Journal of Economics and Management 2026.
Vol.
No.
3, 701-718
https://doi.
org/10.
70716/ecoma.
Predicting Predatory Lending Vulnerability among Female Industrial Workers: Digital Financial Literacy.
Financial Stress.
Self-Efficacy, and Peer Influence Etty Sri Wahyuni1*.
Mercy Reyne Marlina Tirayoh2.
Nolla Puspita Dewi3.
Nurhatisyah4 1 Bachelor in Management.
Universitas Batam.
Indonesia 2 Bachelor in Accounting.
Universitas Batam.
Indonesia 3,4 Doctoral Program in Human Resource Management.
Universitas Batam.
Indonesia *Corresponding Author: ettywahyunie@gmail.
Received: 04/01/2026 | Revised: 12/01/2026 | Accepted: 14/07/2026 | Published: 16/07/2026 Abstract: This study examines the predictive roles of digital financial literacy (DFL), financial stress (FS), financial self-efficacy (FSE), and peer influence (PI) on predatory-lending vulnerability (PLV) among female industrial workers Ai a structurally exposed yet under-studied demographic in Indonesia's digital-credit ecosystem, where 6,991 illegal lending platforms were blocked between 2022 and 2024.
A cross-sectional explanatory survey was administered to 378 female workers in three Batam integrated industrial estates using validated instruments (OECD/INFE DFL.
Prawitz IFDFW.
Lown FSES).
Data were analyzed through PLS-SEM (SmartPLS .
with bootstrapping .
,000 resample.
, following STROBE reporting guidelines.
All four hypotheses were supported .
< .
The model explained 52.
4% of the variance in PLV (QA = .
Financial stress was the strongest predictor ( = .
, followed by peer influence ( = .
, digital financial literacy ( = Ae.
and financial self-efficacy ( = Ae.
, confirming a pushAepull risk architecture.
The single-region cross-sectional design limits causal inference and external validity.
self-reported use of illegal lenders likely understates the true prevalence due to social desirability bias.
The findings apply primarily to feminized manufacturing labor in Southeast Asian export zones.
Originality Ae This is the first integrated PLS-SEM model testing cognitive, affective, psychological, and social predictors of predatory-lending vulnerability among Indonesian female industrial workers, advancing gendersensitive financial-vulnerability research and offering rank-ordered evidence for OJK Regulation No.
11/2024 demand-side interventions aligned with SDG 5 and SDG 8.
Keywords: Predatory Lending.
Digital Financial Literacy.
Financial Stress.
Financial Self-Efficacy.
Peer Influence How to Cite: Wahyuni.
Tirayoh.
Dewi.
, & Nurhatisyah .
Predicting Predatory Lending Vulnerability among Female Industrial Workers: Digital Financial Literacy.
Financial Stress.
Self-Efficacy, and Peer Influence.
Journal of Economics and Management, 4.
, 701Ae718.
https://doi.
org/10.
70716/ecoma.
INTRODUCTION
The rapid digitalization of Indonesia's financial services sector has expanded credit access for low-income groups.
However, it has also triggered a surge in predatory lending practices that disproportionately burden financially fragile populations.
Predatory lending is generally defined as loan arrangements characterizedcharacterized by deceptive marketing, exorbitant interest rates, opaque fees, and aggressive collection techniques that exploit borrowers' financial desperation (Ali et al.
, 2023.
Sukarno et al.
In the Indonesian context, this phenomenon is most visible through illegal peerCopyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
to-peer (P2P) lending platforms Ai locally known as pinjol ilegal Ai and informal door-todoor lenders such as Bank Emok and Bank Berjanji, where monthly interest rates can reach 20%, far exceeding the 1.
75% monthly cap on credit cards set by Bank Indonesia (Yulius, 2023.
OJK, 2.
Between January 2022 and January 2024, the Indonesian Financial Services Authority (Otoritas Jasa Keuangan.
OJK), through the Investment Alert Task Force, blocked 6,991 illegal online lending platforms Ai a number that continues to climb as new operators rebrand or shift servers across jurisdictions (OJK, 2024.
Mahatma & Sari, 2.
The 2024 National Survey of Financial Literacy and Inclusion (SNLIK) reports a national financial literacy index of 65.
43% and an inclusion index of 75.
however, only 12.
of Indonesians actively engage in financial planning, signaling a critical knowledgeAe behavior gap (OJK & BPS, 2024.
Siswanti et al.
, 2.
Wahyuni et al.
further note that the Indonesian financial-literacy landscape exhibits a pronounced urbanAerural divide and an intergenerational gap, rendering lower-income occupational groups particularly susceptible to exploitative digital credit.
Rural literacy lags at 59.
25%, and the disparity is sharpest among lower-income occupational groups (OJK & BPS, 2.
