Content uploaded by Ashish Sedai
Author content
All content in this area was uploaded by Ashish Sedai on Oct 15, 2024
Content may be subject to copyright.
Digital Finance and Gendered Labor Reallocation
Rikhia Bhukta∗Sandhya Garg†Ashish K. Sedai‡
Abstract
This study examines how digital financial services (DFS) drive labor reallocation in
India, leveraging nationally representative data and a two-stage least squares approach.
For men, DFS facilitates a shift from agricultural self-employment to formal wage em-
ployment. For women, it reduces unpaid labor, promoting entry into formal work and
increasing financial autonomy through expanded access to banking, mobile money, and
the internet. Both genders experience increased time spent in paid work and earnings.
Our findings demonstrate the critical role of DFS in transforming labor markets and
advancing gender equity.
JEL classifications: G21, J16, J22, O33, O16
Keywords: Digital payments, Gender, Technology, Labor, Time-use
∗Department of Economics, Indian Institute of Technology, Kanpur
†HDFC Chair of Banking and Finance, Institute of Economic Growth, Delhi
‡Department of Economics, College of Business, University of Texas at Arlington. Email:
ashish.sedai@uta.edu
1 Introduction
In 2023, the global digital payments market has reached an estimated $8.49 trillion and is
projected to exceed $15 trillion by 2027, driven by the rapid adoption of mobile payments,
digital banking and Fintech innovations, particularly in developing countries (Statista,2023;
Insights,2023). This surge in Digital Financial Services (DFS) has significantly trans-
formed economic transactions by lowering transaction costs, expanding financial inclusion,
and enhancing access to credit and consumption. Such developments have proven pivotal
in driving economic growth and alleviating poverty across diverse global contexts (Jack
and Suri,2014;Suri and Jack,2016;Demirguc-Kunt et al.,2018;Higgins,2019;De Mel
et al.,2022;Abiona and Koppensteiner,2022;Shapiro and Mandelman,2021;Agarwal
et al.,2024;Cong, Giesecke, and Kuhnen,2024). Although the macroeconomic impacts
of DFS, such as boosting GDP and reducing economic disparities, are well documented,
their microeconomic implications, particularly concerning labor market dynamics through
a gender-focused lens, require further exploration. This study examines how DFS facilitate
transitions from informal to formal employment and influence the time allocation between
labor and non-labor activities, with a focus on gender disparities.
This research investigates the impact of DFS on employment patterns in India, a country
where the rapid expansion of DFS, coupled with challenges in labor market transitions and
pronounced gender disparities, offers a distinct, yet broadly applicable context for analysis
relevant to other major developing economies.
We first evaluate whether DFS can trigger a structural shift from informal subsistence
activities to formal employment sectors. We investigate whether the mechanisms through
which DFS affect labor market outcomes are differentiated by gender, a critical consideration
in regions marked by pronounced gender inequalities (Kundu,2020;Jayachandran,2015;
Bertrand et al.,2021;Field et al.,2021). We investigate DFS’s impact at the intensive
margins, i.e., on time spent on unpaid and paid work. Additionally, we examine how the
expansion of DFS influences enablers of digital economy, such as access to banking, internet,
and mobile money, specifically women’s financial autonomy.
We hypothesize that the proliferation of DFS facilitates an economic transformation, with
different mechanisms that affect men and women. Increased access to DFS is expected
to reduce agricultural employment for both genders, a transition supported by structural
change theories, which posit that as developing economies modernize, labor shifts from sub-
sistence agriculture to more productive sectors, including non-agricultural self-employment
1
and wage labor (Lewis et al.,1954;Herrendorf, Rogerson, and Valentinyi,2014;Boppart
and Krusell,2020). DFS, by reducing transaction costs and broadening access to credit,
is likely to accelerate this transition by enabling more individuals to access formal credit
markets and improving labor productivity in non-agricultural sectors (Suri and Jack,2016).
For women, DFS presents particularly transformative opportunities. Structural barriers,
including unpaid domestic responsibilities and limited access to financial resources, have
historically restricted women’s participation in formal labor markets (Duflo,2012;Dupas
and Robinson,2013;Barth, Kerr, and Olivetti,2021;Exley and Kessler,2022). By pro-
viding access to essential financial services, DFS can remove these barriers, opening new
pathways to paid and formal employment. This shift is anticipated to significantly enhance
women’s participation in the formal economy, helping them transition away from unpaid
and unrecognized labor, thereby reshaping traditional gender roles. However, the magni-
tude of this change may be limited by the deep-rooted structural challenges women face in
entering formal labor markets (Anderson and Eswaran,2009;Duflo,2012;Bertrand et al.,
2021).
To substantiate these hypotheses, the study employs nationally representative data from
the Periodic Labor Force Survey (PLFS), Indian Time Use Survey (ITUS), and the National
Family Health Survey (NFHS). These datasets provide insights into labor market participa-
tion, time-use patterns, and financial tool utilization, disaggregated by gender, which allows
for an exploration of both the demand- and supply-side effects of DFS on labor markets.
The empirical strategy incorporates a two-stage least squares (2SLS) methodology to control
for potential endogeneity in DFS adoption, using district-level exposure to ’currency chest’
infrastructure as an instrument, following methodologies outlined by Chodorow-Reich et al.
(2020), Crouzet, Gupta, and Mezzanotti (2023), and Das et al. (2023). This approach helps
to isolate the exogenous variation in DFS access driven by historical banking infrastructure
rather than by current economic conditions.
Our findings at the extensive margins offer insights into how DFS influence both the struc-
ture of employment and the allocation of labor across genders. For men, we observe a
statistically significant increase in aggregate employment, primarily driven by a reduction
in agricultural self-employment, while non-agricultural employment (both self-employment
and regular wage/salary employment) increases. For women, the effects are distinct; we do
not find a reduction in aggregate self-employment, but do observe a notable increase in reg-
ular wage/salary employment, majorly driven by a shift away from unpaid self-employment
and casual employment. This suggests that DFS promote a shift from low-productivity
2
agricultural work to higher productivity non-agricultural employment for men, securing
wage-based jobs, and for women, a transition from unpaid roles to formal, wage-earning
employment.
At the intensive margin, DFS significantly increased the time men allocate to both self-
employment and regular wage/salary employment. For women, the impact of DFS is more
pronounced in regular wage/salary employment than in self-employment, reflecting similar
trends observed at the extensive margin. This dual effect indicates that while DFS facilitate
deeper integration into formal labor markets, they also lead to greater involvement in paid
labor market activities. Importantly, the increase in time spent on employment for both
men and women is largely driven by a reduction in time allocated to unpaid care work
and self-care activities, suggesting a shift towards more economically productive endeavors.
Moreover, our analysis reveals significant wage gains across both self-employed and regu-
larly employed individuals. Crucially, the absence of statistically significant differences in
wages between genders suggests that DFS promote wage growth in an equitable manner,
contributing to narrowing the relative gender wage gaps.
Furthermore, the socio-economic transformations induced by DFS are profound, affecting
individual-level outcomes particularly in terms of financial inclusion and autonomy. DFS
significantly increase the ownership of bank accounts, mobile phones, and internet usage,
facilitating greater engagement in the digital economy across genders. For women, in par-
ticular, these technologies have led to a marked increase in financial autonomy—greater
control over financial resources and decision-making within households. This empowerment
of individuals—especially women—represents a broader reallocation of economic power,
wherein DFS not only reshape employment dynamics but also reduce gender disparities in
financial access and decision-making authority.
Disaggregated by gender, age, education and marital status, our results show marked differ-
ences in employment engagement. Younger, more educated individuals, particularly those
adaptable to digital integration, show increased likelihood of paid employment, underscor-
ing the role of digital literacy in harnessing DFS benefits. Unmarried individuals and those
from more privileged social categories also exhibit stronger responses to DFS, highlighting
socio-economic factors as significant mediators of DFS effects. In urban settings, DFS en-
hanced regular wage employment for both men and women, demonstrating more effective
integration of digital services in urban compared to rural areas. Regarding social cate-
gories, DFS impacts vary. Notably, there is a decrease in self-employment among men from
Scheduled Tribes and Scheduled Castes, indicating shifts from traditional livelihoods.
3
The robustness of our findings is ensured through various empirical tests. First, we com-
pare Instrumental Variable (IV) estimates with Ordinary Least Squares (OLS) results to
affirm the direction of the observed effects of DFS on employment. Second, we control for
district-level characteristics that significantly affect the instrumental variable. Third, we
use alternative specifications of DFS exposure—using both transaction (a) counts and (b)
growth of DFS from 2018 to 2019—are tested to confirm the stability of our results.
2 Literature Review
The rapid growth of DFS has become a pivotal area of economic research, particularly for
its potential to enhance financial inclusion, lower transaction costs, and stimulate economic
development in developing nations. Foundational studies by Suri and Jack (2016), Mbiti
and Weil (2015), and Jack and Suri (2014) demonstrate how mobile money platforms such
as M-Pesa have significantly aided economic growth by providing financial services to un-
banked populations, easing credit constraints, and enabling consumption smoothing. These
contributions highlight DFS’s role in poverty alleviation and economic empowerment. Ex-
panding on this, Demirguc-Kunt et al. (2018) and Shapiro and Mandelman (2021) explore
DFS’s potential to mitigate poverty, hunger, and gender inequality and to induce sectoral
labor market shifts globally.
Further, randomized controlled trials (RCTs) provide detailed insights into the demand-side
effects of DFS on financial inclusion. Studies by Dupas and Robinson (2013) and Field et al.
(2021) highlight how access to savings accounts substantially increases women’s financial
autonomy and labor market participation in Kenya and India, respectively. Riley (2018)
illustrates that mobile money enhances risk-sharing and aids in consumption smoothing dur-
ing economic shocks in Tanzania, with pronounced benefits for women. De Mel et al. (2022)
and Cole, Joshi, and Schoar (2024) show that mobile-linked accounts and targeted business
advice significantly improve savings behaviors and entrepreneurial outcomes. While these
RCTs emphasize the benefits of DFS, they tend to have limited geographical scope and
often overlook important factors like time use and labor reallocation.
Prior research has examined the broader impacts of DFS on various economic outcomes.
Studies have addressed savings and insurance (De Mel et al.,2022), liquidity and consump-
tion smoothing (Agarwal et al.,2024;Abiona and Koppensteiner,2022), risk-sharing (Carli
and Uras,2024;Dizon, Gong, and Jones,2020), migration responses and resilience to eco-
nomic shocks (Chen and Zhao,2021;Riley,2018;Suri, Bharadwaj, and Jack,2021;Londo˜no-
4
V´elez and Querubin,2022;Batista and Vicente,2023), intergenerational mobility (Yang
et al.,2024;Abiona and Koppensteiner,2022), agricultural growth (Aker,2010;Aggarwal,
Brailovskaya, and Robinson,2023), remittances (Pazarbasioglu et al.,2020;Moorena et al.,
2020), technology adoption (Crouzet, Gupta, and Mezzanotti,2023), entrepreneurship and
economic development (Dubey and Purnanandam,2023;Beck et al.,2018;Cole, Joshi, and
Schoar,2024), and tax returns (Das et al.,2023). However, these studies largely overlook the
micro-level causal implications for labor markets, particularly in gender-unequal settings.
