USING BIOMETRICS TO EVIDENCE NOTE
DELIVER CASH PAYMENTS
TO WOMEN: EARLY RESULTS
FROM AN IMPACT EVALUATION
IN PAKISTAN1
MARCH 2022
Summary
Pakistan’s Benazir Income Support Program (BISP) provides unconditional cash transfers to women in
7 million of the country’s poorest households. Beginning in 2017, BISP adopted biometric verification
as a part of its payment delivery system, which required female beneficiaries to collect funds in
person and was accompanied by additional changes in payment point location and increased reliance
on human cash-out agents. The goal of these reforms—collectively referred to as the Biometric
Verification System (BVS)—was to streamline the payment process and help ensure that women
beneficiaries received their funds.
This evaluation uses the staggered rollout of BVS to rigorously quantify its impact on beneficiaries’
control of cash, amount of money received, and experience collecting benefits during the early
implementation of these reforms. It provides important lessons for the BISP scheme and for other
policymakers considering the use of this technology for the delivery of social benefits.
BVS Reforms
The BVS reforms involved a number of concurrent changes to the delivery system that were
implemented gradually across districts between 2017 and 2019, including:
• New authentication technology: Under the new system, beneficiaries confirm their identity at
payment points by providing fingerprints that are matched against the national ID database; this
1 This note was prepared by Julia Clark (World Bank), Muhammad Haseeb (University of Geneva), Amen Jalal (London
School of Economics), Bilal Siddiqi (University of California [UC]–Berkeley, CEGA), and Kate Vyborny (Duke University),
based on early results from a study by Haseeb, Jalal, Siddiqi, and Vyborny supported by the World Bank’s Identification for
Development (ID4D) initiative and Development Impact Evaluation department (DIME), under the leadership of Vyjayanti
Desai and Ariana Legovini. This study was approved by Duke Institutional Review Board (IRB) 2020-0285 and the analysis
was preregistered with the Experiments in Governance and Politics (EGAP) registry (20200212AB). It would not have
been possible without the collaboration of Pakistan’s Benazir Income Support Program, including BISP Secretary Ali Raza
Bhutta, former Director General of Monitoring and Evaluation Shujaat Farooq, the Director General Cash Transfers Noor
Rehman, and the Management Information System (MIS) team, and guidance from the World Bank’s Social Protection and
Jobs team, including Gul Najam Jamy, Amjad Zafar, Nina Rosas, and Priyanka Kanth. Alex Quispe and Neha Zaigham pro-
vided excellent research assistance. This note also benefited with inputs of helpful comments from Guadalupe Bedoya, Seth
Garz, Alan Gelb, Harish Natarajan, Robert Palacios, Kanwaljit Singh, Sandip Sukhtankhar, and participants in workshops at
Duke University, Warwick University, UC–Berkeley, the World Bank, IPA, the University of East Anglia, and the American
Economic Association annual meetings.
replaced the previous credentials and authentication mechanism which required them to provide debits cards and
enter PINs.
• In-person collection requirement: Because fingerprints are used for authentication, beneficiaries must come
individually to collect payments; previously many women would send male relatives or community members to
withdraw money on their behalf.
• Increased use of retail payment agents: Most beneficiaries now collect benefits at point-of-sale (PoS) locations
with agents that operate the biometric readers; in the past, many used ATMs, but most of these are not biometrically
enabled, and therefore, not used in the new system.
The impact of the transition to BVS includes both the direct effects of the biometric technology itself, as well as the
effects of changing who collects the money and how.
Main Findings
• Collecting benefits: BVS more than tripled the probability that women collect cash directly (an increase of 46
percentage points).2 Furthermore, it does not appear to reduce access to benefits, including for low-mobility
women.
• Women’s control over cash: For the majority of women who did not personally collect cash under the previous
system, the new system increased their likelihood of deciding how the cash is used (an increase 9 percentage
points).
• Amount of money received: On average, BVS did not significantly change the amount of money beneficiaries
took home. While the switch from debit card to fingerprint authentication increased the amount of cash received
for some beneficiaries, the greater reliance on human retail agents also led to an increase in side payments (by 1.3
percent of the expected payment) that largely canceled out any gains.
• Delivery process: Initially, the new system more than doubled the probability of beneficiaries’ reported difficulties
in withdrawing cash in a single attempt (an increase of 22 percentage points), but this effect dissipated over time.
However, although a majority of beneficiaries still stated they were satisfied with the payment system under BVS,
there was a significant and persistent decrease in satisfaction (from 72 to 54 percentage points).
Policy Implications
• Putting cash directly into the hands of female beneficiaries helps ensure they control how it is used or managed.
Pakistan’s case shows that Biometric verification at the point of delivery can be one mechanism to achieve this,
particularly in contexts where it is not feasible to pay directly into women-owned accounts or ensure direct
collection by women through other means. However, the technology comes with certain risks that should be
carefully weighed against other options.
• Introducing changes to the payment process can create new challenges that must be anticipated and addressed.
Proactive measures are needed to reduce the difficulties and inconveniences beneficiaries face in withdrawing
cash, and ensure that payment processes are designed with their needs and preferences in mind. This could
include providing a choice in payment methods, increasing the availability of cash collection points, and/or paying
benefits directly to agents from bank or mobile money accounts. In all cases, accessible exception handling and
grievance redress mechanisms are needed to address issues and prevent exclusion from benefits due to technology
or process failures.
• Increasing human involvement in the cash delivery process can create new oppor-tunities for rent-seeking.
Changes to the payment process that increase or decrease agent involvement require careful consideration, and
appropriate safeguards and mitigation measures to ensure accessibility and convenience for beneficiaries while
reducing leakage.
2 As described in more detail below, in-person collection by women did not reach 100 percent during the study period as the BVS forms were not
yet fully implemented across districts at this time.
2 id4d.worldbank.org
Introduction
Governments have expanded cash transfer programs to hundreds of millions of recipients in the developing world over
the past decades; these programs continue to increase in their size and scope (World Bank 2018). There is substantial
evidence demonstrating positive impacts of these programs on recipients' well-being.3 However, a key challenge
facing these programs is how to deliver funds effectively to recipients and ensure they reach identified beneficiaries.