Female industrial workers represent a particularly vulnerable demographic.
The International Labor Organization estimates that approximately 80% of garment factory workers in Indonesia are women aged 15Ae35, the majority of whom hold only a senior high school qualification (ILO, 2.
Indonesia exhibits a 28.
24-percentage-point gender gap in labor force participation .
42%) (BPS, 2.
, and women workers shoulder a "double burden" of paid labor and unpaid domestic care, which intensifies financial pressure (Ardini et al.
, 2.
Wahyuni.
Fachrudin, and Silalahi .
documented that Indonesian women's financial behavior is shaped not only by personal income but also by household roles, social expectations, and the availability of financial alternatives Ai factors that magnify exposure to high-cost credit when formal options are inaccessible.
In industrial enclaves such as the Kabil Integrated Industrial Estate in Batam.
OJK Kepulauan Riau has repeatedly issued warnings about workers Ai particularly women Ai being trapped by pinjol ilegal, with documented cases of harassment via fabricated nude images, threats to family members, and psychological breakdown (Antara, 2026.
OJK Kepri, 2.
Table 1.
Snapshot of Indonesia's Predatory Lending and Financial Literacy Landscape .
2Ae2.
Indicator Illegal online lending platforms blocked (Jan 2022 Ae Jan National financial literacy index .
National financial inclusion index .
Rural financial literacy index Indonesians are actively engaged in financial planning Female labor-force participation rate Male labor-force participation rate Women's share of the garment-factory workforce Monthly interest rate Ae informal lender (Bank Emo.
Bank Indonesia's maximum credit-card interest rate Licensed P2P lenders (Pinda.
as of February 2026 Value 6,991 Source OJK .
80% up to 20% 75% per OJK & BPS .
OJK & BPS .
OJK & BPS .
OJK .
BPS .
BPS .
ILO .
Yulius .
Bank Indonesia OJK .
The data in Table 1 collectively constitute a tangible phenomenon gap: a population segment with rising digital access but structurally constrained financial capability is being Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
systematically exploited, generating not only financial loss but also psychological distress, household conflict, and Ai in extreme cases Ai suicide (Adhrianti et al.
, 2022.
Mahatma & Sari, 2.
Empirical research on the predictors of predatory-lending vulnerability among female industrial workers, however, remains sparse and fragmented.
Four constructs have emerged as theoretically and empirically central to financial decision-making in digital ecosystems.
First, digital financial literacy (DFL) Ai defined by the OECD/INFE .
as the combination of knowledge, skills, attitudes, and behaviors required to use digital financial services safely Ai is consistently associated with lower exposure to fraudulent or predatory products (Setiawan et al.
, 2022.
Lyons & Kass-Hanna.
Ali et al.
, 2.
Setiawan et al.
, using Indonesian data, demonstrated that DFL significantly improves saving behavior and curbs impulsive borrowing, while Burchi et al.
found that low DFL leaves individuals more prone to predatory practices.
Wahyuni et al.
extended this argument by showing that even among digitally engaged young Indonesians, gaps between technological adoption and substantive financial knowledge remain wide, especially in rapidly developing economies.
Second, financial stress Ai the subjective experience of pressure arising from current and anticipated financial obligations (Choi et al.
, 2020.
Warmath et al.
, 2.
Ai is a robust driver of reliance on high-cost credit.
Recent evidence from Indonesia shows that financial stress is intensified by rising living costs and wage instability, particularly among urban informal and industrial workers (Zuhroh et al.
, 2025.
Ardini et al.
, 2.
UK-based detection studies confirm a feedback loop in which financial stress heightens predatory borrowing, which in turn deepens stress (Pizzutilo & Mariani, 2.
Third, financial self-efficacy (FSE) Ai Bandura's .
construct adapted by Lown .
Ai captures an individual's confidence in managing personal finances.
Higher FSE predicts greater financial inclusion, more deliberate borrowing, and resistance to consumption-related social pressure (Mindra et al.
, 2017.
Asebedo & Payne, 2.
Recent studies among working young adults confirm that FSE significantly improves financial management behavior, while low FSE correlates with vulnerability to manipulation and dependence on high-cost credit (Goyal et al.
, 2025.
Ribeiro & Cherobim.
Fourth, peer influence operates through social learning (Bandura, 1.
and normative pressure.
Peers shape financial attitudes, signal which products are "safe," and can either buffer or amplify risky borrowing.
Evidence is mixed: Goyal et al.
report a positive peer effect on subjective financial literacy, whereas Ribeiro and Cherobim .
document a significant negative effect of peer influence on financial management Among Generation Z and young workers, peer endorsement of pinjol has been linked to faster adoption of these products (Ardini et al.
, 2.
Despite this growing body of work, three notable gaps remain.