Our study seeks to fill these gaps by providing a micro-level analysis of the gender-differentiated
impacts of DFS on labor supply. We examine how DFS reshape labor allocation within
households, leading to structural shifts from informal to formal employment. For instance,
while Agarwal et al. (2024) focus on the effects of India’s 2016 demonetization on acceler-
ating digital payments, our research extends this analysis by exploring how DFS facilitate
formal employment entry across genders, contributing to the creation of higher-paying job
opportunities. We also build on Shapiro and Mandelman (2021)’s findings that firm-level
digital adoption reduces self-employment by facilitating transitions into salaried jobs. We
highlight a key distinction: this reduction primarily reflects a shift from low-productivity
agricultural work to more productive non-agricultural activities, offering a deeper under-
standing of how DFS promote productivity-enhancing labor reallocations. By addressing
both the extensive and intensive margins of employment and focusing on intra-household
labor dynamics, our analysis provides new insights into how DFS drive inclusive economic
growth and mitigate gender disparities in labor markets.
3 Data
For our analysis, we use five datasets detailed below.
3.1 Indian Time Use Survey (ITUS)
Time-use surveys offer a framework for gauging how much time people spend on various
activities, whether they are paid or unpaid activities. During January-December 2019,
National Statistical Office (NSO) conducted the first pan-India Time Use Survey. ITUS
covered 1,38,799 households, among which 82,897 were from rural areas and 55,902 were
from urban areas. Time use patterns of each household member aged 6 and above were
collected, which aggregated to a total of 4,47,250 individuals.
The gender-disaggregated data on time use patterns in ITUS allow us to study time spent on
5
employment activities by men and women separately. We analyze time spent on all employ-
ment activities, regular employment, self-employment and casual employment, alongside
unpaid activities, learning and socializing activities and self-care. Our sample is restricted
to individuals above age 14. In Table A1, we present a detailed summary statistics of the
intensive margin (conditional on positive time spent in the said activity) of time use vari-
ables, for both men and women. In Table A2, we present another set of time-use statistics,
which we use for additional results and robustness checks.
3.2 Periodic Labor Force Survey (PLFS)
PLFS provides data on the employment and unemployment situation in India. The survey
is conducted annually by the National Statistical Office (NSO) since 2017. For our analysis,
we used the 2019-20 PLFS wave, conducted between July 2019 and June 2020. The pan-
India survey covered 6,913 villages and 5,656 urban blocks, including a total of 4,18,297
individuals. We use gender-disaggregated data on the extensive margin of employment ac-
tivities, including self-employment (disaggregated by agricultural and non-agricultural em-
ployment), unpaid self-employment (which is not a part of the aggregate self-employment),
regular wage/salary employment, and casual employment. We also use information on indi-
vidual wages in regular employment and self-employment.1To make our analysis consistent
with ITUS, we restrict the PLFS sample to individuals over the age of 14. In Table A1, we
present the summary statistics of the variables we use from PLFS.
3.3 National Family Health Survey (NFHS)
NFHS is a nationally representative household survey that captures a wide range of topics
on health and empowerment. For our analysis, we focus on the fifth round of NFHS, which
was conducted in two rounds between 2019 and 2021. From the Pre COVID-19 round, we
draw the following individual-level information for men and women separately: (a) Do they
own a bank account? (b) Do they own a mobile phone? (c) Do they use their mobile phones
for financial transactions? (d) Do they use the internet? and (e) Do they have access to
money that they can use as they wish? The last piece of information is only there for
women, but not for men. The data for women is drawn from the eligible women dataset of
NFHS, where the sample is restricted to women between the age group of 15-49. To make
our results consistent and comparable with women’s data, we restrict our sample to 15-49
age group for men’s data as well.
1It is worth noting that the data for overall wages/earnings from employment are not available in PLFS.
6
3.4 UPI usage data from Phonepe
We extract district-level UPI usage data from Phonepe, which is one of India’s largest UPI
platforms founded in 2015. In September 2021, Phonepe launched ‘Phonepe Pulse’, an
online repository of digital transaction data from Phonepe users, available for each quarter
from the year 2018. Our district-level treatment variable is the average of the total amount
of digital transactions over the years 2018 and 2019. To add additional layers of robustness,
we use the total count of DFS averaged over the same period and the growth of DFS
between 2018 and 2019 as an alternative treatment variable. We observe wide variations in
UPI usage, as shown in Table A2. Furthermore, in Figure A1, we observe a rapid growth
in UPI usage between 2018 and 2019.
3.5 RBI data on Currency Chests
A few scheduled banks have been granted permission by the Reserve Bank to set up currency
chests to make it easier to distribute rupee coins and banknotes. These are repositories
where the Reserve Bank stocks rupee coins and banknotes for distribution to bank branches
within their service region. We obtain district-level data on the number of currency chests
in 2016, curated by Crouzet, Gupta, and Mezzanotti (2023), which was originally obtained
from RBI. This data is available for 512 districts, excluding districts from northeastern
states and union territories. For the purpose of our analysis, we merged these district-level
data with ITUS, PLFS, and NFHS separately.
4 Identification Strategy
To analyze the impact of DFS on gendered time-use patterns, addressing potential endo-
geneity is crucial. Endogeneity may arise from self-selection, where individuals, especially
women, in economically prosperous regions adopt online payment systems and use their
time more efficiently due to existing economic advantages. Additionally, reverse causality
may occur if individuals engaged in employment activities seek out digital transactions to
save time. To mitigate these concerns, we employ an IV approach. We instrument district-
level DFS using the currency chest exposure index, Chestd, for each district. This choice
is inspired by Chodorow-Reich et al. (2020) and Crouzet, Gupta, and Mezzanotti (2023),
followed by Aggarwal, Kulkarni, and Ritadhi (2023) and Das et al. (2023), who leveraged
historical variations in currency chest as a quasi-random instrument in the context of de-
7
monization in 2016.2The chest exposure index, Chestd, derived from Crouzet et al. (2024),
measures the proportion of total deposits in a district held by banks operating currency
chests. This index reflects the significance of these banks in each district, derived from
detailed deposit and branch data. The relationship is given by:
Chestd=Pb∈CdωjDjbd
Pb∈BdωjDjbd
,
where Cdis the set of banks operating currency chests in district d,Bdis the set of all banks
in the district, ωjis a weighting factor for each bank, and Djbd represents the deposits in
bank bof type jin district d.3Chestdcaptures variations in cash availability across districts,
with higher scores indicating easier access to cash, reducing reliance on digital transactions.
The validity of an IV relies on satisfying two primary conditions: relevance and exclusion
restriction. The relevance condition is met as Chestdsignificantly influences the availability
and adoption of digital payment systems. Historical data indicates that the distribution
of chest banks significantly affects the local financial infrastructure’s capability to support
digital transactions (Crouzet, Gupta, and Mezzanotti,2023). Chest banks play a crucial
role in ensuring liquidity and cash management, thereby facilitating the adoption of digital
payments. Empirical evidence supports this claim, showing that a higher number of chest
banks in a district implies greater access to cash and a lower likelihood of cash shortages,
thus enhancing the adoption of digital payments (Aggarwal, Kulkarni, and Ritadhi,2023).
Therefore, Chestdis a strong predictor of digital financial transaction volumes. To sub-
stantiate this assertion within the context of our dataset, we present a two-way scatterplot
with a linear fit in Panel (a) of Figure 1. The inverse relationship between the chest expo-
sure index and our principal treatment variable (log of the amount transacted) is evident,
reflecting a negative correlation coefficient of -0.30.
The overlapping histogram in Panel (b) of Figure 1highlights the distinct distributions of
the chest exposure index (blue) and the normalized amount of DFS (orange). The chest
exposure index is concentrated at lower values, representing limited regional access to the
2As of November 2016, there were approximately 4,000 chests distributed throughout the country, re-
ceiving newly printed notes either directly from one of the 19 Reserve Bank of India (RBI) issue offices or
via around 600 hub chests (Chodorow-Reich et al.,2020).
3To weighting factor, ωjis determined based on factors such as the size of the bank, the volume of cash
transactions it handles, and its operational scope in managing currency chests. For instance, banks with
larger networks and higher volumes of cash handling may be assigned a higher ωj, reflecting their greater
impact on cash availability in the district. This method allows the measure Chestdto account for not just
the presence of banks with currency chests, but also their operational significance in the local cash economy.
8
infrastructure, while the broader distribution of DFS reflects significant variation across
regions. This separation reinforces the validity of using chest exposure as an IV. The
concentrated distribution of chest exposure indicates its exogeneity, as regional access to
currency chests is unlikely to correlate with unobserved factors directly influencing transac-
tion volumes. Meanwhile, the broader, dispersed distribution of digital transactions points
to multiple underlying determinants, suggesting that chest exposure affects transaction vol-
umes indirectly through its role in expanding access to financial infrastructure. The lack of
overlap between the distributions supports the exclusion restriction.
The placement of chest banks, determined by the Reserve Bank of India (RBI) based on
historical and logistical factors, is unrelated to current time-use patterns or economic be-
haviors at the household level (Crouzet, Gupta, and Mezzanotti,2023). Complementarities
in the adoption of DFS play a key role in the validity of our instrument. Complementarities
arise when the benefits of adopting DFS increase with the number of users. These dynam-
ics further justify using district-level exposure to currency chest infrastructure as an IV,
capturing spatially quasi-random variations in the promotion, of DFS.
To further demonstrate that chest exposure in 2016 can serve as an instrument for DFS in
2018-19, we reference the work of Dubey and Purnanandam (2023). Their study utilizes
the district-level exogeneity based on early and late adoption of UPI by lead banks, and
reveals that the higher impact for early adopters persisted well beyond the launch of UPI in
2016. This finding aligns with the evidence of strong network externalities in the adoption
of DFS, as shown by Crouzet, Gupta, and Mezzanotti (2023) and Higgins (2019).
The first stage regression estimates the relationship between the chest exposure index and
the amount of DFS (log), controlling for individual and district-level characteristics. The
second stage regression is:
Yids =βDds +X′
ids +Z′
dsγ+αs+ϵids ,(1)
where Yids is the outcome variable of interest for individual iresiding in district d. The
other variables are as defined in the first stage. In this stage, the main coefficient of interest
(β) captures the causal impact of DFS on the outcome variables. To ensure the robustness
of our results, we conduct an alternative specifications test by using the growth and count
of DFS as the treatment variable.
9
5 Empirical Analysis
We start with estimating the gender-disaggregated impact of digital payments on the like-
lihood (extensive margin) of employment by occupation and gender.