Funds may be taken by individuals outside the household (e.g., “leakage” of benefits captured by officials). They may
also be misdirected by individuals inside the household, which is a particular concern for programs that target women.
Although targeting women in cash transfer programs has documented benefits for women's empowerment (Almas
et al. 2018), paradoxically, the funds may not reach the least empowered women if intercepted by male household
members during the delivery process. Additionally, recipients may incur money, time, or hassle costs in accessing funds
that reduce the net benefits they receive, or in extreme cases, prevent them from collecting benefits at all.
Beginning in 2017, Pakistan's Benazir Income Support Program (BISP) transitioned to a new biometric verification
system (BVS) to deliver cash, with the goal of streamlining the payment process and ensuring targeted beneficiaries
received their funds. As in Pakistan, biometric systems have come into increasing use worldwide for a range of
purposes, from finance to health care to social protection.4 To date, however, there is mixed evidence on the impact
of the adoption of biometric verification in delivering public assistance programs,5 and no quantitative studies have
assessed the causal impact of the use of these systems for unconditional cash transfers or their specific use to ensure
delivery directly to women.
Understanding the causal impact of BVS for BISP beneficiaries is important for informing future directions for the
program and for other government programs worldwide. In this study, we use the staggered rollout of the transition
to the BVS system to quantify its effects. While previous evaluations have studied and documented the positive
impact of the BISP cash transfer on beneficiaries’ overall welfare,6 this study is the first to measure the causal impact
of changes in the payment system on intermediate outcomes for women, including control over cash, the amount of
money received, and the overall payment experience.7
This note provides initial results from this analysis and key takeaways for policy makers in Pakistan and other countries
considering the use of similar systems. The complete and detailed results of this study will be published in a forthcoming
working paper by Haseeb, Jalal, Siddiqi, and Vyborny.
The Benazir Income Support Program and BVS
Adoption
BISP was launched in 2008 as Pakistan's signature safety net program, and provides a range of social protection
vision of an unconditional cash transfer, which
services to 7 million recipients across Pakistan. The key service is the pro
is delivered quarterly to women from eligible families. Women in Pakistan face many constraints and often have limited
agency in making major or even minor choices over their own lives: for example, in our beneficiary sample, only around
half of women are allowed to go alone to any nearby destination including a market, clinic, or friend’s home. The BISP
3 See Manley et al. 2020; Bastagli et al. 2019; Millan et al. 2019 for reviews of the evidence.
4 See, for example, Gelb and Clark 2013; Gelb and Metz 2018, Field et al. 2019; Gine, Goldberg, and Yang 2012, Bossuroy, Delavallade, and Pons
2019, and Dhaliwal and Hanna 2017.
5 Muralidharan, Niehaus, and Sukhtankar 2016 found that biometric smartcards used for a workfare program in India delivered a faster, more
predictable, and less corrupt payment process without adversely affecting program access, but Muralidharan, Niehaus, and Sukhtankar 2020
found requiring biometric authentication in the distribution of subsidized food did not reduce leakage, slightly increased transaction costs for the
average beneficiary, and reduced benefits received by the subset of beneficiaries who had not previously registered for an ID.
6 See, for example, Cheema et al. 2014, 2015, 2016b, 2020; Ambler and De Brauw 2017, 2019; Nawaz and Iqbal 2020.
7 A subsequent paper by Haseeb, Jala, Siddiqi, and Vyborny will extend these findings to studying the effect of BVS reform on downstream out-
comes such as schooling, nutrition, and mobility.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
3
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
program was designed to target female heads of the poorest households in the country, giving them more control over
the funds than if they were directed to the male head of household. The program has had considerable documented
success in reducing poverty and increasing women's empowerment. However, challenges remain in reaching intended
beneficiaries, delivering the service efficiently, and preventing leakage of funds. BISP has introduced several reforms
to improve these outcomes. These included the transition from the initial selection of recipients through discretionary
targeting to a Proxy Means Test (Poverty Score Card)–based targeting system, which improved the quality of targeting,
and experiments with different payment methods and locations.8
Pre-BVS Payment Delivery Methods
Since the inception of the program, BISP has made several reforms to improve its approach to payment delivery,
to ensure that the funds actually reach the selected recipients. Initially, the Pakistan post office distributed cash to
recipients. This system faced high levels of leakage and petty corruption, with reports of postmen demanding a portion
of the cash from recipients before they would deliver the funds (Khan and Qutub 2010; Cheema et al. 2014, 2015). For
this reason, the government began transitioning to debit card–based payments in selected districts in 2012. Under this
system, beneficiaries received a card (Figure 1) that could be used at an ATM or a human payment agent known as
a “Point of Service (PoS)”; PoS agents are typically mobile money agents who provide this service as a side activity.
Figure 1. Pre-treatment: Debit cards
Source: World Bank.
While the debit card system is thought to have reduced petty corruption compared to the post office system, it
raised new delivery challenges. For example, beneficiaries frequently lost cards or forgot their personal identification
number (PINs). When this occurred, the card would be blocked and beneficiaries faced a difficult process to get it
unblocked through the BISP office (Cheema et al., 2015). Cards and PINs were also misappropriated from beneficiaries
through theft or fraud. For example, reports suggest that fraudsters were able to purchase BISP cards and PINs from
beneficiaries; offer assistance with withdrawing funds, and then abscond with the card; or steal the card and forge
the beneficiary's signature, in cases where a payment agent's machine was not PIN-enabled and a signature was
used for verification. The debit card-based system also allowed beneficiaries' family members to withdraw funds on
their behalf. In some cases, this may have added convenience; in others, it may have resulted in male family members
intercepting (and controlling) funds intended to be delivered to female recipients.