First, an empirical gap: most Indonesian studies on pinjol focus on Generation Z students or general urban consumers (Sukarno et al.
, 2024.
Setiawan et al.
, 2022.
Wahyuni et al.
, 2.
, while female industrial workers Ai who combine low wages, the double burden, and high digital exposure Ai have rarely been examined as a distinct population.
Second, a theoretical gap:
existing models examine the four predictors in isolation, lacking an integrated framework that simultaneously tests DFL, financial stress.
FSE, and peer influence as predictors of vulnerability to predatory lending .
ather than merely digital lendin.
Third, a contextual gap: although OJK has launched intensive education campaigns in industrial estates such as Kabil (OJK Kepri, 2.
, policy interventions are not yet evidence-based for this Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
demographic, leaving program effectiveness uncertain.
Addressing these gaps makes a theoretical contribution that extends beyond the novelty of the population under study.
For the consumer financial-vulnerability framework (O'Connor et al.
, 2.
, the proposed model reconceptualizes vulnerability from a static, deficit-based attribute of individuals into a dynamic outcome jointly produced by interacting push forces .
conomic pressure and social influenc.
and pull deficits .
imited literacy and self-efficac.
, thereby specifying the mechanisms through which the frameworkAos dimensions combine rather than merely coexist.
For social cognitive theory (Bandura, 1.
, the model provides a critical boundary test: it examines whether the theoryAos core self-regulatory resources Ai knowledge and self-efficacy Ai retain their protective function under conditions of acute economic scarcity, or whether environmental pressure overrides personal agency in the triadic reciprocal relationship.
In doing so, the study moves both theories from describing who is vulnerable toward explaining how vulnerability is produced, a shift with direct consequences for the design of interventions.
This study therefore aims to examine the predictive roles of digital financial literacy, financial stress, financial self-efficacy, and peer influence on predatory-lending vulnerability among female industrial workers in Indonesia.
Drawing on social cognitive theory (Bandura, 1.
and financial-vulnerability theory (O'Connor et al.
, 2.
, four hypotheses are proposed:
H1: Digital financial literacy negatively predicts predatory-lending vulnerability among female industrial workers.
H2: Financial stress positively predicts vulnerability to predatory lending among female industrial workers.
H3: Financial self-efficacy negatively predicts vulnerability to predatory lending among female industrial workers.
H4: Peer influence positively predicts vulnerability to predatory lending among female industrial workers.
Findings are expected to inform gender-sensitive financial-literacy curricula, workplace-based financial-wellness programs, and regulatory targeting strategies Ai contributing simultaneously to the agenda of SDG 5 (Gender Equalit.
and SDG 8 (Decent Work and Economic Growt.
, and to the broader scholarship on digital financial inclusion in emerging economies.
RESEARCH METHODS
Research Design This study employed a quantitative, cross-sectional, explanatory survey design to examine the predictive roles of digital financial literacy (DFL), financial stress (FS), financial self-efficacy (FSE), and peer influence (PI) on predatory-lending vulnerability (PLV) among female industrial workers.
Reporting follows the STROBE (Strengthening the Reporting of Observational Studies in Epidemiolog.
guideline for cross-sectional studies .
on Elm et al.
, 2.
, ensuring methodological transparency in participant recruitment, measurement, and analysis.
The cross-sectional design was selected because it allows the simultaneous measurement of all latent constructs in a defined population at a single point in time, consistent with prior research on financial vulnerability predictors (O'Connor et al.
, 2019.
Setiawan et al.
, 2.
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
Population and Sampling Instrumentation Participants were female industrial workers aged 18Ae45 years employed at integrated industrial estates in Batam.
Kepulauan Riau (Kabil Integrated Industrial Estate.
Batamindo Industrial Park, and Muka Kuning Industrial Estat.
Inclusion criteria were:
active employment in the manufacturing sector for at least six months.
ownership of a smartphone with internet access.
willingness to provide informed consent.
Exclusion criteria were: .
holding a managerial or supervisory position above first-line foreman, and .
household income exceeding 2y the regional minimum wage (UMK).
The target population comprised an estimated 150,000 female industrial workers in Batam's three principal estates (BP Batam, 2.
A proportionate stratified random sampling technique was applied, with strata defined by industrial estate.
The minimum sample size was calculated using G*Power 3.
1 for multiple regression with fA = 0.
15, =
05, power = .
80, and four predictors, yielding a minimum of 85 respondents (Faul et al.
To meet the stringent requirements of PLS-SEM and accommodate possible nonresponse, a final target of 400 respondents was set, exceeding Hair et al.
's .
"10times rule".
After data cleaning, 378 valid responses were retained for analysis.
Instrument Five established and validated instruments were adapted, translated into Bahasa Indonesia using a back-translation procedure (Brislin, 1.
