Professional Documents
Culture Documents
IMPOVERISHED LIVES
A 2017 DATASET STUDY
A borrower applies for a loan The loan goes through the underwriting and approval Loans are posted to Kiva website
process
Lenders crowdfund the loan in increments of $25 or more Fundraising complete Borrowers repay the loan
Lenders use repayments to fund new loans, donate, or withdraw the money
However, this rather long funding period of 61 days is an anomaly - Kiva generally operates in 30-day
periods after which the loans expire. For public funding, Kiva promises a 93% chance of full-funding for
the 30-day period model. 88% of the loans in the dataset got fully funded within 30 days. There was no
statistical implication that the proportion of loan that gets fully funded within 30 days is at least 93%
1
(Appendix C). This might be due to the fact that when a 60-day period is used instead of a 30-day period,
this causes for some delay due to relative inaction as there is more time for funding.
The means of the amount of loans and amount of funded loans were analyzed using statistical techniques.
The mean of the loan amount was significantly higher than the crowdfunded amount and this will hold
true for 99% within all samples. (Appendix D)
Fig-02: Weekly Performance
Count of loans
600000 800
Sum of loans
With the help of these loans Kiva hopes to change the lives of people spanning the globe. Presently, Kiva
lends to entrepreneurs in 77 countries. (contributors, 2020) Of these 77 countries, 59 were reached within
the week in question. Cambodia, Philippines, the United States, Peru, Rwanda, Kenya and Lebanon were
the top performers in borrowing and accounted for 51% of total loans disbursed. The loan amount in
Cambodia was a whopping $460,800 and 94% of this amount was funded within due time. (Appendix F)
Fig-03: Country-wise total loan amount with cumulative
Nigeria, Madagascar and Togo were the countries with the smallest average loan sizes of $169, $218 and
2
$279 respectively. Here comes the success of Kiva in achieving its goal. With average loans even as small
as these, underprivileged people are getting a shot at changing their lives. The highest average loan
($4,653) was observed in the US and this is expected as the lenders from there would need the money for
bigger businesses compared to those from low-income countries.
Fig-04: Country-wise average loan amount with cumulative
3
Kiva deals with three types of repayment plans: monthly, irregular and bullet. Loan size is observed to
play an important role in determining the type of repayment plan. It was observed that in 95 cases out
100, the average loan amount of these different repayment plans would be different (Appendix H).
Fig 07: Continent-wise total loan amount with
Different people across the globe contribute to the specific loans they deem worthy. The number of
lenders contributing to a loan was strongly and positively related to the loan/funded amount, and
moderately to the duration of the loan in months. (Appendix K). Based on these correlations and factoring
in other variables, a model might be generated to predict the lender count of individual loans.
Lender count= -8.55 - 0.01*Duration + 0.46*term in months+ 2.40*interval monthly + 6.66*interval bullet
+ 0.03*loan amount + 1.70*subcontinent Asia + 2.89*subcontinent Africa - 3.50*subcontinent NA -
5.91*subcontinent SA
The relationship between the duration and the lender count is inverse. With an increase of unit additional
month in the loan duration, almost 46 additional lenders can be expected. Considering the reference
category as an irregular loan interval, there are a requirement 2 and 7 more lenders respectively
compared to irregular loan intervals. On the other hand, keeping the reference category as Europe, we
see that in Asia and Africa, Kiva needs more lenders than they need in Europe. Meanwhile, for processing
loans in North American and South American countries, Kiva needs a smaller number of lenders compared
to European countries. Considering the lender count differences based on different subcontinents, Kiva
might want to extend the funding deadlines for the loans going to Asian and African nations. (Appendix L)
Despite the valiant efforts of Kiva to change peoples’ lives, some people argue that the interest rates of
many microcredit institutions are unreasonably high. As of November 2018, the most frequent interest
rate lay between 20.5-30%. Assuming a 25% interest for the observed loan, the interests can be projected.
For instance, due to longer term and compound interest a borrower from Kenya to fund education might
end up paying more than the principal in the form of interest. So naturally it begs the question, does Kiva
actually helps the borrowers to get funded or maybe it is actually a platform by which microfinance
institutions can conveniently finance themselves for lending. Whatever it is, as the data suggest, Kiva is
helping millions of people across the globe in getting that money, as little as it may be, that might change
their lives.