8 For more information and analysis on these reforms, see Cheema et al. 2015 and Haseeb and Vyborny 2022.
4 id4d.worldbank.org
In summary, neither the post office nor the debit card system fully addressed the concerns of funds being leaked or
intercepted by actors outside or within the target beneficiary's household. In 2016, our baseline year, once the debit
card system had been rolled out across the country, only 20 percent of women in our sample reported that they
collected the transfer themselves. Moreover, 18 percent of beneficiaries reported that they (or the family members or
agents who withdrew money on their behalf) had to unwillingly pay an illegal “fee” or side payment the last time they
received funds (an average of 213 Pakistan rupees [PKR] before BVS). Only 87 percent of beneficiaries received three
or more of the expected four yearly payments, and on average they received only about 74 percent of the total yearly
expected payment amount (approximately 13,906 PKR out of 18,800 PKR) (Cheema et al. 2016a).
Introduction of BVS
In a renewed effort to address these issues, BISP started transitioning to BVS in 2017, equipping payment points
throughout the country with devices to authenticate a beneficiary’s identity by verifying their thumbprint against
the Computerized National Identity Card (CNIC) database before authorizing cash payment. This study focuses on
the transition from the debit card system to this new BVS-based system, which involved a number of changes in the
way that beneficiaries collected their benefits, summarized in Table 1 and described below (Figure 2). Unpacking the
effects of these different components is important to understanding the mechanisms underlying the impact of BVS on
women's control of benefits and the quality of service delivery.
Table 1. Implications of BVS Rollout for BISP
Component Pre-BVS Post-BVS Potential Impact of Change
Confirming Debit card + PIN Biometric • May reduce leakage/fraud to unintended
beneficiary verification using recipients
thumbprint; no
identity • May reduce issues with card loss, PIN errors, and
debit card card blocking
• May increase exclusion if beneficiaries have
difficulty providing fingerprints
Female Not required; Required, • May increase women’s control over cash
beneficiary other family consequence • May increase travel and time costs as women must
member or of biometric
present to come personally
representative verification
receive cash may withdraw for • May lead to exclusion, if some women face social
one or multiple or safety constraints in reaching payment points
recipients
Payment Mixture of ATMs Primarily human • Agents may guide beneficiaries through
point type and human retail retail agents withdrawal, versus ATMs which are confusing for
agents illiterate recipients
and location
• Agents may demand side payments in exchange
for disbursing funds, especially where they face
limited competition
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
5
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
Figure 2. Payment Process under BVS
Source: Karandaaz (undated), reprinted with permission.
Biometric verification replaces debit card/PIN
Payment agents were equipped with biometric readers (Figure 3) that connect with the CNIC database operated by
the National Database & Registration Authority (NADRA) to confirm the beneficiary's identity and authorize the agent
to make the payment. The use of biometric verification may reduce capture of funds by unauthorized individuals, for
example through theft, sale or misuse of the BISP bank card (World Bank 2017). Eliminating debit cards might also
reduce problems with card loss or problems with PIN errors, and debit cards being captured by ATMs or blocked due
to excessive incorrect PIN entries, which were commonly reported under the debit card system (Cheema et al. 2015).
However, it could also cause unintended exclusion or difficulty of withdrawal (Muralidharan et al. 2020), if adequate
exception handling and grievance redress mechanisms are not in place to accommodate beneficiaries who may have
difficulty giving fingerprints (such as elderly persons and manual laborers), or address various technical problems with
the fingerprint readers, applications, or database connections.
Figure 3. Treatment: Biometric Verification
Source: World Bank.
6 id4d.worldbank.org
Beneficiaries must be present at the payment point to withdraw funds
Because fingerprints are used at the point of cash withdrawal, the new system requires that the beneficiary herself
be present. Having women collect benefits directly was one of the major motivations for these reforms, and a major
change from the pre-BVS system in which male family members or agents often picked up cash for beneficiaries (Figure
4). In our baseline data, only about 20 percent of beneficiaries were present when their cash was collected.
BVS could potentially decrease the capture of funds by male household members through multiple mechanisms.
Attending cash withdrawal might also increase women's knowledge about the transfer, including the exact amount due
and how much was withdrawn, increasing their ability to negotiate how the funds are used within the household. In
addition, the simple presence of the female beneficiary during the withdrawal process may also increase the strength
of the messaging to male family members that the transfer is intended for her.
However, requiring beneficiaries to come in person also creates a challenge—each beneficiary must now face the time
and hassle of traveling personally to the payment point. Traveling in public is often challenging for women in Pakistan
due to safety concerns and restrictive gender norms (Field and Vyborny 2016; Sajjad et al. 2018; Field and Vyborny
2020). Under the debit card system, many beneficiaries willingly sent family members or other representatives with
their cards, with a single representative often collecting funds for multiple beneficiaries in a village to save time and
money.
The additional cost or hassle of individual collection could be compounded by limited availability of payment points in
more remote areas where beneficiaries have to travel long distances; pre-BVS, 16 percent of beneficiaries reported that
the person who picked up the payment traveled for over an hour (Cheema et al. 2015). For this reason, it is possible
that some women in inaccessible, unsafe, or socially conservative parts of the country might be unable to access their
payments due to these constraints, resulting in exclusion from the program.
Figure 4. Men in Line to Withdraw BISP Cash Pre-treatment
Source: Pakistan Today newspaper, reprinted with permission.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
7
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
Many recipients switch from ATM to human retail agents
Under the debit card system, beneficiaries were able to withdraw their payments either from an authorized payment
point operated by a human retail agent (PoS), or an ATM. Agents are typically mobile money agents contracted by
telephone operators, who in turn are contracted by BISP's partner banks. Before the BVS transition, about 33.45
percent of the respondents in our sample used ATMs to withdraw the payment. Because only some of these ATMs are
biometrically enabled, 19.4 percent of beneficiaries switched from using an ATM to a human retail agent under BVS.9
The increased reliance on PoS agents could have multiple effects. On the one hand, it may increase their monopoly
power, and thus, their ability to demand side payments (illegal, informal “fees”) in exchange for disbursing cash,
particularly in areas where competition across human retail agents is weak10. On the other hand, agents might also
guide beneficiaries through withdrawal, which may be easier to use for illiterate recipients than ATMs.11
BISP was aware of some of the potential challenges of human retail agents before changing the delivery process, and
worked with its partner banks to try to ensure sufficient coverage of payment points during and after the transition.
Therefore, the transition may have been accompanied by an increase in the number of payment points in some areas.