, and combined into a single All items used a five-point Likert scale .
= strongly disagree.
5 = strongly agre.
, except DFL knowledge items, which were dichotomous.
Sample items, sources, and psychometric properties are summarised in Table 2.
Table 2.
Summary of Instruments.
Sample Items, and Psychometric Properties Construct Digital Financial Literacy (DFL) Items Source OECD/INFE .
Setiawan et al.
Prawitz et al.
Financial Stress
(FS)
Financial SelfEfficacy (FSE) Lown .
Peer Influence (PI) Goyal et al.
Jorgensen & Savla .
Predatory Lending Vulnerability (PLV) O'Connor et al.
Anderloni et al.
Sample Item I know how to verify whether an online lender is registered with OJK.
How often do you worry about being able to meet normal monthly living expenses? It is hard to stick to my spending plan when unexpected expenses arise.
(R)
My co-workers often recommend online loan applications they personally I have considered borrowing from an unverified online lender to cover urgent household needs.
AVE
All Cronbach's and Composite Reliability (CR) values exceeded the .
70 threshold (Hair et al.
, 2.
all Average Variance Extracted (AVE) values exceeded .
50, confirming convergent validity.
Discriminant validity was assessed using the HeterotraitAeMonotrait ratio (HTMT), with all values below .
85 (Henseler et al.
, 2.
A pilot study with 40 respondents .
xcluded from the main analysi.
preceded full deployment.
The adaptation of the instruments to the Indonesian industrial-worker context followed a structured sequence prior to the pilot study.
After forward translation and independent backtranslation (Brislin, 1.
, discrepancies between the original and back-translated Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
versions were reconciled in a consensus meeting, and idiomatic terms were localizedlocalized to the lending vocabulary familiar to respondents .
or example, rendering Auunverified online lenderAy as "pinjol tidak terdaftar OJK").
Content validity was then assessed by a panel of five experts Ai two academics in behavioral finance, one consumer-protection practitioner from OJK, one industrial-relations specialist, and one psychometrician Ai who rated the relevance and clarity of each item on a four-point scale.
Item-level content validity indices (I-CVI) ranged from .
80 to 1.
00, and the scale-level index (S-CVI/Av.
90 for all five constructs, meeting recommended thresholds.
three items were reworded, and one redundant item was removed on the panel's advice before the 40-respondent pilot was conducted to confirm readability and preliminary Procedures and Analysis Plan Data were analyzed using IBM SPSS 28 for descriptive statistics and SmartPLS 4 for variance-based Structural Equation Modeling.
The analysis proceeded in four sequential steps: .
descriptive analysis.
measurement model evaluation .
uter mode.
, assessing indicator loadings, internal consistency reliability, convergent validity, and discriminant validity (HTMT and FornellAeLarcke.
structural model evaluation .
nner mode.
, reporting collinearity (VIF).
RA.
QA, fA, and standardised path coefficients.
hypothesis testing via bootstrapping with 5,000 resamples (Hair et al.
, 2.
PLS-SEM
was selected because the model is predictive, contains both reflective constructs and a non-normally distributed dependent variable, and aligns with prior financialvulnerability research using comparable methods (Setiawan et al.
, 2022.
Goyal et al.
The choice of variance-based over covariance-based SEM (CB-SEM) rests on four methodological considerations.
First, the research objective is prediction and the identification of key driver constructs, rather than the confirmation of an established theory, which is the canonical use case for PLS-SEM (Hair et al.
, 2.
Second, the endogenous construct exhibited a non-normal distribution .
kewness = .
Oe.
, and PLS-SEM's nonparametric bootstrapping imposes no multivariate-normality assumption, whereas maximum-likelihood CB-SEM estimation is sensitive to its violation.
Third, the model integrates constructs adapted from heterogeneous source instruments for the first time in this population, making the exploratoryAepredictive orientation of PLSSEM more defensible than the strictly confirmatory logic of CB-SEM.
Fourth.
PLS-SEM maximizes explained variance in the target construct (RA.
QA), which aligns directly with the studyAos aim of rank-ordering predictors for intervention design.
CB-SEMAos global fit indices would not serve this purpose.
Common method bias was tested using Harman's single-factor test and the full collinearity VIF approach (Kock, 2.
Scope and Limitations of Methods Several methodological boundaries should be acknowledged.
First, the crosectional design precludes strong causal inference.
Second, reliance on self-report measures introduces potential social desirability and recall bias.
Third, geographic scope is restricted to Batam's industrial estates.
Fourth, the focus on female workers, by design, excludes comparisons with males.
Finally, although the instruments were rigorously validated, the PLV measure was adapted from international scales rather than developed Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
RESULTS AND DISCUSSION
Results Respondent Demographic Profile A total of 378 valid responses were retained for analysis.