5.1 Extensive Margin: DFS and Likelihood of Employment
Our findings on aggregate employment,4presented in Figure 2, and in Table A3, indicate
that a 1% increase in district-level DFS leads to a 1.2 percentage point (pp) increase in
aggregate employment for men, while the corresponding effect for women is a 0.6 pp in-
crease, though not statistically significant. Notably, the difference in these estimates is not
statistically significant (see Table A1), suggesting that while the effect for women is smaller,
it nonetheless trends in a positive direction.
Estimates on the impact of DFS on employment structures reveal gender-specific transitions.
For men, a 1% increase in district-level DFS leads to a significant decline in self-employment
(by 2.7 pp), casual employment (by 0.9 pp), and unpaid self-employment (1.2 pp), while a
corresponding increase of 4.8 pp is observed for regular wage employment. This indicates a
stronger movement from self-employment to more formal wage-based roles.
In contrast, for women, the effect of DFS is characterized by a significant shift from un-
paid self-employment and casual employment to regular wage employment, while there is
no change in self-employment (paid).Unpaid self-employment declined more strongly for
women (by 2.3 pp), casual employment decreased (by 1.0 pp), while regular wage employ-
ment rose (by 1.7 pp). Unlike men, women’s transition is from unpaid and casual-insecure
work to more stable wage employment, underscoring a different pathway toward formaliza-
tion in the labor market. These findings suggest that DFS promotes formal employment
for both genders but through distinct transitions—self-employment to regular employment
for men, and unpaid to regular employment for women.
To examine the underlying mechanism of this shift towards formal employment, we decom-
pose employment into its sectoral components. More specifically, we focus on employment
in agricultural and non-agricultural sectors. The estimated coefficients using 2SLS are pre-
sented in Figure 3(and Table A4). The rise in aggregate employment for men (Figure 2),
is driven by a decline in agricultural employment (by 6.3 pp) with a more than propor-
tionate rise in non-agricultural employment (by 7.5 pp). A similar pattern is observed in
4For our extensive margin results, the comparison group of an employed individual is unemployed and
out of labor force individuals.
10
self-employment, where the decline in agricultural self-employment (by 5.2 pp) is bigger
than the gains in non-agricultural self-employment (by 2.5 pp), leading to a decline in ag-
gregate self-employment (by 2.7 pp, Figure 2). This is a case of structural change, where a
shift from subsistence agriculture to a more productive sector is observed.
For women, we observe a shift towards non-agricultural sector, albeit to a lesser extent
as compared to men. We see a reduction in agricultural employment (by 1.4 pp) and an
increase in non-agricultural employment (by 2.0 pp). Although the increase in the latter
is higher, it is not large enough, as reflected in the statistically insignificant increase in
aggregate employment (Figure 2). Additionally, the muted impacts on agricultural self-
employment (decrease by 0.5 pp) and non-agricultural self-employment (increase by 0.5 pp)
is consistent with null impact on aggregate self-employment (Figure 2).
5.2 Intensive Margin: DFS and Time-use in Paid and Unpaid Activities
In this section, we examine the intensive margins of employment (time spent in paid and
unpaid activities, conditional on being employed) using the ITUS, 2019.
The results presented in Panel (a) of Figure 4(Table A5) show that men, who are in any
employment, spent 6.7 minutes more on paid activities with 1% increase in DFS (1.63% rise
over the sample mean). The gains for men in self-employment is 8.7 minutes (2.2% over the
sample mean).5Additionally, there is a 4.5 minute gain in regular-wage employment, which
is a 1% increase relative to the sample mean. The overall increase in employment time,
as shown in Panel (a) of Figure 4, is accompanied by reductions in time-spent on unpaid
activities (caregiving and home production) by 3.8 minutes, self-care (sleeping and related
activities) by 5.8 minutes (Panel (b)), and casual employment by 4.2 minutes, though the
latter is not statistically significant.
For women, the intensive margin results, on aggregate, are largely similar to those for men
(z-score for statistical difference is insignificant (Table A5)). However, the point estimates
for any employment and self-employment, while positive, are not statistically significant.
With regards to regular wage employment, we observe a significant increase of 7.8 minutes
(not statistically different to men); however, this represents a 2.1% rise over the sample
mean, which is notably higher than the 1% increase observed for men.
5At the extensive margin, we observe a decline in self-employment (Figure 2), suggesting that some men
are moving out of self-employment. However, at the intensive margin, those who remain self-employed are
spending more time on self-employment activities, indicating deeper labor market engagement.
11
As shown in Panel (b) of Figure 4, the increase in time spent on paid activities for women,
like men, is driven by a reduction in unpaid activities (by 7.1 minutes), equivalent to a
3.1% decrease over the sample mean, and in self-care (by 5.8 minutes), representing a 0.9%
reduction over the sample mean. However, unlike men, for whom no significant effect on time
spent in learning, socializing, and cultural activities was found, women exhibit a significant
increase in this category (by 8.9 minutes), a 5.5% rise over the sample mean. This shift for
women could be linked to greater use of digital technologies, such as mobile phones and the
internet, as discussed in the next section. These findings suggest a potential reallocation
of women’s time to self-development activities, which may support skill acquisition and
long-term employment opportunities.
Building on the analysis of time allocation in employment, which demonstrates an equitable
increase across genders, we examined the impact of DFS on total annual earnings from self-
employment and regular wage employment.6As shown in Table A6, while the absolute gains
in earnings are higher for men (INR 1,120 for self-employment and INR 1,074 for regular
employment) compared to women (INR 693 and INR 1,125), the percentage increased are
more pronounced for women (13% and 16%) than for men (10% and 13%). However,
insignificant z-scores for statistical differences across genders, indicate equitable growth in
earnings which is commensurate with the increase in time spent on paid activities.
5.3 DFS and Enablers of Digital Economy
So far, we have studied the impacts of DFS on the supply-side shifts in employment patterns.
However, for a more holistic understanding of the impacts of DFS, we explore how the
proliferation of DFS affects the demand for enables of digital economy: access to banks,
mobile phones and the internet. Additionally, we analyze the changes in digital transactions
at an individual level, and the overall impact on women’s financial autonomy.
Using the NFHS data, in Figure 5(Table A7), we show that a 1% increase in DFS led to
a 3 pp and 2.1 pp increase in ownership of bank accounts for men and women, respectively
(statistically indifferent across genders). Mobile phone ownership increased for women by 7.4
pp, which is significantly higher than the increase for men (by 1.1 pp). This underscores the
higher increase in learning, social and cultural activities found in Panel (b) of Figure 4. We
observe a 8.2 pp increase in the use of internet for women, compared to the 6.9 pp increase
for men, however, the difference across gender is statistically insignificant. Additionally, in
6Sourced from PLFS, which does not provide earnings data for aggregate or casual employment.
12
terms of usage of DFS, we observe a 2.8 pp increase in use of mobile for financial transactions
for women, compared to 3.9 pp increase for men, again there is no statistical difference in
the observed effects across genders. Lastly, we examined the impact of DFS on financial
autonomy (money that you can yourself use) to gauge control over financial resources. This
variable is only available for women and the results show a significant increase of 4.9 pp.
In summary, the expansion of DFS not only restructured employment patterns but also
significantly enhanced access to digital enablers, particularly for women. We find substantial
increases in mobile phone ownership, internet usage, and financial autonomy for women,
complementing their rise in self-development activities. These shifts underscore the role of
DFS in promoting equitable access to digital resources and greater financial independence,
with broader implications for gender equity in the labor market.
5.4 Heterogeneity and Robustness
We conduct heterogeneity analyses to explore how DFS impacts employment across various
socioeconomic groups, accounting for factors like age, education, marital status, caste, and
place of residence.
In Figure A2, we find that transitions to regular employment driven by DFS are evident for
men in both rural and urban sectors. For women, however, these effects are predominantly
observed in urban settings, aligning with broader access to digital infrastructure in cities.
In Figure A3, marginalized caste groups exhibit more pronounced shifts towards regular
employment across both genders, indicating a more equitable distribution of DFS benefits
among socially disadvantaged groups.
Further analysis, considering age, education, and marital status (Figures A4,A5,A6),
reveals that young and middle-aged and less-educated individuals primarily drive these
transitions. This is because DFS reduces financial barriers, increases accessibility, and
promotes economic mobility, particularly for those more adaptable to labor market shifts.
Among married individuals, we observe shifts from self-employment to regular employment
for men, while for women, movements from casual employment to regular wage employment
are more frequent. These findings suggest that DFS particularly supports economically
vulnerable and marginalized groups, promoting inclusive participation in the labor market.
Additionally, time allocation between goods and services shows gender-specific trends (Table
A8). Self-employed men are increasingly shifting toward service-oriented work, reflecting a
sectoral reallocation towards services. For women in regular wage employment, we observe
13
an increased engagement in service-related activities with minimal shifts in goods produc-
tion. These patterns, corroborated by reductions in unpaid labor (Table A9), highlight the
potential of DFS to drive structural shifts in labor markets, particularly benefiting women
and self-employed men.
5.4.1 Robustness of IV
To further ensure the exogeneity of our IV, we control for any district-level variable that
could affect the exposure to chest banks. Here, we examine variables that could potentially
affect a district’s chest bank exposure, such as (i) share of villages with ATM in the district,
(ii) share of villages with bank branch in the district, (iii) population density, (iv) number
of illiterate persons, (v) distance to state capital, (vi) credit society to population ratio and
(vii) employment rate. In Table A10, we regress these district-specific variables onto our
chest exposure index, and find that share of villages with ATM and population density are
correlated significantly with chest exposure. Therefore, as a robustness check, we control
for these two factors in our main results, presented in Table A11. Our results do not change
even after including these district-specific controls.
5.4.2 Other Robustness Checks
Furthermore, we carry out an array of robustness checks to maximize the confidence in our
estimates. First, we use OLS to re-estimate our main results. In Table A12, we present
the OLS estimates of the extensive margin of types of employment dummies from PLFS.
Similarly, in Table A13, we show the OLS estimates of intensive margin for paid employment
activities. Our results using OLS are consistent with the 2SLS estimates.
Secondly, we use the average log count of digital transactions between 2018 and 2019 as an
alternative treatment variable, instead of average log amount. The results are produced in
Table A14 for extensive margin of employment and in Table A15 for intensive margin of
employment, again for four types of employment categories. We find that the results are
consistent with our main results.
Thirdly, we use the growth in the amount of digital transactions between 2018 and 2019
as an alternative explanatory variable. Growth is measured as the log difference between
amounts transacted in 2018 and 2019. We present the extensive margin results in Table
A16 and intensive margin results in Table A17, which are again found to be consistent with
our original set of results.
14
6 Conclusion
This study examines the causal effects of Digital Financial Services (DFS) on labor market
dynamics in India, emphasizing differential impacts across genders. Our analysis reveals
that DFS catalyzes significant shifts in employment structures, particularly transitioning
labor towards more productive sectors. For men, this transition is characterized by a pro-
nounced reduction in agricultural self-employment coupled with increased participation in
non-agricultural sectors. For women, the movement away from unpaid self-employment to
regular wage employment marks a crucial step towards access to formal labor markets and
dismantling of traditional gender roles.