This component of the rollout could be a factor in affecting outcomes directly, in addition to the introduction of
biometric technology. Future work by the research team will quantify this effect using geolocated payment point data.
Data and Methodology
As shown in Figure 5, the BVS rollout occurred across the country between early 2017 to late 2019; each district
switched to BVS as local administration and banks were ready to do so, based on their administrative preparations.
The endline data was collected in March 2019, when 91 districts were under the BVS system.
Simply comparing areas with and without BVS could lead to biased conclusions about the impact of the new system,
because areas adopting BVS might be systematically different to begin with. Similarly, a simple comparison of outcomes
before versus after BVS adoption could lead to biased conclusions because there could be trends for other reasons
not related to the new system. We address this by using the staggered rollout of BVS reform to study the effects of
the new system using a difference-in-differences analysis.12 This compares the change in outcomes over time for BISP
beneficiaries in areas that adopted BVS later (after the 2019 survey), versus those who adopted BVS earlier (between
2017–2019).
The primary data used in this analysis are panel survey data provided by BISP and Oxford Policy Management (OPM).13
This is a household level panel survey with four rounds spanning from 2013–2019. OPM sampled BISP-eligible and
-ineligible households; we restrict our sample to BISP beneficiary households only. We focus our analysis on the
balanced panel of 3,074 BISP beneficiary households in 63 districts that are surveyed in both 2016 (pre-BVS) and 2019
(during BVS rollout).14
9 Our data shows that the number of active ATMs decreased from 6,166 in 2016 to 379 in 2019, while the number of active PoS agents saw rapid
growth from 1,712 in 2016 to over 10,700 in 2019.
10 In the baseline data, on average, the beneficiaries were 8.16 and 10.03 kilometres away from the nearest ATM and PoS agent, respectively.
11 The beneficiaries did not receive any financial literacy support from BISP.
12 Difference-in-differences compares average change over time in the outcome variable across treatment and control groups. The estimates
require two key assumptions: the parallel trend assumption and the stable unit treatment value assumption. The parallel trend assumption states
that in the absence of treatment the difference between control and treatment groups should be constant over time and the stable unit treat-
ment assumption requires that there should not be any spillovers across treatment and control groups (please see Angrist and Pischke (2008)
for further details).
13 BISP also provided data on BISP recipients' deposits, withdrawals, and payment points from its Management Information System (MIS) that are
being used to analyze how variation in beneficiaries' access to payment points and competition between payment points affect cash delivery
and leakage.
14 Some households dropped out of follow-up rounds of the survey, but dropout from the later round of the survey is not correlated with the BVS
rollout timing.
8 id4d.worldbank.org
Data from multiple rounds of surveys before BVS was implemented allow us to check whether outcomes of interest
were different or trending differently between the early- and late-adopting districts before the intervention. This is not
the case—areas that began the BVS transition early are statistically similar overall and have similar pre-trends to those
who began the BVS transition late. This supports the conclusion that our estimates represent the causal effect of the
“treatment” of the transition from the debit card–based delivery system to the one that uses BVS. 15
It is important to note that within districts that officially switched to BVS during the 2017–2019 period, not all payment
points were able to immediately adopt this technology. Therefore, non-BVS districts in the graphs below represent
those where no payment points had adopted BVS, while BVS districts are those where some or all had made the
transition as of late 2019. In addition, BISP has taken many steps to further improve the functioning of the payment
system after the last round of survey data was collected in 2019 that are not captured in the survey data or analyzed
here. Thus, the results outlined in this note represent the short-term impact of the initiation of the new system in a
district, rather than the longer-term effect once the switch to BVS is 100 percent complete in an area.
Figure 5. BVS Rollout
Mar-2017 Sep-2017 Dec-2017
Last baseline survey: 2016
62 new BVS districts 4 new BVS districts 11 new BVS districts
Mar-2018 Sep-2018 Late 2019-20
Endline survey: Mar-2019
9 new BVS districts 5 new BVS districts 51 new districts
District in panel survey BVS district Non-BVS districts No data
Results
BVS significantly increased the number of women who personally collected their
benefits, and does not reduce access to benefits, including for low-mobility women.
BVS increased the percent of beneficiaries who were present when their cash was collected to over 65 percent (Figure
6). As noted above, this number does not reach 100 percent because not all payment points immediately implemented
the BVS technology as it was rolled out in a district. It is likely that this number has increased significantly since our
endline survey.
15 For transparency, the study team registered a pre-analysis plan for analysis before the Principal Investigators had access to the 2019 wave of
data (EGAP 20200212AB); this specifies all the outcome variables which will be examined. This note includes key outcomes from that pre-analy-
sis plan; a forthcoming working paper will include results for all the outcomes specified.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
9
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
Figure 6. BVS Increased the Number of Women Who Personally Collected Cash
BVS [ *** ] + 46p.p.
Non-BVS
0 20 40 60 80 100
Percentage of women who personally collected the last cash transfer
Note: Number of households (HHs) = 3,074. The graph shows regression coefficients from a difference-in-differences estimate
controlling for household and province-year fixed effects. Standard errors are robust and clustered at the district level. Graph labels
show change in percentage points (p.p.), with asterisks [*] denoting statistical significance (*p < .1, **p < .05, ***p < .01). Data source:
OPM Panel Data.
As noted above, an important concern about requiring women's presence at the payment point is some women might
be effectively excluded from withdrawing their funds due to social or safety constraints limiting their ability to reach
the payment point. In addition, it is possible that certain groups (e.g., elderly, manual laborers, etc.) will face difficulties
with biometric verification that could lead to failed transactions.
To test whether these issues occurred under BVS, we first split the data into two groups: (1) women who were present
for cash withdrawal at baseline under the old system, and (2) those who were not. Beneficiaries in the first category
were used to traveling to a payment point in person, and the primary change under the BVS system was the introduction
of biometrics for authentication, rather than debit cards.16 For the second category, the BVS transition additionally
required them to change their collection behavior, coming in person to receive the cash. As shown in Figure 7, there is
no significant change in women’s ability to collect their benefits successfully, including for women who were not used
to traveling to the payment point. Examining effects for beneficiaries who live more than 16.3 kilometers (km) from the
nearest payment point (the 80th percentile in terms of distance) gives a similar result, with no statistically significant
difference in accessibility post-treatment (Figure 7, Panel B).