The majority were aged 26Ae35 years .
4%), held a senior high-school qualification .
5%), worked in electronics manufacturing .
7%), and earned monthly wages within A10% of Batam's regional minimum wage of IDR 4,683,049 (BP Batam, 2.
Notably, 61.
4% reported borrowing from at least one online lending platform in the previous 12 months, and 8% reported experience with at least one unverified .
platform Ai confirming the contextual relevance of the study population.
Table 3.
Respondent Demographic Profile .
= .
Characteristic Age Education Industrial Sector Marital Status Tenure Online-loan use .
Category 18Ae25 26Ae35 36Ae45 Junior high school Senior high school Diploma/Bachelor Electronics Garment Plastics/Metal Single Married Divorced/Widowed < 2 years 2Ae5 years > 5 years Yes .
egal onl.
Yes .
Frequency Percentage Descriptive statistics for the five latent constructs are reported in Table 4.
Mean scores indicate that respondents reported moderate-to-high financial stress (M = 3.
= 0.
, moderate peer influence (M = 3.
SD = 0.
, and relatively low predatorylending vulnerability (M = 2.
SD = 1.
, though the high standard deviation on PLV signals substantial variation across the sample.
Table 4.
Descriptive Statistics of Latent Constructs .
= .
Construct Digital Financial Literacy (DFL) Financial Stress (FS) Financial Self-Efficacy (FSE) Peer Influence (PI) Predatory Lending Vulnerability (PLV) Min Max Mean Skewness Kurtosis Measurement Model (Outer Mode.
Evaluation Indicator-level outer loadings (Table .
ranged from .
712 to .
892, all exceeding the recommended threshold of .
70 (Hair et al.
, 2.
, indicating that each item shared adequate variance with its assigned construct.
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
Table 5.
Outer Loadings of Reflective Indicators Construct
DFL
FSE
PLV
Indicator DFL1
DFL2
DFL3
DFL4
DFL5
DFL6
DFL7
DFL8
DFL9
DFL10
DFL11
DFL12
FS1
FS2
FS3
FS4
FS5
FS6
FS7
FS8
FSE1
FSE2
FSE3
FSE4
FSE5
FSE6
PI1
PI2
PI3
PI4
PI5
PLV1
PLV2
PLV3
PLV4
PLV5
PLV6
PLV7
Loading t-value p-value < .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
< .
Internal consistency reliability and convergent validity are reported in Table 6.
All Cronbach's coefficients (.
858Ae.
DijkstraAeHenseler's rho_A (.
864Ae.
, and Composite Reliability (.
898Ae.
exceeded the .
70 threshold, and all AVE values (.
608Ae .
50, confirming convergent validity (Hair et al.
, 2.
Table 6.
Construct Reliability and Convergent Validity Construct Digital Financial Literacy (DFL) Financial Stress (FS) Financial Self-Efficacy (FSE) Peer Influence (PI) Predatory Lending Vulnerability (PLV) Threshold (Hair et al.
, 2.
Cronbach's AVE Ou .
Ou .
Ou .
Ou .
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
Discriminant validity was assessed using two approaches.
First, the FornellAeLarcker criterion (Table .
confirmed that the square root of each construct's AVE .
iagonal values, in bol.
exceeded its correlation with every other construct.
Table 7.
Discriminant Validity Ae FornellAeLarcker Criterion
DFL
DFL
FSE
PLV
FSE
PLV
Note.
Diagonal values .
old in original outpu.
= OoAVE.
off-diagonal values = construct Second, the HeterotraitAeMonotrait (HTMT) ratio (Table .
, considered a stricter test of discriminant validity (Henseler et al.
, 2.
, yielded values ranging from .
278 to .
all below the conservative .
85 cutoff.
The highest HTMT .
etween FS and PLV) signals a strong but distinct association, consistent with the structural model's expected strongest Table 8.
Discriminant Validity Ae HeterotraitAeMonotrait (HTMT) Ratio
FSE
PLV
DFL
FSE
Note.
All HTMT values < .
85 (Henseler et al.
, 2.
Common Method Bias and Collinearity Harman's single-factor test produced a first factor explaining 31.
7% of total variance Ai below the 50% threshold (Podsakoff et al.
, 2.
Full collinearity VIFs ranged 624Ae2.
, and inner model VIFs (Table .
421Ae2.
indicating that common method bias and multicollinearity are not concerns.
Table 9.
Inner Model Collinearity Statistics (VIF) Predictor
Digital Financial Literacy (DFL) Financial Stress (FS) Financial Self-Efficacy (FSE) Peer Influence (PI) Outcome PLV
PLV
PLV
PLV
VIF
Threshold < 5.
0 ue < 5.
0 ue < 5.