On the intensive margin, DFS markedly increases labor time across both self and regular
wage employment, yielding significant wage gains for both genders. Importantly, these
gains are comparable for women, thereby contributing to narrowing the gender wage gap
and enhancing gender equality in labor market outcomes. Moreover, the reallocation of
time from unpaid domestic and caregiving activities to productive employment highlights a
shift in household labor dynamics that is particularly beneficial for women, reinforcing the
role of DFS in promoting gender equity.
Furthermore, DFS bolsters financial autonomy for both men and women by enhancing access
to banking services and enabling mobile-based financial transactions. For men, this access
facilitates income diversification, particularly into more productive non-agricultural sectors,
fostering broader economic growth. For women, increased access to financial tools—banks,
mobile phones, the internet, and DFS—empowers them to engage more actively in wage-
based employment and formal financial systems, thereby expanding their economic agency.
For women in particular, the adoption of formal financial instruments promotes greater
financial independence and supports more equitable economic participation.
In summary, this paper illustrates that DFS serves as a potent catalyst for structural changes
in the labor market, effectively reshaping traditional gender roles and reducing disparities
in labor and financial inclusion. The profound potential of DFS to foster labor market
participation and drive substantive social change underscores its capacity to promote gender
equity and economic empowerment for women. The enduring impacts of these dynamics
suggest that the widespread adoption of DFS is pivotal for achieving gender-equitable and
sustainable labor market outcomes.
15
7 Figures
(a) Scatterplot of Chest Exposure and
Amount of DFS (Log)
(b) Distribution of Chest Exposure and Nor-
malized DFS
Figure 1: Panel (a): Correlation between the Chest Exposure Index and the Amount of Digital
Financial Transactions (log-transformed), trimmed at 1%. The scatterplot presents individual data points
along with a fitted trend line and 95% confidence intervals (shaded). The negative slope indicates a moderate negative
correlation with a coefficient of -0.30, suggesting that higher chest exposure is associated with lower digital financial
transaction volumes. Panel (b): Distribution of the Chest Exposure Index (blue) and the Normalized
Amount of Digital Financial Transactions (orange). The normalized transaction amounts are scaled between
0 and 1 for comparability. Both variables are displayed using 30 bins, and the distinct shapes of the distributions
indicate that the Chest Exposure Index is concentrated at lower values, while the digital financial transactions are
more widely spread.
16
Figure 2: Impact of DFS on Employment Structure: 2SLS Estimates. Notes: The figure presents 2SLS
estimates of the impact of DFS on various employment categories, differentiated by gender. Coefficient estimates and
95% confidence intervals are shown. Square markers represent estimates for men, while diamond markers represent
estimates for women. The vertical dashed line at zero indicates the reference point for no effect. The dependent
variable in each model is a dummy variable. For all employment, if a person is employed as per the Usual Principal
Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for those who are unpaid self-employed, unemployed
or out of labor force. Self-employment dummy takes value ‘1’ if a person is self-employed (own-account worker or
employer), ‘0’ otherwise. Similarly, the regular and casual wage dummies are defined. Y-axis shows the outcome
variables of employment in aggregate employment, self-employment, regular wage employment, causal employment
and unpaid self-employment. The X-axis represents the local average treatment effects in percentage points. Data:
PLFS, 2019-2020, RBI Chest Exposure, 2016 and Phone Pay data (2018, 2019).
17
Figure 3: Impact of DFS on Employment and Self-Employment by Sector: 2SLS Estimates.
Notes: This figure presents 2SLS estimates analyzing the impact of DFS on employment in the agricultural and
non-agricultural sectors. The coefficient estimates and 95% confidence intervals are displayed. Squares represent
estimates for men, and diamonds represent estimates for women. The vertical dashed line indicates the reference
point for no effect. Agricultural self-employment is a dummy variable that captures activities related to farming and
allied sectors, while non-agricultural self-employment covers informal businesses and entrepreneurial activities. Y-axis
shows the outcome variables. The X-axis represents the local average treatment effects in percentage points. Data:
PLFS, 2019-2020, RBI Chest Exposure, 2016 and Phone Pay data (2018, 2019).
Figure 4: Impact of DFS on Time Spent on Paid and Unpaid Activities (minutes per day): 2SLS
Estimates. Notes: The figure presents 2SLS estimates of the impact of DFS on time allocation to paid and unpaid
activities. Square markers represent men’s estimates, and diamond markers represent women’s estimates. Coefficient
estimates and 95% confidence intervals are displayed. The vertical dashed line indicates no effect. Y-axis shows
variables for paid (Panel (a)) and unpaid (Panel (b)) activities, for Men (M) and Women (W). Analysis for paid
activities is conducted separately for those employed in any category, self-employed, and regular wage employed.
The analysis on unpaid activities (including unpaid domestic services, learning, social, and cultural activities, and
self-care) is reported only for employed persons. The X-axis represents the local average treatment effects in minutes
per day. Data: ITUS, 2019, RBI Chest Exposure, 2016 and Phone Pay data (2018, 2019).
18
Figure 5: Impact of DFS on Enablers of Digital Economy: 2SLS Estimates. Notes: This figure shows
2SLS estimates for the impact of digital transactions on enablers of digital economy. The indicators include owning a
bank account, owning a phone, using a mobile for financial transactions, using the internet, and access to money for
personal use. Coefficient estimates and 95% confidence intervals are presented. Squares represent estimates for men,
while diamonds represent estimates for women. The vertical dashed line at zero indicates no effect. Y-axis shows the
outcomes variables. All these dependant variables are in the form of dummy variables where ‘yes; is coded as ‘1’ and
‘no’ as ‘0’. The X-axis represents the local average treatment effects in percentage points. Data: NFHS (2019), RBI
Chest Exposure (2016) and Phone Pay data (2018, 2019).
19
References
Abiona, Olukorede and Martin Foureaux Koppensteiner. 2022. “Financial inclusion, shocks,
and poverty: Evidence from the expansion of mobile money in Tanzania.” Journal of
Human Resources 57 (2):435–464.
Agarwal, Sumit, Pulak Ghosh, Jing Li, and Tianyue Ruan. 2024. “Digital Payments and
Consumption: Evidence from the 2016 Demonetization in India.” The Review of Financial
Studies 37 (8):2550–2585. URL https://doi.org/10.1093/rfs/hhae005.
Aggarwal, Bhavya, Nirupama Kulkarni, and SK Ritadhi. 2023. “Cash is king: The role of
financial infrastructure in digital adoption.” The Review of Corporate Finance Studies
12 (4):867–905.
Aggarwal, Shilpa, Valentina Brailovskaya, and Jonathan Robinson. 2023. “Saving for mul-
tiple financial needs: evidence from lockboxes and mobile money in Malawi.” Review of
Economics and Statistics 105 (4):833–851.
Aker, Jenny C. 2010. “Information from markets near and far: Mobile phones and agricul-
tural markets in Niger.” American Economic Journal: Applied Economics 2 (3):46–59.
Anderson, Siwan and Mukesh Eswaran. 2009. “What determines female autonomy? Evi-
dence from Bangladesh.” Journal of development Economics 90 (2):179–191.
Barth, Erling, Sari Pekkala Kerr, and Claudia Olivetti. 2021. “The dynamics of gender
earnings differentials: Evidence from establishment data.” European Economic Review
134:103713.
Batista, Catia and Pedro C Vicente. 2023. “Is mobile money changing rural Africa? Evi-
dence from a field experiment.” Review of Economics and Statistics :1–29.
Beck, Thorsten, Haki Pamuk, Ravindra Ramrattan, and Burak R Uras. 2018. “Payment
instruments, finance and development.” Journal of Development Economics 133:162–186.
Bertrand, Marianne, Patricia Cortes, Claudia Olivetti, and Jessica Pan. 2021. “Social
norms, labour market opportunities, and the marriage gap between skilled and unskilled
women.” The Review of Economic Studies 88 (4):1936–1978.
Boppart, Timo and Per Krusell. 2020. “Labor supply in the past, present, and future: a
balanced-growth perspective.” Journal of Political Economy 128 (1):118–157.
20
Carli, Francesco and Burak R Uras. 2024. “E-money, risk-Sharing, and welfare.” European
Economic Review :104832.
Chen, Boou and Chunkai Zhao. 2021. “Poverty reduction in rural China: Does the digital
finance matter?” PLOS ONE 16 (12):1–27. URL https://doi.org/10.1371/journal.
pone.0261214.
Chodorow-Reich, Gabriel, Gita Gopinath, Prachi Mishra, and Abhinav Narayanan. 2020.
“Cash and the economy: Evidence from India’s demonetization.” The Quarterly Journal
of Economics 135 (1):57–103.
Cole, Shawn, Mukta Joshi, and Antoinette Schoar. 2024. “Heuristics on Call: The Impact of
Mobile-Phone-Based Business-Management Advice.” The World Bank Economic Review
38 (3):580–597. URL https://doi.org/10.1093/wber/lhad038.
Cong, Lin William, Kay Giesecke, and Cami Kuhnen. 2024. “Call for Papers—Management
Science Virtual Special Issue on Digital Finance.” Management Science 70 (8):Vi–Vii.
Crouzet, Nicolas, Pulak Ghosh, Apoorv Gupta, and Filippo Mezzanotti. 2024. “Demo-
graphics and Technology Diffusion: Evidence from Mobile Payments.” SSRN URL
https://ssrn.com/abstract=4778382. Available at SSRN.
Crouzet, Nicolas, Apoorv Gupta, and Filippo Mezzanotti. 2023. “Shocks and technology
adoption: Evidence from electronic payment systems.” Journal of Political Economy
131 (11):3003–3065.
Das, Satadru, Lucie Gadenne, Tushar Nandi, and Ross Warwick. 2023. “Does going cashless
make you tax-rich? Evidence from India’s demonetization experiment.” Journal of Public
Economics 224:104907.
De Mel, Suresh, Craig McIntosh, Ketki Sheth, and Christopher Woodruff. 2022. “Can
mobile-linked bank accounts bolster savings? evidence from a randomized controlled
trial in Sri Lanka.” Review of Economics and Statistics 104 (2):306–320.
Demirguc-Kunt, Asli, Leora Klapper, Dorothe Singer, and Saniya Ansar. 2018. The Global
Findex Database 2017: Measuring financial inclusion and the fintech revolution. World
Bank Publications.
Dizon, Felipe, Erick Gong, and Kelly Jones. 2020. “The effect of promoting savings on
informal risk sharing: experimental evidence from vulnerable women in Kenya.” Journal
of Human Resources 55 (3):963–998.
21
Dubey, Tamanna Singh and Amiyatosh Purnanandam. 2023. “Can Cashless Payments Spur
Economic Growth?” Available at SSRN 4373602 .
Duflo, Esther. 2012. “Women Empowerment and Economic Development.” Journal of
Economic Literature 50 (4):1051–1079.
Dupas, Pascaline and Jonathan Robinson. 2013. “Savings constraints and microenterprise
development: Evidence from a field experiment in Kenya.” American Economic Journal:
Applied Economics 5 (1):163–192.