Figure 7. BVS Does Not Reduce Access to Benefits, Including for Low-Mobility Women
+1 p.p. +1.3 p.p. +1.8 p.p.
100 +0.5 p.p. 100
Percentage of women who withdrew
90 90
withdrew funds last 12 months
Percentage of women who
80 80
funds last 12 months
70 70
60 60
50 50
40 40
30 30
20 20
10 10
Non-BVS: Non-BVS:
0 BVS: 0 BVS:
Present at Not present Below 80th Above 80th
baseline at baseline percentile percentile
At baseline, present at At baseline, Distance to
payment point? closest Payment Point
16 We do not observe whether the beneficiary travelled alone or in a group or with a chaperone to receive funds from a payment point.
10 id4d.worldbank.org
By requiring beneficiaries to collect cash in person, BVS increased women’s control
over the cash transfer.
To test whether the BVS rollout increased women's control over the BISP cash, we use a self-reported measure from the
OPM survey. Survey enumerators ask BISP beneficiaries to name one person “who in general decides how the money
you personally receive from the BISP cash transfer is spent.” We then examine whether the BVS rollout increases the
probability that the respondent names herself as the main decision-maker.
Among women who were not present for cash withdrawal at baseline, BVS increased control of cash from 65 to 74
percent, a 14 percent (9 percentage point) increase (Figure 8) that is statistically significant at the 99 percent level.
For those beneficiaries who withdrew cash personally at baseline, 80 percent say they decide how to use the funds at
baseline. The BVS transition does not change this, consistent with expectations that the mechanism which increases
control is participating personally in payment withdrawal (rather than the technology itself). Given that 75 percent
of sampled beneficiaries (around 2,296 women) do not withdraw cash personally at baseline, this suggests that over
500,000 female BISP beneficiaries experienced an increase in control over cash due to the reform. Future research will
explore whether this leads to any detectable changes in other indicators of women’s empowerment or on how BISP
funds are spent.
Figure 8. For Women Who Previously Did Not Withdraw in Person, BVS Increased Control over Cash
100
- 1.6 p.p
+ 9.3 p.p
[ *** ]
80
Percentage of women who
decide use of cash
60
40
20
Non-BVS:
0
BVS:
Present at baseline Not present at baseline
At baseline, present at payment point?
Note: Number of HHs—collected personally at baseline = 778. Number of HHs—did not collect personally at baseline = 2,296. The
graph shows regression coefficients from a difference-in-differences estimate controlling for household and province-year fixed
effects. Standard errors are robust and clustered at the district level. Graph labels show change in percentage points (p.p.), with
asterisks [*] denoting statistical significance (*p < .1, **p < .05, ***p < .01). Data source: OPM Panel Data.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
11
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
On average, BVS did not significantly change the net amount of money that women
take home—for some, it slightly increased the amount of cash received; however,
increased reliance on human retail agents also led to a small increase in side
payments that canceled out these benefits.
Figure 9 shows the overall impact of the BVS transition on the funds received by beneficiaries over the last year, the
amount spent on travel to collect the payment, and side payments or illegal fees paid during collection17. For the
average beneficiary, the BVS transition does not lead to a statistically detectable change in net amount received,
although the point estimate is positive. BVS also does not change the reported travel costs that each beneficiary or
her representative incurred in collecting the payment.18 Together, these findings provide evidence that the system is
not imposing significant additional costs on beneficiaries to access the full funds to which they are entitled.
Figure 9. On Average, BVS Does Not Change the Net Amount of
Cash Received by Beneficiaries
15,000
+ PKR 513 (4%) + PKR 279 (2.3%)
(USD 2.94) (USD 1.60)
Total amount per year (PKR)
10,000
5,000
+ PKR 257 (130%) + PKR 69 (12%)
(USD 1.47) (USD 0.22)
[ *** ]
Non-BVS:
BVS:
0
Side payments Total amount net of Travel costs Total amount net of
side payments side payments and
travel costs
Note: Number of HHs = 3,074. The graph shows regression coefficients from a difference-in-differences estimate controlling
for household and province-year fixed effects. For reference, BISP households on average have a per-adult equivalent monthly
consumption expenditure of PKR 3,064 or ~US$18.5 (OPM 2016 report). The asterisks [*] denote statistical significance (*p < .1, **p <
.05, ***p < .01). Absence of asterisks implies that we found no statistically detectable effect. Data source: OPM Panel Data.
At the same time, we see that BVS does lead to a small but statistically significant increase in reported side payments
(an increase of about 257 PKR per year, or 1.3 percent of the payment amount—Figure 9). Overall, it appears this was
primarily caused by the increased reliance on human retail agents under the BVS system. In Figure 10, for example,
we can see that nearly half of the respondents in BVS districts in the endline report pay a side payment, and nearly all
are to retail agents. BISP has worked to address this issue, but there are limitations on their capacity to monitor PoS
agents in the field.19
17 USD amounts in Figure 9 are according to the exchange rate of February 8, 2022.
18 Note that there could still be an increase in travel costs due to each beneficiary having to travel separately, rather than a group of beneficiaries
sending a single representative with all their cards; this is not captured in these data.
19 For example, there are thousands of agents operating across the country and a short period of time in between payment tranches. In addition,
there is a lack of a direct accountability between BISP and the agents, as they are contracted by telephone companies, who are themselves con-
tracted by the banks that BISP has contracted to implement the payment system.