0 ue < 5.
0 ue Structural Model (Inner Mode.
Evaluation The structural model produced an RA of .
524 for predatory-lending vulnerability (Table .
, indicating that 52.
4% of the variance in PLV is jointly explained by DFL.
FS.
FSE, and PI Ai a substantial effect (Cohen, 1988.
Hair et al.
, 2.
The adjusted RA (.
closely matched the unadjusted value, suggesting that the model is not over-fitted.
QA of 342 .
ell above .
confirmed adequate predictive relevance.
Table 10.
Coefficient of Determination (RA) and Predictive Relevance (QA) Endogenous Construct Predatory Lending Vulnerability (PLV) RA RA Adjusted QA Interpretation Substantial/strong Note.
RA thresholds: .
19 weak, .
33 moderate, .
67 substantial (Chin, 1.
QA > 0 indicates predictive relevance (Hair et al.
, 2.
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
Effect sizes .
A) for each predictor are reported in Table 11.
Financial stress exerted a medium effect .
A = .
, peer influence a small-to-medium effect .
A = .
, digital financial literacy a small effect .
A = .
, and financial self-efficacy the smallest effect .
A = .
All four effects, however, were statistically significant and substantively Table 11.
Effect Sizes .
A) of Predictors on PLV Predictor Financial Stress (FS) Peer Influence (PI) Digital Financial Literacy (DFL) Financial Self-Efficacy (FSE) Effect-size category Medium Small-to-medium Small Small Rank Note.
fA thresholds: .
02 small, .
15 medium, .
35 large (Cohen, 1.
Hypothesis Testing Hypotheses were tested via bootstrapping with 5,000 resamples and bias-corrected 95% confidence intervals.
As summarised in Table 12, all four hypotheses were supported at p < .
001, with no confidence interval crossing zero.
Table 12.
Path Coefficients and Hypothesis Testing Results .
= .
Path DFL Ie PLV FS Ie PLV FSE Ie PLV PI Ie PLV Ae.
Ae.
< .
< .
< .
< .
Decision Supported Supported Supported Supported The strongest predictor of PLV was financial stress ( = .
, followed by peer influence ( = .
, digital financial literacy ( = Ae.
, and financial self-efficacy ( = Ae.
The two negative predictors (DFL and FSE) confirm a protective role.
the two positive predictors (FS and PI) confirm a risk-amplifying role.
Discussion Financial Stress as the Dominant Risk Driver The finding that financial stress is the strongest predictor of PLV ( = .
corroborates international evidence that economic strain pushes individuals toward high-cost credit as a coping mechanism (Choi et al.
, 2020.
Pizzutilo & Mariani, 2.
For female industrial workers in Batam, persistent financial stress is structurally produced:
stagnant wages relative to the cost of living, household-level expenses such as children's school fees and remittances, and unpredictable expenditure shocks.
This finding extends Zuhroh et al.
's .
Indonesian MSME results into the manufacturing-labor context and reinforces the feedback-loop hypothesis: stress Ie predatory borrowing Ie deepened Beyond corroborating this loop, the dominance of financial stress over the cognitive and psychological predictors is theoretically consistent with the scarcity framework advanced by Mullainathan and Shafir .
, which holds that acute financial scarcity imposes a Aubandwidth taxAy that captures attention and depletes the cognitive resources otherwise available for deliberate, forward-looking decision-making.
On this account, financial stress does not merely coexist with poor borrowing decisions.
actively narrows the decision frame to the immediate obligation, foregrounds present needs over future costs, and thereby crowds out the reflective evaluation on which digital financial literacy and financial self-efficacy depend.
For female industrial workers facing time-sensitive household shocks Ai unpaid school fees, medical emergencies, or delayed Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
remittances Ai the psychological urgency of resolving the shock can override an accurate appraisal of a lender's legitimacy or interest burden even when the requisite knowledge and confidence are present.
This explains why a resource that is economic in origin outperforms predictors that are cognitive (DFL) and psychological (FSE) in nature:
scarcity operates upstream of both, constraining the very capacity through which knowledge and self-efficacy would otherwise exert their protective effect.
The dominance of financial stress also acquires a distinctly gendered and contextual character among Indonesian female industrial workers that a generic scarcity account does not fully First, wages in Batam's export-oriented estates are anchored to the regional minimum wage, so most respondents operate at a structurally fixed income ceiling with virtually no scope for overtime-independent income growth.
financial shocks, therefore, cannot be absorbed through earnings adjustments and translate immediately into stress.
Second, the "double burden" positions women as de facto household financial managers who are culturally expected to reconcile school fees, food expenditure, and remittances to extended family, concentrating the psychological weight of household deficits on precisely the population studied here.