Exley, Christine L and Judd B Kessler. 2022. “The gender gap in self-promotion.” The
Quarterly Journal of Economics 137 (3):1345–1381.
Field, Erica, Rohini Pande, Natalia Rigol, Simone Schaner, and Charity Troyer Moore.
2021. “On her own account: How strengthening women’s financial control impacts labor
supply and gender norms.” American Economic Review 111 (7):2342–2375.
Herrendorf, Berthold, Richard Rogerson, and Akos Valentinyi. 2014. “Growth and struc-
tural transformation.” Handbook of economic growth 2:855–941.
Higgins, Sean. 2019. “Financial technology adoption.” Northwestern University: Kellogg
School of Management .
Insights, Fortune Business. 2023. “Digital Payment Market Size, Share Trends
Analysis Report, 2023-2030.” URL https://www.fortunebusinessinsights.com/
digital-payment-market-102743. Accessed: 2024-09-12.
Jack, William and Tavneet Suri. 2014. “Risk sharing and transactions costs: Evidence from
Kenya’s mobile money revolution.” American Economic Review 104 (1):183–223.
Jayachandran, Seema. 2015. “The roots of gender inequality in developing countries.”
Annual Review of Economics 7 (1):63–88.
Kundu, Anirban. 2020. “Impact of trade liberalisation on formal–informal interlinkages in
India: does sectoral labour mobility matter?” Journal of Economic Structures 9 (1):42.
Lewis, William Arthur et al. 1954. “Economic development with unlimited supplies of
labour.” .
Londo˜no-V´elez, Juliana and Pablo Querubin. 2022. “The impact of emergency cash assis-
tance in a pandemic: Experimental evidence from Colombia.” Review of Economics and
Statistics 104 (1):157–165.
22
Mbiti, Isaac and David N Weil. 2015. “Mobile banking: The impact of M-Pesa in Kenya.”
In African successes, Volume III: Modernization and development. University of Chicago
Press, 247–293.
Moorena, Lolita, Mikaela Rabb, Aliyyah Rusdinar, Simone Schaner, Caroline Tangoren,
and Natalie Theys. 2020. “Towards inclusive digital finance in indonesia.” Inclusive
Financial Innovation Initiative (IFII) White Paper .
Pazarbasioglu, Ceyla, Alfonso Garcia Mora, Mahesh Uttamchandani, Harish Natarajan,
Erik Feyen, and Mathew Saal. 2020. “Digital financial services.” World Bank 54:1–54.
Riley, Emma. 2018. “Mobile money and risk sharing against village shocks.” Journal of
Development Economics 135:43–58.
Shapiro, Alan Finkelstein and Federico S Mandelman. 2021. “Digital adoption, automa-
tion, and labor markets in developing countries.” Journal of Development Economics
151:102656.
Statista. 2023. “Digital Payments Worldwide - Statistics Facts.” URL https://www.
statista.com/statistics/871273/worldwide-digital-payments-market-size/.
Accessed: 2024-09-12.
Suri, Tavneet, Prashant Bharadwaj, and William Jack. 2021. “Fintech and household
resilience to shocks: Evidence from digital loans in Kenya.” Journal of Development
Economics 153:102697.
Suri, Tavneet and William Jack. 2016. “The long-run poverty and gender impacts of mobile
money.” Science 354 (6317):1288–1292.
Yang, Xiaolan, Lele Zhang, Xiaoyue Hong, and Wenchao Li. 2024. “Can digital financial
inclusion facilitate intergenerational income mobility? Evidence from China.” Journal of
Comparative Economics .
23
Online Appendix
Table A1: Summary Statistics
(1) (2) (3) (4) (5) (6) (7)
Men Women Difference in Mean
N Mean SD N Mean SD
A. PLFS 2019-2020 Extensive Margin
All Employment 127,677 0.633 0.482 125,580 0.162 0.368 0.47 ***
- Employed in Agriculture 127,677 0.195 0.396 125,580 0.065 0.246 0.130 ***
- Employed in Non-agriculture 127,677 0.439 0.496 125,580 0.097 0.295 0.342 ***
Self Employed 127,677 0.303 0.459 125,580 0.057 0.232 0.246 ***
- Employed in Agriculture 127,677 0.145 0.352 125,580 0.029 0.169 0.116 ***
- Employed in Non-agriculture 127,677 0.158 0.364 125,580 0.029 0.164 0.130 ***
Unpaid Self Employed 127,677 0.054 0.227 125,580 0.069 0.253 -0.015 ***
Regular Employed 127,677 0.184 0.387 125,580 0.058 0.001 0.126 ***
Casual Employed 127,677 0.146 0.353 125,580 0.046 0.209 0.100 ***
Intensive Margin
Wages for self-employment activities 38,714 10888.1 10043.1 7,186 4844.3 5838.9 6043.8 ***
Wages for Regular salary/wage activities 42,204 8980.5 14534.6 13,119 6997.2 13058.5 1983.4 ***
B. TUS Data 2019-2020 Intensive Margin
Time spent on Paid Activities (In minutes)
Employed 82,444 414.06 112.55 16,535 344.54 116.28 69.52 ***
Self Employed 35,763 395.77 126.54 4,465 307.46 127.47 88.31 ***
Regular Employed 21,087 442.59 97.80 5,353 365.07 111.99 77.52 ***
Casual Employed 25,594 416.10 96.99 6,717 352.83 105.69 63.27 ***
Time spent on Other Activities for employed persons (In minutes)
Unpaid Activities 82,444 45.90 71.97 16,535 227.33 130.99 -181.43 ***
Learning, Social & Cultural 82,444 202.57 112.11 16,535 159.39 97.29 43.19 ***
Self-Care 82,444 700.81 87.21 16,535 660.70 82.47 40.10 ***
C. NFHS Data 2019
Own Bank account 45,626 0.854 0.352 53,117 0.775 0.417 0.079 ***
Own Mobile 45,626 0.905 0.293 53,117 0.532 0.498 0.372 ***
Used Internet 45,626 0.515 0.499 53,117 0.294 0.456 0.221 ***
Use mobile for financial transaction 41,296 0.246 0.431 28,284 0.202 0.401 0.045 ***
Financial Autonomy 53,117 0.493 0.499
Notes: This table presents summary statistics for both Men and Women, comparing various employment measures across
agricultural and non-agricultural sectors. Columns (1)-(3) report statistics for Men, while columns (4)-(6) correspond to
Women. Column (7) highlights the difference in means between Men and Women, with significance levels denoted by ***
(p < 0.01). Employment measures cover the extensive margin (i.e., whether the individual is employed in agriculture, non-
agriculture, self-employment, regular employment, casual employment or unpaid self-employment) and the intensive margin
(i.e., wages for employment activities, time spent on paid activities, and time spent on other activities such as unpaid work
and self-care). The intensive margin includes data from the Time Use Survey (TUS) 2019-2020, and the extensive margin
is derived from the Periodic Labour Force Survey (PLFS) 2019-2020. The last section (C) incorporates indicators defining
enablers of digital economy from the National Family Health Survey (NFHS). Differences between Men and Women are
statistically tested, with significance levels marked as * (p < 0.10), ** (p < 0.05), and *** (p < 0.01).
1
Table A2: Additional Summary Statistics
Panel A: Additional summary statistics from TUS data
Men Women Difference in Mean
N Mean SD N Mean SD
Intensive Margin (Minutes Spent)
Production of goods
Employed 82,444 184.27 197.03 16,535 170.32 180.12 13.95 ***
Self Employed 35,763 188.32 188.09 4,465 180.20 162.92 8.12 ***
Regular Employed 21,087 66.47 153.35 5,353 41.75 120.68 24.71 ***
Casual Employed 25,594 275.68 190.87 6,717 266.22 167.87 9.45 ***
Production of Services
Employed 82,444 132.24 194.20 16,535 76.98 149.83 55.26 ***
Self Employed 35,763 166.87 203.77 4,465 104.06 162.60 62.80 ***
Regular Employed 21,087 128.07 194.32 5,353 100.67 166.89 27.40 ***
Casual Employed 25,594 87.30 169.26 6,717 40.10 114.89 47.20 ***
Panel B: Summary Statistics for DFS and Banking at the District Level
N Mean SD Min 25th 75th Max
Amount (log) 512 21.59 1.18 16.45 20.81 22.24 26.53
Count (log) 512 14.16 1.24 8.76 13.37 14.82 19.41
Chest Exposure Index 512 0.569 0.189 0.107 0.44 0.70 1
Growth in amount of transactions 512 1.107 0.29 -0.90 0.98 1.22 2.53
Status ATM 512 0.031 0.07 0 0.004 0.03 0.5
Status Commercial Banks 512 0.08 0.13 0 0.027 0.08 0.85
Density 510 452.2 357.6 2 205 607 2913
Distance to state capital 512 0.216 0.134 0 0.118 0.293 0.787
Credit Societies per 1000 persons 512 0.043 0.089 0 0.007 0.041 1.178
Worker population Ratio 512 0.409 0.068 0.258 0.352 0.463 0.572
Illiterate Persons (log) 512 13.35 0.772 9.18 12.90 13.91 14.93
Notes: Panel A presents additional summary statistics from the Time Use Survey (TUS) 2019-2020 data, disaggregated
by gender, focusing on both extensive and intensive margins. The extensive margin reports the proportion of individuals
employed, self-employed, and working in regular or casual employment. The intensive margin covers time spent on pro-
duction of goods and services, highlighting gender differences in hours worked across employment types. Panel B provides
summary statistics for outcome and control variables related to financial transactions, including the amount and growth of
transactions, as well as the district-level characteristics. Differences in means between Men and Women are tested, with
significance levels denoted by * (p < 0.10), ** (p < 0.05), and *** (p < 0.01). All statistics are presented with their
corresponding sample sizes (N), means, standard deviations (SD), and range (Min, Max, 25th, 75th percentiles) where
applicable.
2
Table A3: Impact of Digital Economy on Employment Structure (2SLS) using PLFS data:
Extensive Margin
Type of employment dummy (2SLS)
(1) (2) (3) (4) (5)
All Employment Self-Employment Regular Employment Casual Employment Unpaid Self-Employment
Panel A: Men
Amount (log) 0.012*** -0.027*** 0.048*** -0.009* -0.012***
SE [0.004] [0.006] [0.006] [0.005] [0.003]
P-val [0.001] [0.000] [0.000] [0.065] [0.000]
N 127677 127677 127677 127677 127677
F-stat from first stage 61.74 61.74 61.74 61.74 61.74
Sample Mean 0.633 0.303 0.183 0.146 0.054
Panel B: Women
Amount (log) 0.006 0.000 0.017*** -0.010*** -0.023***
SE [0.005] [0.004] [0.003] [0.003] [0.005]
P-val [0.235] [0.992] [0.000] [0.001] [0.000]
N 125580 125580 125580 125580 125580
F-stat from first stage 65.907 65.907 65.907 65.907 65.907
Sample Mean 0.161 0.057 0.058 0.046 0.069
Z-Score 0.937 3.744*** 4.621*** 0.171 1.886**
Notes: (i) These models estimate the impact of digital transactions on the incidence of employment by gender. (ii) The
dependent variable in each model is a dummy variable. In column (1), if a person is employed as per the Usual Principal
Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for those who are unpaid self-employed, unemployed or out of
labor force. (iii) Self-employment dummy takes value ‘1’ if a person is self-employed (own-account worker or employer), ‘0’
otherwise. Similarly, the regular and casual wage dummies are defined. (iv) The main explanatory variable in all models is
the district-level amount transacted via digital platform of Phonepay. It is computed as an average of value of transactions
for the years of 2018 and 2019. (v) All models are estimated on the basis of 2SLS model, where the main explanatory
variable is instrumented by the district-level exposure to the bank chests. Standard errors are clustered and presented in
brackets. The Z-score tests for significant gender differences in the coefficients across employment types. Control variables
include individual and household characteristics. F-statistics from the first-stage regression are reported for robustness.