12 id4d.worldbank.org
Figure 10: Descriptive Statistics: Most Side Payments Go to Retail Agents
Note:
- 19% of baseline households (n=548) report having to pay side payments Baseline (2016)
- 48% of endline households (n=1,354) report having to pay side payments Endline (2019)
50 47.5
40
% of households
30
20
9.96
10
8.96 8.42
5.47 5.27 6.65 5.08
3.78 2.24 2.8 2.52 2.99
.28 .14 .56 .421 .095 .665 .76 .428 .38 .76 1.38
0
Pakistan Bank sta Security Retail Other Other Pakistan Bank sta Security Retail Other Other
postman guard agent government postman guard agent government
o cial o cial
Note: Number of HHS = 3,074. DataNon-BVS
source: OPM Panel data. BVS
Splitting the sample into two groups—those who previously collected payments from a human agent, and those
who previously collected from an ATM and switched to a human agent—helps unpack these results. Although both
groups experienced a small but statistically significant increase in side payments, this was larger for beneficiaries who
switched from an ATM to a human agent, as shown in Figure 11. Furthermore, beneficiaries who already used human
retail agents— i.e., those for whom the intervention only added biometric verification—experienced a statistically
significant increase in the net amount of money received net of this side payment (826 PKR per year).
Figure 11. Payment Effects Differ for Areas Using Human Retail Agent Versus ATM at Baseline
+ PKR 826
(USD 4.74)
14,000 14,000 + PKR 135
Amount paid in side paymentsin last year (PKR)
[ *** ]
(USD 0.77)
Amount net of side payments received in
12,000 12,000
10,000 10,000
last year (PKR)
8,000 8,000
6,000 6,000
4,000 4,000
2,000 + PKR 323 + PKR 235 2,000
(USD 1.85) (USD 1.35)
[ *** ] [ *** ] Non-BVS:
BVS:
0 0
Retail Agent Retail Agent Retail Agent Retail Agent
unavailable available unavailable available
At baseline, Retail Agents are At baseline, Retail Agents are
available in enumeration area? available in enumeration area?
Note: Number of HHs without retail agents in enumeration area at baseline = 863. Number of HHs with retail agents in enumeration
area at baseline = 2,211. The graph shows regression coefficients from a difference-in-differences estimate controlling for household
and province-year fixed effects. Standard errors are robust and clustered at the district level. The asterisks [*] denote statistical
significance (*p < .1, **p < .05, ***p < .01). Data source: OPM Panel Data.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
13
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
In addition, for beneficiaries where the reform added only biometric verification, there is a small but significant increase
in the proportion of beneficiaries who receive any amount of money (Figure 12). The reason for this effect of biometric
verification on the receipt of money is unclear; it could occur because some beneficiaries whose cards were blocked,
stolen, or lost under the debit card system were able to access funds again after the BVS transition obviated the debit
card system. More research is needed to explore this finding further.
These findings suggest that increased reliance on human retail agents added opportunities for leakage, which canceled
out the small increase in cash received as a result of switching from debit card authentication to biometric verification.
Figure 12. In Areas Using Retail Agents at Baseline, BVS Led to a Small Increase in
Probability of Receiving Any Funds
+ 2.6 p.p.
[ *** ]
100 -0.6 p.p.
withdrew funds last 12 months
Percentage of women who
80
60
40
20
Non-BVS:
0 BVS:
Retail Agent Retail Agent
unavailable available
At baseline, Retail Agents are avaliable in enumeration area?
Note: Number of HHs without retail agents in enumeration area = 863. Number of HHs with retail agents in enumeration area = 2,211.
The graph shows regression coefficients from a difference-in-differences estimate controlling for household and province-year fixed
effects. Standard errors are robust and clustered at the district level. Graph labels show change in percentage points (p.p.), with
asterisks [*] denoting statistical significance (*p < .1, **p < .05, ***p < .01). Data source: OPM Panel Data.
BVS increased some challenges collecting funds in the short term and decreased
eficiary satisfaction for women newly collecting payments in person.
ben
Before the BVS transition, around 18 percent of recipients reported they had to make multiple attempts to withdraw
the last transfer they received; the transition leads to an average 10 percentage point increase in this number. However,
this increase seems to disappear over time (Figure 13), suggesting it was due to temporary challenges with the new
system. While these issues are important in terms of beneficiary experience and hassle, the above results demonstrate
that they did not cause an overall reduction in beneficiaries’ ultimate ability to collect their benefits.
Beneficiaries may fail to withdraw cash on the first attempt for a number of reasons, some of which may be related
to the BVS technology (such as a failure in fingerprint verification, device malfunction, or loss of connection with the
central system), and others to the switch in payment modalities (such as congestion of recipients at the payment point
or agents running out of cash). Approximately 2.8 percent of respondents reporting payments under the BVS system
reported that they had problems directly related to the biometric reader the last time they withdrew cash; 9.5 percent
reported having to return due to a long queue, which could itself occur because of technology failures, or because of
delays as payment agents or recipients learn how to use the new system, particularly in the early phases of rollout.20
20 The mean failure rates should be interpreted with caution when comparing them to failure rates reported from other contexts: some recipients
may report a problem with a long queue, which is itself due to technology issues; in addition, the survey question asks about “the last time you
withdrew cash,” which might lead enumerators or respondents to exclude occasions when the respondent did not successfully withdraw funds at
all due to technology failures.
14 id4d.worldbank.org
These effects are larger for older beneficiaries (Figure 14), which could be the result of multiple factors such as eroded
fingerprints or other difficulties with the technology.
Figure 13. BVS Increased the Percentage of Beneficiaries Who Had to Make More Than One Attempt
to Collect Payment, But These Effects Dissipate Over Time
100
90
Percentage of women who made more than
80
one attempt to collect payment
70
60 + 22 p.p.
[ *** ]
50
40
+ 5.8 p.p.
30
20
10
0
5 or fewer More than 5
Number of tranches since treatment
Note: The graph shows regression coefficients from a difference-in-differences estimate controlling for household and province-year
fixed effects. Standard errors are robust and clustered at the district level. The asterisks [*] denote statistical significance (*p < .1, **p
< .05, ***p < .01). Data source: OPM Panel Data.
Figure 14. Older Beneficiaries Were More Likely to Report Multiple Attempts
100
Percentage of women who made more than
one attempt to collect payment
80
60
+ 11.8 p.p.
[ *** ]
40 + 7.4 p.p.
20
0
Below median age Above median age
At baseline, beneficiary is above median Age?
