.262785 2.8428703 6492
1 | 5.352292 3.1605045 2356
------------+------------------------------------
Total | 4.5528933 2.969938 8848
ttest hhsizem, by (remittances)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 6492 4.262785 .0352832 2.84287 4.193618 4.331952
1 | 2356 5.352292 .0651132 3.160505 5.224607 5.479977
---------+--------------------------------------------------------------------
combined | 8848 4.552893 .0315737 2.969938 4.491002 4.614785
---------+--------------------------------------------------------------------
diff | -1.089507 .0704906 -1.227685 -.9513291
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -15.4561
Ho: diff = 0 degrees of freedom = 8846
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
VI. Regression on remittances received using the created independent variables.
We can now run a logistic regression for the categorical variable remittances controlling for the characteristics displayed below. While there is a strong relationship between most of the independent variables on the receipt of remittances, our model shows a small R sq and it is difficult to determine any relationships of causality. However, the odds ratios show a distinct difference for gender of household head, location, and race for households that receive remittances as compared to those that do not. For example, households that receive remittances are more likely to be female headed and located in rural areas as shown from their the odds ratios which are signficantly greater than 1.
logistic remittances femalehead agehead educationlevelhead hhsizem race rural if household==1
Logistic regression Number of obs = 7128
LR chi2(6) = 718.52
Prob > chi2 = 0.0000
Log likelihood = -3312.3169 Pseudo R2 = 0.0978
------------------------------------------------------------------------------
remittances | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
femalehead | 2.509773 .1605082 14.39 0.000 2.214101 2.84493
agehead | 1.010594 .0022594 4.71 0.000 1.006175 1.015032
educationl~d | .9492634 .0097912 -5.05 0.000 .9302656 .9686491
hhsizem | 1.05578 .0105221 5.45 0.000 1.035357 1.076606
race | .8649627 .038124 -3.29 0.001 .7933776 .9430069
rural | 1.520474 .1067506 5.97 0.000 1.325004 1.744781
------------------------------------------------------------------------------
Remittances Sent
The variable for total monthly remittance expenditure is mxtrem. This is a continuous variable, which shows the amount of remittances sent by each household. In order to examine the characteristics of households that send remittances, we can generate a remittances sent variable if remittances are greater than 0 (The following command will give a value of 1 to respondents who send remittances and a 0 for those who do not).
gen remittancesset =.
replace remittances=1 if mxtrem >0 & household==1
replace remittances=0 if mxtrem <=0 & household==1
In order to display the mean remittances sent for all households, we can use the sum command:
sum mxtrem if household==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
mxtrem | 8822 36.07137 154.0136 0 5416.667
I. Race
The mean amount of remittances for each race can be shown by combining the tab and sum commands.
tab race if household==1, sum (mxtrem)
19 | Summary of tot month remittance
:population | expenditur
group | Mean Std. Dev. Freq.
------------+------------------------------------
01-afric | 32.933179 106.22933 6056
02-colou | 20.413481 109.0701 654
03-india | 34.70068 335.73636 245
04-white | 29.3492 215.66526 1166
------------+------------------------------------
Total | 31.463685 139.4425 8121
If we would like to show this visually, we can graph this data into a bar chart. The following graph shows that Africans and Indians send the largest amount of remittances out of all four races.
graph bar (mean) mxtrem if household==1, over(race) blabel(bar, format(%9.1f)) ytitle(Av erage amount of remittances ) title(Average Amount of Remittances Sent ) subtitle(By Households)
tab race remittancessent, row
+----------------+
| Key |
|----------------|
| frequency |
| row percentage |
+----------------+
19 |
:populatio | remittancessent
n group | 0 1 | Total
-----------+----------------------+----------
01-afric | 4,945 1,124 | 6,069
| 81.48 18.52 | 100.00
-----------+----------------------+----------
02-colou | 600 58 | 658
| 91.19 8.81 | 100.00
-----------+----------------------+----------
03-india | 231 14 | 245
| 94.29 5.71 | 100.00
-----------+----------------------+----------
04-white | 1,085 87 | 1,172
| 92.58 7.42 | 100.00
-----------+----------------------+----------
Total | 6,861 1,283 | 8,144
| 84.25 15.75 | 100.00
Display a pie chart showing remittances received according to race:
graph pie remittancessent if household==1, over(race) angle(90) plabel(1 percent, color(white) format(%9.1f)) plabel(2 percent, color(white) format(%9.1f)) plabel(4 percent, color(white) format(%9.1f)) title(Remittances Sent by Race)
Using the variable newrace, which was created in the previous section, we can run a t-test for the amount of remittances sent to see if there is a significant difference between the White and Black Africans. Based on the t-test below, we can accept the null hypothesis that there is no difference between the means of the two groups.
ttest mxtrem if household==1, by (newrace)
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 1166 29.3492 6.315837 215.6653 16.95751 41.74089
1 | 6056 32.93318 1.365059 106.2293 30.25718 35.60918
---------+--------------------------------------------------------------------
combined | 7222 32.35454 1.532811 130.2619 29.34978 35.3593
---------+--------------------------------------------------------------------
diff | -3.58398 4.165929 -11.75042 4.58246
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = -0.8603
Ho: diff = 0 degrees of freedom = 7220
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.1948 Pr(|T| > |t|) = 0.3896 Pr(T > t) = 0.8052
II. Income Quartile
Next, we will display the average amount of remittances sent according to income quartiles.
tab incquart if household==1, sum (mxtrem)
RECODE of |
totminc |
(total | Summary of tot month remittance
monthy | expenditur
income) | Mean Std. Dev. Freq.