***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. The sample is drawn from the
PLFS 2019-2020 dataset.
3
Table A4: Impact of Digital Economy on Employment by Sector using PLFS Data: Ex-
tensive Margin
(1) (2) (3) (4)
All Employment Self-Employment
Agricultural Non-agricultural Agricultural Non-agricultural
Panel A: Men
Amount (log) -0.063*** 0.075*** -0.052*** 0.025***
SE [0.007] [0.008] [0.006] [0.005]
P-val [0.000] [0.000] [0.000] [0.000]
N 127677 127677 127677 127677
F-stat from first stage 61.74 61.74 61.74 61.74
Sample Mean 0.194 0.438 0.145 0.157
Panel B: Women
Amount (log) -0.014*** 0.020*** -0.005 0.005
SE [0.004] [0.005] [0.003] [0.003]
P-val [0.001] [0.000] [0.153] [0.114]
N 125580 125580 125580 125580
F-stat from first stage 65.907 65.907 65.907 65.907
Sample Mean 0.064 0.096 0.029 0.027
Z-score 6.078*** 5.830*** 7.006*** 3.430***
Note: This table shows the impact of the digital economy on employment outcomes in agricultural and non-agricultural
sectors, also with a focus on self-employment. Columns (1) and (2) report the effects on agricultural and non-agricultural
employment, while columns (3) and (4) analyze self-employment, differentiating between agricultural and non-agricultural
activities. Agricultural self-employment is a dummy variable that captures activities related to farming and allied sectors,
while non-agricultural self-employment covers informal businesses and entrepreneurial activities. The main explanatory
variable in all models is the district-level amount transacted via digital platform of Phonepay. It is computed as an average
of value of transactions for the years of 2018 and 2019. All models are estimated on the basis of 2SLS model, where the
main explanatory variable is instrumented by the district-level exposure to the bank chests. Standard errors are clustered
and presented in brackets. The Z-score tests for significant gender differences in the coefficients across employment types.
Control variables include individual and household characteristics. F-statistics from the first-stage regression are reported
for robustness. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. The sample is
drawn from the PLFS 2019-2020 dataset.
4
Table A5: Impact of digital transactions on time spent on paid and unpaid activities: Intensive Margin
2SLS and Control Variables
(1) (2) (3) (4) (5) (6) (7)
Any employment Self-employed Regular wage Casual Employment Unpaid Activities Learning, Social and Cultural Self-Care
Panel A: Men
Amount (log) 6.774** 8.774*** 4.591* -4.148 -3.838*** -0.528 -5.840***
SE [2.661] [3.038] [2.650] [3.573] [1.202] [2.696] [2.182]
P-value [0.011] [0.004] [0.083] [0.246] [0.001] [0.845] [0.007]
N 82444 35763 21087 25594 82444 82444 82444
F-stat from first stage 73.767 82.859 37.22 100.487 73.767 73.767 73.767
Sample Mean 414 396 443 416 46 203 701
Controls Yes Yes Yes Yes Yes Yes Yes
Panel B: Women
Amount (log) 4.389 8.573 7.808** -0.633 -7.142** 8.967*** -5.837**
SE [3.580] [5.811] [3.312] [6.477] [3.531] [3.135] [2.844]
P-value [0.220] [0.140] [0.018] [0.922] [0.043] [0.004] [0.040]
N 16535 4465 5353 6717 16535 16535 16535
F-stat from first stage 58.549 74.528 40.382 45.426 58.549 58.549 58.549
Sample Mean 345 307 365 353 227 159 661
Controls Yes Yes Yes Yes Yes Yes Yes
Z-Score 0.535 0.031 0.758 0.475 0.886 2.296** 0.000
Notes: (i) These models estimate the impact of digital transactions on time spent on paid (columns 1-3) and unpaid
(columns 4-6) activities by gender. (ii) Analysis for paid activities is conducted separately for those employed in any
category, self-employed, and regular wage employed. The analysis in columns (4-6) on unpaid activities (including unpaid
domestic services, learning, social, and cultural activities, and self-care) is reported only for employed persons. (iii) The
main explanatory variable is the district-level amount transacted via digital platform (PhonePe), averaged over 2018 and
2019. (iv) All models are estimated using 2SLS where the main explanatory variable is instrumented with district-level
exposure to bank chests. (v) Control variables are included in all models. (vi) Robust standard errors are clustered at the
district level (in parentheses). (vii) Asterisks *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
5
Table A6: Impact on Wages (2SLS) using PLFS: Intensive Margin
(1) (2) (3) (4)
Earnings from self-employment Activity Earnings from Regular Wage Activity
Men Women Men Women
Amount (log) 1120.380*** 693.508** 1074.966*** 1125.994***
SE [245.749] [288.406] [271.260] [369.764]
P-val [0.000] [0.016] [0.000] [0.002]
N 38714 7186 42204 13119
F-stat from first stage 65.483 30.526 54.3 42.106
Sample Mean 10888 4844 8981 6997
Z-Score 1.127 0.111
Notes: (i) These models estimate the impact of digital transactions on the average wage. (ii) The dependent variable
in models (1 & 2) and (3 & 4) is gross earning for last 30 days from self-employment activity; and from regular
salaried/wage activity respectively. (iii) Models 1 & 2 are estimated for those who are self-employed. Models 3 &
4 are estimated for those who are working in any type of wage labour. (iv) The main explanatory variable is the
district-level amount transacted via digital platform (PhonePe), averaged over 2018 and 2019. (v) All models are
estimated on the basis of 2SLS model, where the main explanatory variable is instrumented by the district-level
exposure to the bank chests. (vi) Robust standard errors are clustered at the district level. (vii) The asterisks *,
**, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
6
Table A7: Impact of DFS on Enablers of Digital Economy using NFHS Data: 2SLS
Estimation
(1) (2) (3) (4) (5)
Owning a bank
account
Owning a
phone
Using mobile
for financial
transactions
Use of
internet
Money that
you can
yourself use
Panel A: Men
Amount (log) 0.030*** 0.011* 0.039*** 0.069***
SE [0.010] [0.006] [0.014] [0.015]
P-val [0.002] [0.077] [0.005] [0.000]
N 45626 45626 41296 45626
F-stat from first stage 31.375 31.375 29.794 31.375
Sample Mean 0.854 0.905 0.246 0.515
Panel B: Women
Amount (log) 0.021** 0.074*** 0.028* 0.082*** 0.049***
SE [0.009] [0.017] [0.017] [0.015] [0.014]
P-val [0.023] [0.000] [0.094] [0.000] [0.000]
N 53117 53117 28284 53117 53117
F-stat from first stage 32.766 32.766 20.385 32.766 32.766
Sample Mean 0.775 0.532 0.201 0.293
Z-Score 0.669 3.495*** 0.499 0.613
Notes: (i) These models estimate the impact of average value of digital transactions on five indicators namely:
owning a mobile; owning a phone; using phone for financial transactions; use of internet; and money that the
women can herself use. (ii) All these dependant variables are in the form of dummy variables where ‘yes; is coded
as ‘1’ and ‘no’ as ‘0’. (iii) The main explanatory variable is the district-level amount transacted via digital platform
(PhonePe), averaged over 2018 and 2019. (iv) All models are estimated using 2SLS model and include control
variables. (v) The main explanatory variable is instrumented by the district-level exposure to the bank chests. (vi)
Robust standard errors are clustered at the district level and given in the parentheses. (vii) The asterisks *, **, and
*** indicate significance at the 10%, 5%, and 1% levels, respectively. (viii) Date source: NFHS and Phonepay.
7
Figure A1: District-level maps of DFS growth.
Notes: We show the growth rates of amount of digital transactions between 2018 and 2019.
Here, -0.90 to 0.97 represents growth of DFS between -90% and 97%. Regions covered in darker
shades mean higher growth in DFS. Growth rate for the darkest shade is between 123% and
396%. Data: Phone Pay Pulse, UPI 2018 and 2019.
8
Figure A2: Impact of DFS on Employment (Extensive Margin) by Sector
Notes: This figure presents the effect of DFS on employment outcomes, disaggregated by
gender and sector (rural/urban), across various employment categories. The y-axis represents
the percentage point change in employment outcomes, and the x-axis shows the sector (rural
or urban). The blue points indicate the estimated coefficients, while the vertical lines represent
95% confidence intervals. All models are estimated using the 2SLS method, where the main
explanatory variable is the district-level amount transacted via the digital platform (PhonePe),
averaged over 2018 and 2019. Each panel corresponds to a specific employment category: self-
employed, casual employment, and regular wage employment for both men and women.
9
Figure A3: Impact of DFS on Employment (Extensive Margin) by Caste Groups
Notes: This figure presents the effect of DFS on employment outcomes, disaggregated by
gender and caste (SC/ST/OBC/General), across various employment categories. The y-axis
represents the percentage point change in employment outcomes, and the x-axis shows the
caste (SC/ST/OBC/General). The blue points indicate the estimated coefficients, while the
vertical lines represent 95% confidence intervals. All models are estimated using the 2SLS
method, where the main explanatory variable is the district-level amount transacted via the
digital platform (PhonePe), averaged over 2018 and 2019. Each panel corresponds to a specific
employment category: self-employed, casual employment, and regular wage employment for
both men and women.
10
Figure A4: Impact of DFS on Employment (Extensive Margin) by Age terciles
Notes: This figure presents the effect of DFS on employment outcomes, disaggregated by gender
and age terciles, across various employment categories. The y-axis represents the percentage
point change in employment outcomes, and the x-axis shows the age terciles. The blue points
indicate the estimated coefficients, while the vertical lines represent 95% confidence intervals.
All models are estimated using the 2SLS method, where the main explanatory variable is
the district-level amount transacted via the digital platform (PhonePe), averaged over 2018
and 2019. Each panel corresponds to a specific employment category: self-employed, casual
employment, and regular wage employment for both men and women. In this figure, we see
that the impact of DFS on employment is stronger for young people, especially young women.