4
References
contributors, K., 2020. KIVA. [Online]
Available at: https://www.kiva.org/about
Tables of Appendices
Appendix A: Descriptive Statistics of Loans Posted ...................................................................................... A
Appendix B: Confidence Interval of Proportion of Loans Funded ................................................................ B
Appendix C: Z test for Proportion of Loan Funded Within 30 Days .............................................................. C
Appendix D: Comparison of Means of Loan Amount and Funded Amount .................................................D
Appendix E: Comparison of Weekday and Weekend Properties ................................................................. E
Appendix F: Country-Wise Performance ....................................................................................................... I
Appendix G: Sector-Wise Performance ........................................................................................................ K
Appendix H: Comparison of Means of Repayment Term ............................................................................. L
Appendix I: Comparison of Means of Loans from Different Continents .....................................................M
Appendix J: Continent-Wise Performance ................................................................................................... N
Appendix K: Correlation Matrix ................................................................................................................... O
Appendix L: Regression Model of Number of Lenders ................................................................................. P
Appendix M: Relationship Between Gender and Sector ............................................................................. Q
5
Appendix A: Descriptive Statistics of Loans Posted
loan_amount
Mean 821.8412
Standard Error 22.38504
Median 475
Mode 225
Standard Deviation 1427.738
Sample Variance 2038435
Kurtosis 385.1119
Skewness 14.0647
Range 49950
Minimum 50
Maximum 50000
Sum 3343250
Count 4068
Confidence Level(95.0%) 43.88694
loan_amount
Mean 702.9386
Standard Error 11.30178
Median 475
Mode 225
Standard Deviation 715.4127
Sample Variance 511815.3
Kurtosis 9.377523
Skewness 2.637568
Range 5025
Minimum 50
Maximum 5075
Sum 2816675
Count 4007
Confidence Level(95.0%) 22.15778
The average loan amount with outliers is 821.84 whereas without outliers 702.94. Without the outliers,
loan amounts show relatively low yet high positive skewness (2.63). A total of 61 outliers were found.
A
Appendix B: Confidence Interval of Proportion of Loans Funded
96.7% of the total loan has been fully funded. The confidence interval at 95% confidence level is 0.005. It
can be said with 95% confidence that from at least 96.2% to maximum 97.3% of the loans will be fully
funded
B
Appendix C: Z test for Proportion of Loan Funded Within 30 Days
This dataset has a timeframe of 60 days and 88% of the loans have been funded within 30 days of
posting.
Here, H0: 93% or more of all loans are funded within 30 days
H1: Less than 93% of all loans are funded within 30 days
Here Z value exceeds Z Critical. We successfully reject the null hypothesis. We do have significant
statistical evidence (95% C.L.) that 93% or more of all loans are not funded within 30 days.
C
Appendix D: Comparison of Means of Loan Amount and Funded Amount
funded_amount loan_amount
Mean 776.6654376 821.8411996
Variance 1870446.113 2038434.512
Observations 4068 4068
Pearson Correlation 0.964118599
Hypothesized Mean Difference 0
df 4067
t Stat -7.599941548
P(T<=t) one-tail 1.82775E-14
t Critical one-tail 2.32726513
P(T<=t) two-tail 3.65549E-14
t Critical two-tail 2.577038726
H1: Means of Funded amounts and Loan amounts are not equal
As P-value<0.001, we reject the null hypothesis. At 99% confidence level, there is significant statistical
evidence that the means of loan amount and funded amount are different.
D
Appendix E: Comparison of Weekday and Weekend Properties
Weekly Performance
E
F-Test Two-Sample for Variances
Variable 1 Variable 2
Mean 144.5 755.8
Variance 40.5 26517.7
Observations 2 5
df 1 4
F 0.001527282
P(F<=f) one-tail 0.029301017
F Critical one-tail 0.004452692
Variable
Variable 1 2
Mean 144.5 755.8
Variance 40.5 26517.7
Observations 2 5
Hypothesized Mean Difference 0
df 4
-
t Stat 8.378069095
P(T<=t) one-tail 0.000555115
t Critical one-tail 2.131846786
P(T<=t) two-tail 0.001110231
t Critical two-tail 2.776445105
For F Test: H0: Variances of loan counts for weekends and weekdays are equal
H1: Variances of loan counts for weekends and weekdays are not equal
P<0.05 attests to the fact we do not have significant statistical evidence that Variances of loan counts for
weekends and weekdays are equal. We reject the null hypothesis.
H0: Means of loan counts for weekdays and weekends are equal
H1: Means of loan counts for weekdays and weekends are not equal
As P-value<0.05, we do not have significant statistical evidence to accept H0. This means that the means
for both of the samples are not equal.