------------+------------------------------------
first qua | 7.5786898 47.291882 1992
second qu | 27.926962 78.961371 2145
third qua | 65.269661 152.76022 2132
fourth qu | 44.650712 247.93051 2131
------------+------------------------------------
Total | 36.822103 155.16207 8400
We can show this visually using a bar graph of income quartiles per each race. According to the following graph, it seems that the mean amount of remittances sent tends to be from higher income quartiles for most races.
graph bar (mean) mxtrem if household==1, over(incquart, relabel(1 "Q1" 2 "Q2" 3 "Q3" 4 "Q4")) by(race) ytitle(Mean of Remittances Sent)
Similar to remittances received, we should also look at the % of households who send remittances by income quartiles for each race. According to the graph below, the highest % of households who send remittances are households within the highest income quartiles. This holds true for each race.
graph bar (mean) remittancessent if household==1, over(incquart, relabel(1 "Q1" 2 "Q2" 3 "Q3" 4 "Q4")) by(race) ytitle(% Receive Remittances) title(% who sent remittances) name(RemittancesSent)
III. Location
If we would like to see remittances sent by location, we can also use the tab command along with col.
tab metro remittancessent if household==1, col
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
metro - |
urban - | remittancessent
rural | 0 1 | Total
-----------+----------------------+----------
Rural | 3,756 615 | 4,371
| 51.29 40.33 | 49.40
-----------+----------------------+----------
Urban | 1,584 369 | 1,953
| 21.63 24.20 | 22.07
-----------+----------------------+----------
Metro | 1,983 541 | 2,524
| 27.08 35.48 | 28.53
-----------+----------------------+----------
Total | 7,323 1,525 | 8,848
| 100.00 100.00 | 100.00
Similar to the results displayed for remittances received, rural area makes up the largest % of all locations for remittances sent.
graph pie remittancesset if household==1, over(metro) angle(90) plabel(_all percent, orientation(horizontal) color(white) format(%9.1f)) title(% of Remittances Sent by Location) subtitle(For Households)
With the mxtrem command, we can generate the amount of remittances sent and compare between location within different races.
graph bar (mean) mxtrem if household==1, over(metro, relabel(1 "rural" 2 "urban" 3 "metro")) by(race) ytitle(Mean of Remittances Sent)
By looking within race, we can see that the amount of remittances sent is larger for most races (excluding Indians) for both urban and metro locations as compared to the mean amount of remittances sent from rural areas. By performing t-tests for each metro location compared to the other locations for remittances sent, we can see that there is a statistically significant difference for Africans. This is true for all metro locations only for Africans. The t-test displayed below provides an example of the difference between rural and non-rural locations for Black Africans.
bysort race: ttest remittancessent, by (newmetro1)
------------------------------------------------------------------------------------------
-> race = 01-afric
Two-sample t test with equal variances
------------------------------------------------------------------------------
Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
---------+--------------------------------------------------------------------
0 | 1976 .2889676 .0101997 .4533976 .2689644 .3089708
1 | 4093 .1351087 .0053439 .3418814 .1246319 .1455856
---------+--------------------------------------------------------------------
combined | 6069 .1852035 .0049869 .3884946 .1754275 .1949795
---------+--------------------------------------------------------------------
diff | .1538589 .0104581 .1333573 .1743605
------------------------------------------------------------------------------
diff = mean(0) - mean(1) t = 14.7119
Ho: diff = 0 degrees of freedom = 6067
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 0.0000
IV. Gender
Now, we can compare the amount of remittances sent by household head according to gender. We can use the same variable femalehead, which was generated in the above analysis.
tab femalehead remittancessent if household==1, col
+-------------------+
| Key |
|-------------------|
| frequency |
| column percentage |
+-------------------+
| remittancessent
femalehead | 0 1 | Total
-----------+----------------------+----------
0 | 5,346 1,298 | 6,644
| 73.00 84.78 | 75.04
-----------+----------------------+----------
1 | 1,977 233 | 2,210
| 27.00 15.22 | 24.96
-----------+----------------------+----------
Total | 7,323 1,531 | 8,854
| 100.00 100.00 | 100.00
Using a bar graph, we can visually display the % of households who send remittances by the gender of the household head.
graph hbar (mean) remittancessent if household==1, over(femalehead, relabel(1"male" 2"female")) title(% of Household heads who send remittances)
Based on the horizontal bar chart above, we can see that male headed households make up a larger % that send remittances as compared to femal headed households.
IV. Education Level
In this section, we can use the variable generated that displays the education level of each household head and applies this to members of households. In order to see the mean education level of household head for those who receive remittances and those who do not receive remittances, we can use the following command.