11
Figure A5: Impact of DFS on Employment by Education Categories
Notes: This figure presents the effect of DFS on employment outcomes, disaggregated by gender
and education categories, across various employment categories. E du 1 represents persons
with up to primary education, Edu 2 are those with middle, secondary, higher secondary or
Diploma, Edu 3 is for graduates and above. The y-axis represents the percentage point change
in employment outcomes, and the x-axis shows the education categories. The blue points
indicate the estimated coefficients, while the vertical lines represent 95% confidence intervals.
All models are estimated using the 2SLS method, where the main explanatory variable is
the district-level amount transacted via the digital platform (PhonePe), averaged over 2018
and 2019. Each panel corresponds to a specific employment category: self-employed, casual
employment, and regular wage employment for both men and women.
12
Figure A6: Impact of DFS on Employment by Marital Status (Unmarried/Ever-
Married)
Notes: This figure presents the effect of digital transactions on employment outcomes, dis-
aggregated by gender and marital status, across various employment categories. The y-axis
represents the percentage point change in employment outcomes, and the x-axis shows the
marital status (unmarried/ever-married). The blue points indicate the estimated coefficients,
while the vertical lines represent 95% confidence intervals. All models are estimated using the
2SLS method, where the main explanatory variable is the district-level amount transacted via
the digital platform (PhonePe), averaged over 2018 and 2019. Each panel corresponds to a
specific employment category: self-employed, casual employment, and regular wage employ-
ment for both men and women.
13
Table A8: Division of time spent on production of goods and services: 2SLS Model
(1) (2) (3) (4)
Any employment Self Employed Regular Wage Casual Wage
Panel A: Men
Dependent Variable: Time spent on production of goods
Amount (log) -17.769*** -18.463*** 5.492 -12.371*
SE [4.664] [5.115] [4.682] [6.559]
P-value [0.000] [0.000] [0.241] [0.059]
N 82444 35763 21087 25594
F-stat from first stage 73.767 82.859 37.22 100.487
Sample Mean 184 188 66 276
Controls Yes Yes Yes Yes
Dependent Variable: Time spent on production of services
Amount (log) 13.302** 21.113*** 9.569 5.619
SE [4.915] [6.022] [9.754] [5.163]
P-value [0.007] [0.000] [0.327] [0.277]
N 82444 35763 21087 25594
F-stat from first stage 73.767 82.859 37.22 100.487
Sample Mean 132 167 128 87
Controls Yes Yes Yes Yes
Panel B: Women
Dependent Variable: Time spent on production of goods
Amount (log) -12.132** 1.119 9.992* -7.729
SE [5.929] [6.728] [5.170] [9.832]
P-value [0.041] [0.868] [0.053] [0.432]
N 16535 4465 5353 6717
F-stat from first stage 58.549 74.528 40.382 45.426
Sample Mean 170 180 42 266
Controls Yes Yes Yes Yes
Dependent Variable: Time spent on production of services
Amount (log) 20.267*** 8.67 20.047** 14.817**
SE [5.026] [6.227] [8.905] [5.953]
P-value [0.000] [0.164] [0.024] [0.013]
N 16535 4465 5353 6717
F-stat from first stage 58.549 74.528 40.382 45.426
Sample Mean 77 104 101 40
Controls Yes Yes Yes Yes
Notes: (i) These models estimate the impact of digital transactions on the time spent on production of goods
and services. (ii) Analysis is conducted separately for those who are employed in any category, self-employed,
regular-wage employed and casual/other-wage employed. (iii) The dependent variable in Section 1 & 2 is time spent
on Employment in household enterprises to produce goods; and Employment in household enterprises to provide
services respectively. (iv) The main explanatory variable in all models is the district-level amount transacted via
digital platform of Phonepay. It is computed as an average of value of transactions for the years of 2018 and
2019. (v) All the models are estimated using 2SLS model and include control variables. (vi) The main explanatory
variable (amount) is instrumented by the district-level exposure to the bank chests. (vii) Robust standard errors are
clustered at the district level and given in the parentheses. (viii) The asterisks *, **, and *** indicate significance
at the 10%, 5%, and 1% levels, respectively. 14
Table A9: Impact of digital transactions on time spent on other (unpaid) activities (2SLS):
By Employment Category
(1) (2) (3)
Unpaid
Activities
Learning, Social
and Cultural Self-Care
Panel A1: Self-employed men
Amount(log) -5.000** -0.671 -6.767***
SE [1.935] [3.212] [2.300]
P-value [0.010] [0.834] [0.003]
N 35763 35763 35763
F-stat from first stage 82.859 82.859 82.859
Sample Mean 53 217 708
Panel A2: Regular wage-employed men
Amount(log) -1.829 -2.03 -2.847
SE [1.158] [3.067] [2.724]
P-value [0.114] [0.508] [0.296]
N 21087 21087 21087
F-stat from first stage 37.22 37.22 37.22
Sample Mean 35 190 675
Panel B1: Self-employed women
Amount(log) -4.244 9.287** -11.120***
SE [4.378] [4.718] [3.417]
P-value [0.332] [0.049] [0.001]
N 4465 4465 4465
F-stat from first stage 74.528 74.528 74.528
Sample Mean 264 177 670
Panel B2: Regular wage-employed women
Amount(log) -11.553*** 3.192 1.053
SE [3.551] [3.454] [3.143]
P-value [0.001] [0.355] [0.738]
N 5353 5353 5353
F-stat from first stage 40.382 40.382 40.382
Sample Mean 195 165 648
Notes: This table presents the impact of digital transactions on time spent on unpaid activities, learning/social/cultural
activities, and self-care, estimated using a 2SLS model. The analysis is broken down by employment category, including
self-employed and regular wage-employed individuals, with results separately reported for men (Panels A1 and A2) and
women (Panels B1 and B2). The main independent variable is the log of digital transactions, and the dependent variables
reflect the time spent on each activity category (in minutes). Standard errors (SE) are reported in brackets, and p-values
indicate the significance of the estimates. Significance levels are denoted by *** (p < 0.01), ** (p < 0.05), and * (p < 0.10).
All regressions include individual and household controls, and the F-statistics from the first stage are provided to assess the
strength of the instrumental variable. Sample means indicate the average time spent on each activity for the corresponding
employment category. The sample sizes (N) reflect the number of observations used in each regression.
15
Table A10: Chest Exposure and District-specific variables
Dependent Variable: Exposure to Chest Banks
Coefficient SE P-val
Share of villages with ATM 0.615** 0.209 0.003
Share of villages with Bank Branch 0.045 0.117 0.699
Population Density -0.0001*** 0.00003 0.004
Illiterate Persons (log) -0.014 0.012 0.240
Distance to State capital 0. 061 .0632 0.337
Credit society to population ratio -0.100 0.092 0.279
Employment Rate -0.103 0.153 0.498
N 510
Notes: This table presents estimates from an OLS regression of the district-specific charac-
teristics on the chest exposure index. Share of villages with ATM is the ratio of number of
villages with at least one ATM to the total number of villages in the district. Share of villages
with bank branch is the ratio of number of villages with at least one bank branch to the total
number of villages in the district. Density of the population density in the district. Illiterate
persons (log) is the logarithm of total number of illiterate persons in the district. Distance to
state capital is the total kilometre distance between the district and the state capital. Credit
society to population ratio is the number of credit societies per 1000s. Employment rate is
the ratio of working population to total population. Data for these district-level variables
are taken from Crouzet, Gupta, and Mezzanotti (2023). Significance levels are denoted by
*** (p < 0.01), ** (p < 0.05), and * (p < 0.10).
16
Table A11: Impact of Digital Economy on Employment Structure (2SLS) using PLFS
Data and Additional District-level Controls: Extensive Margin
Type of employment dummy (2SLS)
(1) (2) (3) (4)
All Employment Self-Employment Regular Employment Casual Employment
Panel A: Men
Amount (log) 0.012** -0.021*** 0.043*** -0.011*
SE [0.005] [0.007] [0.008] [0.007]
P-val [0.013] [0.004] [0.000] [0.096]
N 127532 127532 127532 127532
F-stat from first stage 45.898 45.898 45.898 45.898
Sample Mean 0.633 0.303 0.184 0.146
Panel B: Women
Amount (log) 0.009 0.002 0.014*** -0.007*
SE [0.007] [0.006] [0.004] [0.004]
P-val [0.189] [0.770] [0.000] [0.074]
N 125418 125418 125418 125418
F-stat from first stage 47.825 47.825 47.825 47.825
Sample Mean 0.161 0.057 0.058 0.046
Z-score 0.349 2.495** 3.242*** 0.496
Notes: (i) These models estimate the impact of digital transactions on the incidence of employment by gender, with
additional district-specific controls (Share of villages with ATMS, share of villages with bank branches, population
density and education category) (ii) The dependant variable in each model is a dummy variable. In column (1), if
a person is employed as per the Usual Principal Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for
those who are unpaid self-employed, unemployed or out of labour force. (iii) Self-employment dummy takes value
‘1’ if a person is self-employed (own-account worker or employer), ‘0’ otherwise. Similarly, the regular and casual
wage dummies are defined. (iv) The main explanatory variable in all models is the district-level amount transacted
via digital platform of Phonepay. It is computed as an average of value of transactions for the years of 2018 and
2019. (v) All models are estimated on the basis of 2SLS model, where the main explanatory variable is instrumented
by the district-level exposure to the bank chests. Data for district-level controls are taken from the dataset curated
by Crouzet, Gupta, and Mezzanotti (2023). Standard deviations (SD) are reported in parentheses, and differences
between Men and Women are statistically tested, with significance levels marked as * (p < 0.10), ** (p < 0.05), and