F
F-Test Two-Sample for Variances
Variable 1 Variable 2
Mean 132125 615800
Variance 351125000 19138725000
Observations 2 5
df 1 4
F 0.018346311
P(F<=f) one-tail 0.101199998
F Critical one-tail 0.004452692
Variable 1 Variable 2
Mean 132125 615800
Variance 351125000 19138725000
Observations 2 5
Hypothesized Mean Difference 0
df 4
-
t Stat 7.644413402
P(T<=t) one-tail 0.000786602
t Critical one-tail 2.131846786
P(T<=t) two-tail 0.001573203
t Critical two-tail 2.776445105
For F Test: H0: Variances of loan amounts for weekends and weekdays are equal
H1: Variances of loan amounts for weekends and weekdays are not equal
P>0.05 attests to the fact we do have significant statistical evidence that Variances of loan amounts for
weekends and weekdays are equal. We fail to reject the null hypothesis.
H0: Means of loan amounts for weekdays and weekends are equal
H1: Means of loan counts for weekdays and weekends are not equal
As P-value<0.05, we do not have significant statistical evidence to accept H0. This means that the means
for both of the samples are not equal.
G
F-Test Two-Sample for Variances
Weekends Weekdays
Mean 914.3598616 814.7658111
Variance 1011780.926 2116531.847
Observations 289 3779
df 288 3778
F 0.478037185
P(F<=f) one-tail 7.99361E-15
F Critical one-tail 0.862914526
Weekends Weekdays
Mean 914.3598616 814.7658111
Variance 1011780.926 2116531.847
Observations 289 3779
Pooled Variance 2038280.921
Hypothesized Mean Difference 0
Df 4066
t Stat 1.143005371
P(T<=t) one-tail 0.126551828
t Critical one-tail 1.645228472
P(T<=t) two-tail 0.253103656
t Critical two-tail 1.960547596
For F Test: H0: Variances of the average loan for weekends and weekdays are equal
H1: Variances of the average loan for weekends and weekdays are not equal
As P<0.05, we reject the null hypothesis. It attests to the fact that; we do not have significant statistical
evidence that variances of average loans for weekends and weekdays are equal.
H0: Means of the average loan for weekdays and weekends are equal
H1: Means of the average loan for weekdays and weekends are not equal
As P-value>0.05, we fail to reject H0. It is can be said with significant statistical evidence that the average
loan for weekdays and weekends are equal.
H
Appendix F: Country-Wise Performance
I
Nicaragua 117 $104,025 $101,250 $889 97%
Nigeria 35 $5,925 $5,925 $169 100%
Pakistan 198 $92,575 $92,575 $468 100%
Palestine 57 $94,450 $94,450 $1,657 100%
Paraguay 3 $6,050 $6,050 $2,017 100%
Peru 131 $222,300 $222,300 $1,697 100%
Philippines 778 $246,100 $246,100 $316 100%
Puerto Rico 1 $10,000 $10,000 $10,000 100%
Rwanda 46 $203,650 $178,600 $4,427 88%
Senegal 39 $54,525 $54,525 $1,398 100%
Sierra Leone 52 $47,050 $44,125 $905 94%
Tajikistan 124 $73,225 $70,650 $591 96%
Tanzania 1 $2,475 $2,475 $2,475 100%
Thailand 1 $400 $400 $400 100%
Togo 28 $7,825 $7,825 $279 100%
Turkey 26 $10,675 $10,675 $411 100%
Uganda 141 $73,125 $72,150 $519 99%
United States 49 $228,000 $135,450 $4,653 59%
Vietnam 57 $69,925 $69,925 $1,227 100%
Yemen 7 $4,375 $4,375 $625 100%
Zambia 5 $8,250 $8,250 $1,650 100%
Zimbabwe 12 $17,500 $17,500 $1,458 100%
J
Appendix G: Sector-Wise Performance
K
Appendix H: Comparison of Means of Repayment Term
SUMMARY
Groups Count Sum Average Variance
monthly 2458 2015225 819.8637 927036.3
irregular 1410 1077000 763.8298 3801480
bullet 200 251025 1255.125 3085416
ANOVA
Source of Variation SS df MS F P-value F crit
3.05E-
Between Groups 42301689 2 21150845 10.42411 05 2.997941
Within Groups 8.25E+09 4065 2029031
H1: Means of loans pertaining to different payment terms are not equal
As P-value<0.05, we reject H0. It is can be said with significant statistical evidence that the average loan
of different payment terms are not equal.
L
Appendix I: Comparison of Means of Loans from Different Continents
SUMMARY
Groups Count Sum Average Variance
Asia 2151 1306500 607.3919 325337.2
Africa 1125 910150 809.0222 4170508
Europe 67 94350 1408.209 475878.6
North America 378 515275 1363.161 3417129
South America 347 516975 1489.841 3455997
ANOVA
Source of Variation SS df MS F P-value F crit
5.27E-
Between Groups 3.88E+08 4 96936704 49.83872 41 2.374119
Within Groups 7.9E+09 4063 1945008
As P-value<0.05, we reject H0. It is can be said with significant statistical evidence that the loans from
different continents are not equal.
