Evidence from BRAC TUP Program In Bangladesh ASSET AND SKILL TRANSFER IMPROVE THE ECONOMIC LIVES OF ULTRA-POOR YOUTH? Evidence from BRAC TUP Program In Bangladesh By Saad Noor Quayyum

CAN ASSET AND SKILL

TRANSFER IMPROVE THE

ECONOMIC LIVES OF ULTRA-

POOR YOUTH?

Evidence from BRAC TUP Program

In Bangladesh

By

Saad Noor Quayyum (IMF)

Tanvir Sobhan (BRAC University)

BACKGROUND

TUP AND GRADUATION MODEL (GM)

BACKGROUND


The Graduation Approach has its foundation in a

model named TUP created by BRAC, which tweaked

the program over the years before initiating its own

randomized studies.

CGAP and the Ford Foundation launched 10 pilots in

eight countries in 2006.

Currently more than 33 countries have adopted and

successfully implemented the ultra poor graduation

model.

BACKGROUND


Recently Science has published a research article,

which tested the effectiveness of “Graduation model”

in six countries by following 21,000 of the world’s

poorest people for three years.

The data showed this approach led to large and lasting

impacts on their standard of living.

INTRODUCTION

BRAC's Targeting Ultra Poor (“TUP”) program focuses

on the poorest women in rural Bangladesh.

TUP program in Bangladesh represents a significant

attempt to lift large numbers of women, and their

dependents, out of extreme poverty.

TUP program is designed in such a way that ultra-

poor women can break out of seasonal wage

employment and enter into a more stable self-

employment.

INTRODUCTION

This study focuses on the effects of the TUP program

on the youth defined as those between the ages of 18

and 25.

We look at how the program affects occupational

choices, food security, and income and asset

accumulation among the ultra-poor youth.

TUP PROGRAM DESCRIPTION

BRAC central office selects the most vulnerable

districts in rural Bangladesh based on the food

security maps of the World Food Program.

BRAC employees from local branch offices within

those districts select the poorest communities in their

branch.

Program officers then use a participatory wealth

ranking that aggregates the private information of

communities to classify households into wealth

classes.

PROGRAM DESCRIPTION

Participant women including the youth receive

productive assets, such as cows, goats, and poultry

etc., accompanied by intensive skills training. They

also receive subsistence allowance for the first 40

weeks after the livestock asset transfer.

THE LIVES OF ULTRA POOR YOUTH AT BASELINE

Treated Communities Control Communities Differences in Mean

Whole

sample

18 to 25

years old

Whole

sample

18 to 25

years old

18 to 25

years old

Whole

sample

Mean age of primary female 39.18 21.99 40.55 22.35 -0.36 -1.37

(13.63) (2.33) (12.56) (2.32)

Size of household 3.39 3.53 3.10 3.60 -0.08 0.29

(1.69) (1.02) (1.71) (1.06)

Whether primary female 7.3% 24.0% 6.7% 27.0% -3.0% 0.6%

can read and write (%) (0.26) -0.43 -0.25 -0.45

Primary female's number of 0.59 2.01 0.52 2.23 -0.22 0.07

year of schooling (1.65) -2.64 -1.60 -2.84

Food security (%) 46.0% 51.0% 36.0% 40.0% 11.0% 10.0%

(0.49) (0.50) (0.48) (0.49)


Treated Communities Control Communities Differences in Mean

Whole sample 18 to 25 years

old

Whole

sample

18 to 25

years old

18 to 25

years old

Whole

sample

Land Ownership (%) 6.7% 6.5% 6.2% 4.7% 1.8% 0.5%

(0.25) (0.25) (0.24) (0.21)

Net asset 16184 10820 17920 10164 656 -1736

(29,480) (15,751) (48,870) (15,171)

Income 4355 2186 5629 2844 -658 -1274

(5,261) (3,726) (6,028) (4,355)

Baseline sample size 4,009 642 2,637 339

Table 1: Summary statistics of treatment and control groups at baseline

About 16% of the treatment group and 12% of the

control group were youth (18 – 25 years of age).

Overall educational attainment is very low among

ultra-poor women.

Only 24% of youth in the treatment group and 27% of

the youth in the control group can read and write

compared to approximately 7% of all women in the

whole sample for both treatment and control groups.


About 49% of the youth in the treatment group could

not afford two meals a day, while about 60% of the

youth in control group could not afford two meals a day

most of the time.

It is found that land ownership rate is very low, while

only about 6.5% of the whole sample in the treatment

and control groups own any land, this portion is even

smaller (about 5%) for the control group of the youth

sample.