*** (p < 0.01).
17
Table A12: Extensive Margin for Types of Employment Dummies, OLS Model (PLFS)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
All
employment
Agricultural
Employment
Non-agricultural
Employment
Self-
employment
Agricultural
Self-employment
Non-agricultural
Self-employment
Unpaid
Self-employment
Regular
employment
Casual
Employment
Panel A: Men
Amount(log) 0.009*** -0.049*** 0.057*** -0.027*** -0.040*** 0.013*** -0.012*** 0.047*** -0.012***
SE [0.002] [0.004] [0.004] [0.003] [0.003] [0.002] [0.001] [0.004] [0.003]
P-value [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
N 127677 127677 127677 127677 127677 127677 127677 127677 127677
Sample Mean 0.633 0.194 0.438 0.303 0.145 0.157 0.054 0.183 0.146
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Panel B: Women
Amount(log) 0.004 -0.015*** 0.019*** -0.006*** -0.008*** 0.003*** -0.018*** 0.017*** -0.007***
SE [0.003] [0.002] [0.003] [0.002] [0.002] [0.001] [0.002] [0.002] [0.002]
P-value [0.172] [0.000] [0.000] [0.002] [0.000] [0.010] [0.000] [0.000] [0.000]
N 125580 125580 125580 125580 125580 125580 125580 125580 125580
Sample Mean 0.161 0.064 0.096 0.057 0.029 0.027 0.069 0.058 0.046
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Notes: (i) These models estimate the impact of digital transactions on the incidence of employment by gender. (ii)
The dependant variable in each model is a dummy variable. In column (1), if a person is employed as per the Usual
Principal Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for those who are unpaid self-employed,
unemployed or out of labour force. (iii) Self-employment dummy takes value ‘1’ if a person is self-employed (own-
account worker or employer), ‘0’ otherwise. Similarly, the regular and casual wage dummies are defined. (iv) The
main explanatory variable in all models is the district-level amount transacted via digital platform of Phonepay. It
is computed as an average of value of transactions for the years of 2018 and 2019. (v) All models are estimated on
the basis of 2SLS model, where the main explanatory variable is instrumented by the district level exposure to the
bank chests. (vi) Robust standard errors are clustered at the district level and given in the parentheses. (vii) The
asterisks *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
18
Table A13: Impact of digital transactions on time spent on paid activities: Intensive
Margin (OLS Model)
(1) (2) (3) (4)
All employment Self-employment Regular employment Casual employment
Panel A: Men
Amount (log) 8.585*** 9.599*** 4.957*** 2.727
SE [1.478] [1.647] [1.278] [1.732]
P-value [0.000] [0.000] [0.000] [0.116]
N 82444 35763 21087 25594
R-Square 0.052 0.049 0.071 0.059
Sample Mean 414 396 443 416
Controls Yes Yes Yes Yes
Panel B: Women
Amount (log) 6.148*** 7.266* 11.282*** -1.749
SE [1.698] [3.220] [1.667] [2.560]
P-value [0.000] [0.024] [0.000] [0.495]
N 16535 4465 5353 6717
R-Square 0.031 0.034 0.091 0.043
Sample Mean 344 307 365 353
Controls Yes Yes Yes Yes
Notes: This table presents the impact of digital transactions on time spent on paid activities, estimated using an Ordinary
Least Squares (OLS) model. The dependent variable is the time spent (in minutes) on different forms of paid employment,
including all employment, self-employment, regular employment, and casual employment. The time spent is conditional on
individuals spending more than zero minutes on the respective employment type, hence focusing on the intensive margin.
Panel A shows results for Men, and Panel B for Women. The main independent variable is the log of digital transactions, and
standard errors (SE) are reported in brackets. The table reports p-values for each estimate, with significance levels denoted
by *** (p < 0.01), ** (p < 0.05), and * (p < 0.10). Sample means refer to the average time spent on each employment
type. All regressions include controls for individual and household characteristics, with sample sizes (N) specified for each
employment type.
19
Table A14: Impact of Digital Economy on Employment Structure (2SLS) at extensive
margin (PLFS): Alternative explanatory variable
(1) (2) (3) (4) (5) (6) (7) (8) (9)
All
employment
Agricultural
Employment
Non-agricultural
Employment
Self-
employment
Agricultural
Self-employment
Non-agricultural
Self-employment
Unpaid
Self-employment
Regular
employment
Casual
Employment
Panel A: Men
Count (log) 0.012*** -0.064*** 0.076*** -0.027*** -0.052*** 0.025*** -0.012*** 0.049*** -0.009*
SE [0.004] [0.007] [0.009] [0.006] [0.006] [0.005] [0.003] [0.006] [0.005]
P-val [0.001] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.065]
N 127677 127677 127677 127677 127677 127677 127677 127677 127677
F-stat from First Stage 56.745 56.745 56.745 56.745 56.745 56.745 56.745 56.745 56.745
Sample Mean 0.633 0.194 0.438 0.303 0.145 0.157 0.054 0.183 0.146
Panel B: Women
Count (log) 0.006 -0.014*** 0.020*** 0 -0.005 0.005 -0.024*** 0.017*** -0.010***
SE [0.005] [0.004] [0.005] [0.004] [0.003] [0.003] [0.005] [0.003] [0.003]
P-val [0.234] [0.001] [0.000] [0.992] [0.153] [0.111] [0.000] [0.000] [0.001]
N 125580 125580 125580 125580 125580 125580 125580 125580 125580
F-stat from First Stage 60.472 60.472 60.472 60.472 60.472 60.472 60.472 60.472 60.472
Sample Mean 0.161 0.064 0.096 0.057 0.029 0.027 0.069 0.058 0.046
Notes: (i) These models estimate the impact of digital transactions on the incidence of employment by gender. (ii)
The dependant variable in each model is a dummy variable. In column (1), if a person is employed as per the Usual
Principal Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for those who are unpaid self-employed,
unemployed or out of labour force. (iii) Self-employment dummy takes value ‘1’ if a person is self-employed (own-
account worker or employer), ‘0’ otherwise. Similarly, the regular and casual wage dummies are defined. (iv) The
main explanatory variable in all models is the district-level number of transactions via digital platform of Phonepay.
It is computed as an average of count of transactions for the years of 2018 and 2019. (v) All models are estimated
on the basis of 2SLS model, where the main explanatory variable is instrumented by the district level exposure to
the bank chests. (vi) Robust standard errors are clustered at the district level and given in the parentheses. (vii)
The asterisks *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
20
Table A15: Impact of digital transactions on time spent on paid activities in intensive
margin: Alternate explanatory variable
Explanatory variable: Count of digital transactions
(1) (2) (3) (4)
Any employment Self Employed Regular Wage Casual Wage
Panel A: Men
Count (log) 6.736** 8.732*** 4.618* -4.072
SE [2.659] [3.024] [2.684] [3.498]
P-value [0.011] [0.004] [0.085] [0.244]
N 82444 35763 21087 25594
F-stat from first stage 69.451 78.817 33.016 102.426
Sample Mean 414 396 443 416
Controls Yes Yes Yes Yes
Panel B: Women
Count (log) 4.322 8.427 7.876** -0.61
SE [3.518] [5.710] [3.324] [6.251]
P-value [0.219] [0.140] [0.018] [0.922]
N 16535 4465 5353 6717
F-stat from first stage 55.804 73.642 35.884 46.973
Sample Mean 345 307 365 353
Controls Yes Yes Yes Yes
Notes: (i) These models estimate the impact of digital transactions on the time spent on productive activities
for Men. (ii) Analysis is conducted separately for those who are employed in any category, self-employed, regular-
wage employed and casual-wage employed. (iii) The dependent variable in each model is the sum of time spent
on Employment in corporations, government and non-profit institutions; Employment in household enterprises to
produce goods; and Employment in household enterprises to provide services. (iv) The main explanatory variable in
all models is the district-level number of transactions via digital platform of Phonepay. It is computed as an average
of count of transactions for the years of 2018 and 2019. (v) All the models are estimated using 2SLS model and
include control variables. (vi) The main explanatory variable (count) is instrumented by the district-level exposure
to the bank chests. (vii) Robust standard errors are clustered at the district level and given in the parentheses.
(viii) The asterisks *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
21
Table A16: Impact of Growth in Digital Economy on Employment Structure (2SLS) at
extensive margin (PLFS)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
All
employment
Agricultural
Employment
Non-agricultural
Employment
Self-
employment
Agricultural
Self-employment
Non-agricultural
Self-employment
Unpaid
Self-employment
Regular
employment
Casual
Employment
Panel A: Men
Growth in Amount 0.055*** -0.297*** 0.352*** -0.127*** -0.242*** 0.115*** -0.057*** 0.226*** -0.044*
SE [0.018] [0.053] [0.063] [0.035] [0.045] [0.026] [0.016] [0.043] [0.024]
P-val [0.002] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.067]
N 127677 127677 127677 127677 127677 127677 127677 127677 127677
F-stat from First Stage 33.095 33.095 33.095 33.095 33.095 33.095 33.095 33.095 33.095
Sample Mean 0.633 0.194 0.438 0.303 0.145 0.157 0.054 0.183 0.146
Panel B: Women
Growth in Amount 0.03 -0.067*** 0.098*** 0 -0.022 0.022 -0.112*** 0.079*** -0.049***
SE [0.026] [0.023] [0.026] [0.021] [0.016] [0.015] [0.031] [0.018] [0.017]
P-val [0.238] [0.004] [0.000] [0.992] [0.164] [0.134] [0.000] [0.000] [0.004]
N 125580 125580 125580 125580 125580 125580 125580 125580 125580
F-stat from First Stage 33.388 33.388 33.388 33.388 33.388 33.388 33.388 33.388 33.388
Sample Mean 0.161 0.064 0.096 0.057 0.029 0.027 0.069 0.058 0.046
Notes: (i) These models estimate the impact of digital transactions on the incidence of employment by gender. (ii)
The dependant variable in each model is a dummy variable. In column (1), if a person is employed as per the Usual
Principal Activity Status, the dummy takes value ‘1’. It takes value ‘0’ for those who are unpaid self-employed,
unemployed or out of labour force. (iii) Self-employment dummy takes value ‘1’ if a person is self-employed (own-
account worker or employer), ‘0’ otherwise. Similarly, the regular and casual wage dummies are defined. (iv) The
main explanatory variable is growth in amount from 2018 to 2019. (v) All models are estimated on the basis of
2SLS model, where the main explanatory variable is instrumented by the district level exposure to the bank chests.
(vi) Robust standard errors are clustered at the district level and given in the parentheses. (vii) The asterisks *,
**, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.
22
Table A17: Impact of growth in amount transacted on time spent on productive activities:
2SLS
Dependent variable: Time spent on productive activities
(1) (2) (3) (4)
Any employment Self employed Regular wage Casual Wage
Panel A: Men
Growth in amount (2019 over 2018) 42.336** 56.550** 24.568 -28.18
SE [19.376] [22.062] [15.617] [24.162]
P-value [0.029] [0.010] [0.116] [0.243]
N 82444 35763 21087 25594
F-stat from first stage 23.887 23.065 24.639 18.234
Sample Mean 414 396 443 416
Controls Yes Yes Yes Yes
Panel B: Women
Growth in amount (2019 over 2018) 30.15 61.764 39.836** -6.104
SE [24.934] [44.504] [17.154] [62.607]
P-value [0.227] [0.165] [0.020] [0.922]
N 16535 4465 5353 6717
F-stat from first stage 26.628 24.247 32.458 13.944
Sample Mean 345 307 365 353
Controls Yes Yes Yes Yes
Notes: (i) These models estimate the impact of growth in digital transactions on the time spent on productive
activities separately for Men (Panel A) and Women (Panel B). (ii) Analysis is conducted separately for those
who are in any type of employment, self-employment, regular-wage and casual/other wage employment. (iii) The
dependent variable in each model is the sum of time spent on Employment in corporations, government and non-profit
institutions; Employment in household enterprises to produce goods; and Employment in household enterprises to
provide services. (iv) The main explanatory variable in all models is the growth in district-level amount transacted
via digital platform of Phonepay. It is computed as growth in the value of transactions in 2019 over 2018. (vi) All
models are estimated using 2SLS model and include control variables. (vii) The main explanatory variable (growth
in amount) is instrumented by the district-level exposure to the bank chests. (viii) Robust standard errors are
clustered at the district level and given in the parentheses. (ix) The asterisks *, **, and *** indicate significance at
the 10%, 5%, and 1% levels, respectively.
23