Third, many female workers are internal migrants with thin local collateral, limited formal credit histories, and weak eligibility for bank credit, so the feasible choice set during a shock often collapses to informal or digital lenders Ai making stress not only a motivational push but also a channel-selection Under these conditions, economic pressure plausibly dominates because it is chronic, gender-concentrated, and paired with restricted formal alternatives, a configuration less pronounced in the student and general-urban samples that dominate prior Indonesian studies.
Peer Influence as a Social Amplifier The significant positive effect of peer influence ( = .
supports Bandura's .
social cognitive theory and aligns with findings from Goyal et al.
and Ardini et al.
In the industrial context, workers cluster in dense social networks Ai co-workers in the same production line, dormitory mates.
WhatsApp groups Ai through which loan applications are recommended, downloaded together, and endorsed via referral codes.
Two mechanisms appear to operate: descriptive normalization .
hen peers borrow from pinjol, the practice is perceived as routin.
and information laundering .
ven unverified platforms gain legitimacy when introduced by trusted peer.
This finding aligns with Ribeiro and Cherobim's .
negative-effect result, suggesting that peer influence in low-information environments tends to amplify rather than mitigate financial risk.
The magnitude of this effect is best understood in relation to the distinctive social ecology of integrated industrial estates, which differs from the more diffuse networks of student or general-urban samples in three respects.
First, spatial and temporal density: assemblyline work, shared shift schedules, and dormitory co-residence produce sustained, highfrequency contact among workers occupying near-identical economic positions, so that a borrowing solution adopted by one worker is observed, discussed, and imitated within a tightly bounded reference group.
Second, economic homogeneity: because co-workers face common wage ceilings and synchronous expenditure shocks .
or example, simultaneous school-enrolment period.
, peer endorsement carries unusual persuasive weight Ai the recommending peer is seen as facing the same constraints and therefore as a credible source of a "tested" solution.
Third, closed-loop information channels: estatebased WhatsApp groups and referral-code economies circulate loan applications within a self-contained informational environment largely insulated from independent Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
verification, allowing unlicensed platforms to accrue legitimacy purely through repeated intra-network transmission.
Taken together, these features transform the ordinary sociallearning mechanism described by Bandura .
into an accelerated diffusion process in which risky credit products spread with limited friction, thereby explaining why peer influence emerges as the second-strongest determinant of vulnerability in this Digital Financial Literacy as a Protective Factor The negative effect of DFL on PLV ( = Ae.
confirms its protective role and is consistent with Setiawan et al.
Lyons and Kass-Hanna .
Burchi et al.
and Wahyuni et al.
Workers with higher DFL are better able to verify whether a lender is registered with OJK, recognize hidden fees, identify aggressive collection tactics, and distinguish licit pindar from illicit pinjol.
However, the modest effect size .
A = .
is theoretically informative: knowledge alone is necessary but insufficient.
This echoes the "knowledgeAeaction gap" reported by Kass-Hanna and Lyons .
The theoretical significance of this gap becomes clearer when read alongside the financial-stress finding.
Digital financial literacy is predominantly a Aucold-stateAy resource: it is acquired and rehearsed under calm conditions and presupposes the deliberative bandwidth required to apply it.
Borrowing decisions among stressed workers, however, are frequently made in Auhot,Ay affectively loaded states of scarcity in which that bandwidth is compromised (Mullainathan & Shafir, 2.
Financial knowledge can therefore raise the probability of a sound decision without guaranteeing it, because the structural pressure that motivates borrowing simultaneously erodes the cognitive conditions under which knowledge is This reframes the modest effect size not as a weakness of the DFL construct but as evidence of a boundary condition: education can shift the intention to avoid predatory credit, yet the intentionAebehavior gap widens precisely when economic pressure is most It follows that financial-literacy programs are likely to underperform when delivered in isolation and substantially more effective when paired with interventions that alleviate the underlying scarcity itself.
This inference directly informs the practical implications developed below.
Financial Self-Efficacy as a Psychological Buffer The negative path from FSE to PLV ( = Ae.
supports the findings of Mindra et al.
Asebedo and Payne .
, and Wahyuni et al.
that confidence in managing one's finances predicts more deliberate financial choices.
Workers with higher FSE are more likely to reject the perception that pinjol is the only option, to negotiate alternatives .
ooperative loans, wage advances, family borrowin.
, and to resist peerdriven herd behavior.
Although the fA of .
052 is small, the consistent direction reinforces the value of psychological resources alongside knowledge-based interventions.
Read together with the digital-financial-literacy result, this pattern indicates that the two protective resources share a common vulnerability: both are internal capacities whose expression depends on the availability of deliberative bandwidth, and both are therefore liable to be overridden when financial stress is high.
Integrated Interpretation Taken together, the results portray PLV as the product of an interaction between push factors .
inancial stress, peer influenc.
and pull factors .
ow DFL, low FSE).