Note: The graph shows regression coefficients from a difference-in-differences estimate controlling for household and province-year
fixed effects. Standard errors are robust and clustered at the district level. The asterisks [*] denote statistical significance (*p < .1, **p
< .05, ***p < .01). Effects are significantly different between groups (p < 0.1). Data source: OPM Panel Data.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
15
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
Overall, beneficiaries’ reported high levels of satisfaction with the BISP payment method, with most women saying
they are “very satisfied”; however, BVS decreased the proportion of women who reported they are very satisfied from
72 to 54 percent, on average. This effect does not disappear after several payment tranches, suggesting that it may
not be entirely driven by temporary difficulties with the new technology, which were eliminated over time (see Figure
13). We might expect that beneficiaries who experience increased control over BISP cash would improve satisfaction
and thus mitigate this effect; however, this effect is no smaller for beneficiaries who did not control cash at baseline.
Having to collect payments in person could itself reduce satisfaction, due to the hassle, cost or insecurity of travel.
The estimated effect on satisfaction appears somewhat larger for women who were collecting funds in person for the
first time than those who had collected funds in person before; however, this difference is not significantly different,
so these results are inconclusive. Importantly, the satisfaction question was asked of the beneficiary, whether or not
she was the one to collect the funds; because the BVS transition led many women to withdraw funds personally who
otherwise would not travel to the payment point, the reform may have simply led to greater beneficiary awareness of
the hassles involved, which previously were only observed by a male family member. Future research will explore this
in greater detail by exploring how these effects differ depending on the accessibility of nearby payment points.
Policy Implications and Future Research
The results of the study have three important implications for policy makers considering the adoption of similar
technologies and processes for delivering public assistance programs, and particularly those intended to reach low-
mobility women.
First, putting cash directly into women’s hands can help ensure that they control
how it is used.
For programs like BISP that are intended to reach the most marginalized women, delivering funds directly to them
can intensify the empowerment benefits of transfer. When benefits have frequently collected by male members of
households, this study demonstrates that using biometric verification at the point of delivery is one method to ensure
collection by women because the nature of the technology requires the presence of a specific individual.
At the same time, the use of biometric verification at the point of transaction requires careful consideration given
the cost, as well as potential data protection concerns and exclusion risks, and may not be appropriate or advisable
depending on the context.21 For example, it may be possible to increase women's agency over cash benefits by making
payments directly into bank or mobile money accounts that they directly access and control; however, constraints on
women's mobility combined with limited coverage of financial services may also make these options infeasible in some
settings.
Second, changes to the payment process—particularly changing locations
or mandating in-person collection when people could previously send
representatives—can create new challenges for beneficiaries that must be
anticipated and addressed.
Any changes to the payment process should be carefully considered to avoid introducing new barriers or hassles and
ensure they align with beneficiary’s needs and preferences. If biometric verification or any digital technologies are
21 For a deeper discussion of the potential benefits and risks of biometric technology, see the ID4D Practitioner's Guide (http://id4d.worldbank.org/
guide) and the forthcoming Biometrics Handbook.
16 id4d.worldbank.org
used at the point of service delivery, appropriate exception handling and grievance redress mechanisms must be in
place to avoid denial of services due to a technical failure.
Increasing payment point coverage could address the hassle of required in-person collection. Another potential
approach could be increasing the number of mobile agents who visit beneficiaries in their villages, reducing the hassle
they may experience in traveling. Karandaaz, a leading organization working on payment systems in Pakistan, has
advocated such an approach (Karandaaz 2020). Anecdotally, some payment agents have taken the initiative to do this
themselves; however, this is not a formal policy and GPS "ring-fencing" controls on payment devices are designed to
reduce device mobility to combat fraud.
Since the period of data covered in this study, BISP has already taken multiple steps to help to combat this issue,
including expanding the number and coverage of human payment points, enhancing monitoring procedures, and
providing an alternative payment approach through the bank branch in cases of fingerprint failure. In addition, BISP has
worked with NADRA to update the fingerprint records of beneficiaries who had problems with biometric verification.
Third, increasing human involvement in the delivery chain may have unintended
sequences, creating new opportunities for leakage.
con
For BISP areas that switched from ATMs to retail agents who operate biometric readers, these new agents appear to
have increased demands for side payments from beneficiaries.22 Policies and strategies must be in place to help reduce
or offset incentives and opportunities for rent seeking among payment agents. This could include increasing cov erage
and competition between payment agents to reduce their relative power, and implementing top-down and/or bottom-
up safeguards and sanctions, such as audits of payment points, awareness raising campaigns, and social accountability
measures. The effectiveness of such strategies needs to be carefully tested and evaluated.23
The results of study show that the adoption of the BVS by the BISP cash transfer program had clear benefits in terms
of women’s control over resources. At the same time, it also had challenges that BISP has been actively working
to address since the completion of the study baseline. Further research on BISP's ongoing efforts to improve the
payment process will provide additional lessons for other counties seeking to design effective and inclusive delivery
systems for social benefits.
22 It is possible that increased competition between payment agents can limit their ability to demand side payments. We are evaluating this in
on-going work.
23 Another option that has been discussed would be to introduce more biometric ATMs, which could confirm the identity of the beneficiary without
introducing human retail agents who may be able to demand side payments. However, switching from human payment agents to biometric ATMs
could also have unintended consequences. In the case of Pakistan, many BISP recipients face difficulties when withdrawing from ATMs without
the support of an agent because most are illiterate and do not know how to use ATMs. In addition, if the biometric verification fails on a given
attempt, the beneficiary may have no source of information on how to proceed. In addition, biometric ATMs would require substantial new infra-
structure at payment points, which could be costly, particularly in countries with low connectivity and sparse ATM coverage to begin with.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
17
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN
References
Almas, I., Armand, A., Attanasio, O., and Carneiro, P. 2018. Measuring and Changing Control: Women's Empowerment
and Targeted Transfers. Economic Journal, 128 (612), F609–F639.
Ambler, K., and De Brauw, A. 2017. The Impacts of Cash Transfers on Women's Empowerment. (1702).
Ambler, K., and De Brauw, A. 2019. Household Labor Supply and Social Protection: Evidence from Pakistan's BISP
Cash Transfer Program. IFPRI Discussion Paper (March).