The net assets and income of ultra-poor females in

general is also lower (i.e. The average net asset of the

treatment group was about 10,800 Taka and the

average net asset of the control group was about

10,200 Taka in 2007)

Average income ranged from 2200 Taka to 2800 Taka

from the treatment to the control groups of youth,

compared to 4355 Taka and 5629 Taka respectively for

the treatment and control groups of the whole sample.


Baseline survey (2007) Treatment Control

Whole sample 18 to 25 years old Whole sample 18 to 25

years old

Specialized in wage employment (%) 25.70 28.30 30.60 31.20

Specialized in self-employment (%) 30.30 29.40 29.30 28.80

Engaged in both occupations (%) 26.40 24.50 27.20 26.50

Out of the labor force (%) 17.6 17.80 12.9 13.50

Table 2: Occupational choices at baseline


From table 2, it is quite evident that the choice of

occupation by the youth is not much different than the

whole sample and is consistent across treatment and

control groups.

In all cases the difference in the means is less than 1/6

of standard deviation, or well below the 0.25 suggested

by Imbens and Woolridge (2009), suggesting that

randomization had been successful in yielding fairly

similar control and treatment groups.


EVALUATION STRATEGY & DATA

BRAC implemented a large-scale randomized control

trial.

This RCT experiment was carried out in 20 treatment

and 20 control branches in 2007.

All communities within the 20 treatment branches are

treated in 2007 and all communities within the 20

control branches are kept as controls until 2011.

Over the four years from baseline to end line, there is

a 13% attrition rate among eligible households.

ESTIMATION TECHNIQUE

We observe the individuals in our sample at three different times, in 2007 right before the treatment, in 2009 and 2011.

In order to find the causal effect of the program on non-indicator outcome such as income and net asset, we estimate the following model:

is the welfare measure for individual i in subdistrict d at time t.

is a dummy variable which takes the value of 1 if individual i in the subdistrict d received the treatment and zero otherwise.


is a dummy variable which takes the value of 1 for

2009 and 𝑊𝑡11 is a dummy variable which takes the

value of 1 for 2011.

The parameters of interest are and which

measures the causal effect of the program two and four

years from the first time the program is administered.


To see the impact of the program on whether the

households have two meals to eat every day most of

the time, we estimate the following probit model:

𝑝𝑟𝑜𝑏𝑖𝑡(𝐹𝑖𝑑𝑡) = 𝛼 + 𝛿𝑇𝑖𝑑 + 𝛽1𝑊𝑡09𝑇𝑖𝑑 + 𝛽2𝑊𝑡

11𝑇𝑖𝑑 + 𝛾1𝑊𝑡

09 + 𝛾2𝑊𝑡11 + 𝑒𝑖𝑑𝑡

where, 𝐹𝑖𝑑𝑡takes the value of 1 if the household have

two meals a day most of the time and zero otherwise.

(1) (2)

Youth Sample Whole Sample

Marginal Effect of Program in 2009 0.122*** 0.113***

(0.0381) (0.0267)

Marginal Effect of Program in 2011 0.124*** 0.112***

(0.0374) (0.0267)

Observations 2,956 20,250

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 3: Food Security

FINDINGS: IMPACT OF TUP PROGRAM ON

ULTRA-POOR YOUTH (FOOD SECURITY)

Youth who were part of the TUP program were on

average 0.12 less likely to be hungry i.e. their

probability of having two meals a day increased by

0.12. Table 3 contains the result from the probit

regression.

The table shows the marginal effect of the program on

the treated youth in Column 1 and the whole sample in

Column 2. The effect is large and statistically

significant in both cases.

The coefficients for the youth and for the whole sample

are close to each other.



The effect of the program is about the same two and

four years from its onset.

After the first two years, BRAC stopped providing

training to the treatment group.

The impact of the asset transfer and the skills training

on this particular measure of food security thus seems

to continue even after the intervention period, which

indicates that the effect of the program on reducing

hunger among the ultra-poor youth is more permanent

than temporary.



(1) (2)

Youth Sample Whole Sample

𝛽1, Program effect after 2 years 0.120** 0.214***

(0.0514) (0.0243)

𝛽2, Program effect after 4 years 0.107** 0.202***

(0.0489) (0.0250)


R-squared 0.122 0.082


*** p<0.01, ** p<0.05, * p<0.1

Table 4: Effect of the Program on Real Net Asset


ULTRA-POOR YOUTH (NET ASSET)



The effect of the program on net asset accumulation is

much higher for participant youth than non-

participants. Participants increased household net

assets by 12% in 2009 and about 11% in 2011.

Net asset was deflated using the rural CPI for

Bangladesh.