M
Appendix J: Continent-Wise Performance
N
Appendix K: Correlation Matrix
O
Appendix L: Regression Model of Number of Lenders
SUMMARY OUTPUT
Regression Statistics
Multiple R 0.93
R Square 0.86
Adjusted R Square 0.86
Standard Error 14.69
3934.0
Observations 0
ANOVA
Signific
df SS MS F ance F
5352540 59472 275
Regression 9 .55 6.73 7.13 0
3 846426. 215.7
Residual 924 45 1
3 6198967
Total 933 .00
P-
Coeffic Standar valu Lower Upper Lower Upper
ients d Error t Stat e 95% 95% 99.0% 99.0%
Intercept -8.55 2.22 -3.85 0.00 -12.91 -4.20 -14.28 -2.83
Duration in hours (from
posted to funded) -0.01 0.00 -5.59 0.00 -0.01 0.00 -0.01 0.00
term_in_months 0.46 0.04 12.24 0.00 0.38 0.53 0.36 0.55
IntervalMonthly 2.40 0.53 4.56 0.00 1.37 3.43 1.04 3.75
IntervalBullet 6.66 1.26 5.30 0.00 4.19 9.12 3.42 9.90
144.6
loan_amount 0.03 0.00 6 0.00 0.03 0.03 0.03 0.03
SubcontinentAsia 1.70 2.06 0.82 0.41 -2.34 5.74 -3.61 7.00
SubcontinentAfrica 2.89 2.10 1.37 0.17 -1.24 7.01 -2.53 8.31
SubcontinentNA -3.50 2.18 -1.61 0.11 -7.77 0.77 -9.11 2.12
SubcontinentSA -5.91 2.23 -2.65 0.01 -10.28 -1.54 -11.65 -0.17
The regression shown above is intended to retrieve the impacts of different variables on Lender count.
The adjusted R square shows that it is a good fit to the model. The negative value of the intercept denotes
that intercept does not have any managerial interpretation, but is there just to fit the model.
P
Appendix M: Relationship Between Gender and Sector
Observed Frequency
Agricul Ar Cloth Constru Educa Entertain Fo Hea Hous Manufact Persona Ret Servi Transport Grand
ture ts ing ction tion ment od lth ing uring l Use ail ces ation Total
52
female 689 86 132 9 118 2 7 44 139 11 87 490 182 29 2545
male 237 6 6 10 53 4 75 26 42 5 44 73 40 33 654
Grand 60
Total 926 92 138 19 171 6 2 70 181 16 131 563 222 62 3199
Expected Frequency
AgriculArt Cloth Constru Educa Entertain Foo Hea Hous Manufac Persona Ret Servi Transpor Grand
ture s ing ction tion ment d lth ing turing l Use ail ces tation Total
73. 109. 136.0 478. 55. 144. 447. 176. 2545.0
female 736.69 19 79 15.12 4 4.77 93 69 00 12.73 104.22 90 61 49.32 0
18. 28.2 123. 14. 37.0 115. 45.3
male 189.31 81 1 3.88 34.96 1.23 07 31 0 3.27 26.78 10 9 12.68 654.00
Grand
Total 926 92 138 19 171 6 602 70 181 16 131 563 222 62 3199
Row Agricul Art Cloth Constru Educa Entertain Fo Hea Hous Manufact Persona Ret Servi Transpor Grand
Labels ture s ing ction tion ment od lth ing uring l Use ail ces tation Total
2.2 4.8 2.4 3.9
female 3.09 4 4.49 2.47 2.39 1.61 3 5 0.17 0.23 2.84 6 0.16 8.38 39.33
8.7 17.4 18. 9.5 15.
male 12.01 2 9 9.63 9.31 6.27 78 5 0.67 0.91 11.07 40 0.64 32.59 153.05
192.38
Grand 10. 21.9 23. 12. 19.
Total 15.10 96 8 12.10 11.70 7.88 60 00 0.85 1.15 13.91 36 0.80 40.97
Q
DOF 13
Observed Chi 192.38
Chi Critical 5.89
P value 5.0E-34
H0: Borrowers’ gender and sector of the loan are not dependent.
Here, degree of freedom is 13. As p<.05 or Chi critical< Chi observed, we reject the null hypothesis. So, we have significant statistical evidence
that there is relationship between the borrowers’ gender and sector.