Angrist, J.D. and Pischke, J.S., 2008. Mostly Harmless Econometrics. Princeton University Press.
Bastagli, F., Hagen-Zanker, J., Harman, L., Barca, V., Sturge, G., and Schmidt, T. 2019. The Impact of Cash Transfers: A
review of the evidence from low- and middle-income countries. Journal of Social Policy, 48 (3), 569–594.
Bossuroy, T., Delavallade, C. and Pons, V. 2019. Biometric Tracking, Healthcare Provision, and Data Quality:
Experimental Evidence from Tuberculosis Control. NBER Working Paper 26388, p. No. 26388.
Cheema, I., Farhat, M., Binci, M., Asmat, R., Javeed, S., and O'Leary, S. 2020. Benazir Income Support Programme:
Evaluation Report. Tech. Rep. March.
Cheema, I., Farhat, M., Hunt, S., Javeed, S., Keck, K. and O'Leary, S. 2015. Benazir Income Support Programme: Second
Impact Evaluation Report. Oxford Policy Management report (December).
Cheema, I., Farhat, M., Hunt, S., Javeed, S., Pellerano, L., and O'Leary, S. 2014. Benazir Income Support Programme:
First follow-up impact evaluation report. Oxford Policy Management report.
Cheema, I., Farhat, M., Hunt, S., Javeed, S., Pellerano, L., and O'Leary, S. 2016a. Benazir Income Support Programme:
Final Impact Evaluation Report. Oxford Policy Management report (June).
Cheema, I., Hunt, S., Javeed, S., Lone, T., and O'Leary, S. 2016b. Benazir Income Support Programme: Final
Impact Evaluation Report. Oxford Policy management report (June).
Dhaliwal, I., and Hanna, R. 2017. The devil is in the details: The successes and limitations of bureaucratic reform in
India. Journal of Development Economics, 124, 1–21.
Field, E., and Vyborny, K. 2016. Female Labor Force Participation: Pakistan Country Study for the Asian Development
Bank.
Field, E., and Vyborny, K. 2020. Women's mobility and labor supply: Experimental evidence from Pakistan. ADB
Working Paper (1629317), 1–20.
Field, E. M., Pande, R., Rigol, N., Schaner, S. G., and Moore, C. T. 2019. On her own account: How strengthening
women's financial control affects labor supply and gender norms. NBER Working Paper 26294.
Gelb, A., and Clark, J. 2013. Identification for Development: The Biometrics Revolution. SSRN Electronic Journal
(January 2013).
Gelb, A., and Metz, A. D. 2018. Identification Revolution: Can Digital ID Be Harnessed for Development? Center for
Global Development.
Gine, X., Goldberg, J., and Yang, D. 2012. Credit Market Consequences of Improved Personal Identification: Field
Experimental Evidence from Malawi. American Economic Review, 102 (6), 2923–2954.
Haseeb, M., and Vyborny, K. 2022. Data, discretion and institutional capacity: Evidence from cash transfers in
Pakistan. Journal of Public Economics.
18 id4d.worldbank.org
Karandaaz. Undated. A human centered design study on biometric cash withdrawal system for BISP beneficiaries.
Karandaaz. 2020. Industry Note: COVID-19 and Its Impact on Pakistan's Digital Financial Services Landscape.
(January), 1–6.
Khan, S. N., and Qutub, S. 2010. The Benazir Income Support programme and the Zakat programme: A political
economy analysis of gender. Working Paper, Overseas Development Institute.
Manley, J., Balarajan, Y., Malm, S., Harman, L., Owens, J., Murthy, S., Stewart, D., Winder-Rossi, N. E., and Khurshid, A.
2020. Cash transfers and child nutritional outcomes: A systematic review and meta analysis.
Millan, T. M., Barham, T., Macours, K., Maiuccio, J. A., and Stampini, M. 2019. Long-term impacts of conditional cash
transfers: Review of the evidence. World Bank Research Observer, 34 (1), 119–159.
Muralidharan, K., Niehaus, P., and Sukhtankar, S. 2016. Building state capacity: Evidence from biometric smartcards in
India. American Economic Review, 106 (10), 2895–2929.
Muralidharan, K., Niehaus, P., and Sukhtankar, S. 2020. Identity verification standards in welfare programs:
Experimental evidence from India. Working paper.
Nawaz, S., and Iqbal, N. 2020. The impact of unconditional cash transfer on fuel choices among ultra-poor in
Pakistan: Quasi-experimental evidence from the Benazir Income Support Program. Energy Policy.
Sajjad, F., Field, E., Anjum, G. A., and Vyborny, K. 2018. Overcoming Barriers to Women's Mobility: Improving
Women's Access to Public Transport in Pakistan. GLM—LIC Policy Brief No. 24 (24).
World Bank. 2017. Pakistan national social protection program: Environmental and Social Systems Assessment Final
Report.
World Bank. 2018. The State of Social Safety Nets 2018.
About ID4D
The World Bank Group’s Identification for Development (ID4D) Initiative uses global knowledge and expertise
across sectors to help countries realize the transformational potential of digital identification systems to achieve the
Sustainable Development Goals. It operates across the World Bank Group with global practices and units working on
digital development, social protection, health, financial inclusion, governance, gender, and legal, among others.
The mission of ID4D is to enable all people to access services and exercise their rights by increasing the number of
people who have an official form of identification. ID4D makes this happen through its three pillars of work: thought
leadership and analytics to generate evidence and fill knowledge gaps; global platforms and convening to amplify
good practices, collaborate, and raise awareness; and country and regional engagement to provide financial and
technical assistance for the implementation of robust, inclusive, and responsible digital identification systems that are
integrated with civil registration.
The work of ID4D is made possible with support from the World Bank Group, Bill & Melinda Gates Foundation, Omidyar
Network, the UK government, the French government, and the Norwegian Agency for Development Cooperation
(Norad).
To find out more about ID4D, visit id4d.worldbank.org. To participate in the conversation on social media, use the
hashtag #ID4D.
USING BIOMETRICS TO DELIVER CASH PAYMENTS TO WOMEN:
19
EARLY RESULTS FROM AN IMPACT EVALUATION IN PAKISTAN