We also subtracted 9,500 Taka from the real net asset

of participants in 2009 and 2011 to take into account

the initial asset transfer from BRAC in 2007.

2007 was used as the base year.

from Column one in the table that the coefficients of interest 𝛽1and 𝛽2 are both statistically significant.

The log linear model allows the coefficients of interest to be interpreted as percentage change.

Program participating increased the household net assets of youths by 12% in 2009 and about 11% in 2011. Note that these increases are net of the initial 9,500 transfer of asset.

We run a F-test to see if 𝛽1and 𝛽2are equal to each other. Indeed we cannot reject the null hypothesis that they are equal to each other. Thus the effect of the program on net asset seems to be more of permanent nature than temporary.



(1) (2)

18 to 25 years old Whole Sample

𝛽1, Program effect after 2 years 0.459*** 0.250***

(0.0791) (0.0356)

𝛽2, Program effect after 4 years 0.155* 0.167***

(0.0795) (0.0352)


R-squared 0.089 0.037


*** p<0.01, ** p<0.05, * p<0.1


ULTRA-POOR YOUTH (REAL INCOME)

Table 5: Effect of the Program on the Real Income

Table 5 contains the results of estimating equation 1

with log of real income as the dependent variable.

Column 1 contains the results for the regression on the

youth sample. Program participation seems to increase

their income by 45% at the end of two years. This is

large and economically significant effect.

Effect of program participation seems to diminish over

time as the estimate of 𝛽2 is about 0.15. The

hypothesis that 𝛽1is equal to 𝛽2can be rejected at the

95 percent confidence level using an F-test.





The youth thus seems to lose some of their income

generating capacity when then the BRAC training

stops (but still has higher income than those that did

not participate in the program).

Similar pattern of 𝛽1 being greater than 𝛽2 can also be

seen in Column 2 for the whole sample.

In the whole sample, program participation increase

real income by 25 percent after the first two years and

about 16 percent after four years.

Thus in the long run the program participation seems

to increase income by about 15 percent in both the

sample.

We found that four years after the program,

participants maintained the occupational choices they

made two years after program initiation. In contrast,

the distribution of occupations across control

communities, for both youth and all women, remained

the same across the four years (Figure 3 and 4).

This suggests that the effect of the TUP program does

not coincide with an underlying process of change in

occupational structure for the poorest women in rural

Bangladesh.


ULTRA-POOR YOUTH


ULTRA-POOR YOUTH

0

10

20

30

40

50

60

Specialised in

wage

employment

(%)

Specialised in

self

employment

(%)

Engaged in

both

occupations

(%)

Out of the

labor force (%)

Figure 1: Treatment communities at baseline, midline and end line : Whole

sample

2007 2009 2011

0

10

20

30

40

50

60

Specialised in

wage

employment

(%)

Specialised in

self

employment

(%)

Engaged in

both

occupations (%)

Out of the labor

force (%)

Figure 2: Treatment communities at baseline, midline and end line :

Youth sample

2007 2009 2011


ULTRA-POOR YOUTH

In Figure 1 and 2, we can see the dramatic change in

the occupational structure of the ultra-poor in treated

communities, both for youth and for whole samples

relative to their counterparts in control communities

(Figure 3 and 4).


ULTRA-POOR YOUTH


ULTRA-POOR YOUTH

0

10

20

30

40

Specialised in

wage

employment

(%)

Specialised in

self

employment

(%)

Engaged in

both

occupations

(%)

Out of the

labor force (%)

Figure 3: Control communities at baseline, midline and end line :

Whole sample

2007 2009 2011


ULTRA-POOR YOUTH

0

5

10

15

20

25

30

35

40

Specialised in

wage

employment

(%)

Specialised in

self

employment

(%)

Engaged in

both

occupations

(%)

Out of the

labor force (%)

Figure 4: Control communities at baseline, midline and end line :

Youth sample

2007 2009 2011

CONCLUSION

The improvement in food security and the accumulation of

assets does not diminish even two years after the program

intervention ends.

Although income generation falls two years after program

intervention ends, the program participants still have

higher income than those who do not participate.

The biggest limitation we faced in our search for the

differential impact the BRAC TUP program has on the

youth population is the fact that it was not exclusively

targeted to youth population.

We need to rethink a separate intervention strategy- a

different set of skills and training program- to cater the

potentials of the youth and lift them out of poverty.

THANK YOU!

Evidence from BRAC TUP Program In Bangladesh ASSET AND SKILL TRANSFER IMPROVE THE ECONOMIC LIVES OF ULTRA-POOR YOUTH? Evidence from BRAC TUP Program In Bangladesh By Saad Noor Quayyum

Documents