The model accounts for 52.
4% of the variance Ai a substantial figure for a behavioral construct Ai yet leaves 47.
6% to other unexplored predictors, such as financial socialization in the Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
family of origin, religiosity, impulsivity, and structural workplace factors.
The dominance of financial stress over the four predictors is theoretically meaningful: cognitive and psychological interventions, while necessary, must be embedded within structural improvements in wages and household economic stability.
This unexplained share merits more systematic mapping as a foundation for subsequent research.
At the individual level, time preference and impulsivity, materialism, and attitudes toward debt are plausible omitted drivers of high-cost borrowing.
at the household level, intra-household bargaining power, spousal income volatility, and the number of dependants may condition both stress and the feasibility of alternatives.
at the environmental level, exposure to targeted digital-loan advertising, the density of informal lender networks around industrial estates, and access barriers to formal microcredit may shape the choice architecture within which vulnerability materialises.
Protective candidates Ai workplace social capital, participation in savings cooperatives or arisan rotating savings groups, and religiosity as a normative brake on interest-bearing debt Ai deserve equal attention, since interventions may be more tractable by strengthening buffers than by removing The dominance of financial stress over the four predictors is theoretically meaningful: cognitive and psychological interventions, while necessary, must be embedded within structural improvements in wages and household economic stability.
To move beyond a purely descriptive synthesis, the four predictors can be organized into a sequential pushAepull decision architecture (Figure .
that specifies how they interact rather than merely how they rank.
Within this architecture, financial stress functions as the activating condition: it generates the borrowing impulse and, through the scarcity mechanism described above, simultaneously lowers the deliberative capacity available to evaluate that impulse.
Peer influence then operates as the facilitating channel, converting a generalized impulse into a specific action by supplying an accessible, socially legitimated product and reducing the perceived search cost of finding one.
Digital financial literacy and financial self-efficacy act as regulatory gates positioned between impulse and action Ai DFL screening the product's legitimacy and cost, and FSE sustaining the confidence to pursue safer alternatives.
Crucially, these protective gates are not bypassed at random.
they are systematically weakened by the same financial stress that initiates the sequence, because scarcity depletes the bandwidth on which both depend.
The model is therefore better understood as an interaction than as a set of parallel main effects: the push factors do not simply add to the pull factors but condition their operation, such that the protective value of literacy and self-efficacy is greatest when stress is low and progressively attenuated as stress rises.
So conceived, predatory-lending vulnerability is the predictable outcome of a pressured impulse .
inancial stres.
channeled by social facilitation .
eer influenc.
through protective gates (DFL.
FSE) that scarcity has already partially disabled.
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
Figure 1.
Integrated pushAepull decision architecture of predatory-lending vulnerability.
Note.
Financial stress is the activating push that also depletes the deliberative bandwidth on which the protective gates (DFL.
FSE) depend.
peer influence is the facilitating channel.
values are standardized path coefficients.
Implications Theoretical Implications The study contributes to financial-vulnerability literature by providing the first integrated empirical test of cognitive (DFL), affective (FS), psychological (FSE), and social (PI) predictors of PLV within a single PLS-SEM model in Southeast Asia.
It extends Bandura's .
social cognitive theory into the predatory-lending domain and operationalizes O'Connor et al.
's .
consumer financial vulnerability framework specifically for digital predatory lending.
Practical Implications For OJK, the current education-only model should be redesigned to focus on the dominant predictor .
inancial stres.
Workplace-based financial-wellness programs should combine emergency-savings facilitation, employer-sponsored cooperative-credit alternatives, and short-burst micro-learning modules on DFL.
For employers, salary-ondemand schemes and partnerships with licensed pinjol could pre-empt the resort to illegal pinjol.
For trade unions and women's organizationsorganizations, peer-led "financial buddy" programs can turn peer influence into a protective force.
The rankordering of effects provides a direct template for prioritizing these interventions.
Because financial stress is the dominant driver ( = .
fA = .
, the highest-leverage measures are those that relieve liquidity pressure at its source Ai employer-facilitated emergency-savings schemes, salary-on-demand access, and cooperative-credit alternatives Ai rather than information provision alone.
Because peer influence is the second-strongest determinant ( = .
fA = .
and operates through estate-based networks, the second priority is to repurpose those same networks as protective infrastructure through peer-led Aufinancial buddyAy schemes and trained workplace financial champions, converting the diffusion mechanism identified above from a risk multiplier into a safeguard.
Digital financial literacy and financial self-efficacy, whose Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
smaller effects ( = Oe.
218 and Oe.
fA = .
078 and .
reflect their status as bandwidthdependent protective gates, are best delivered not as standalone curricula but as shortburst micro-learning embedded within the stress-relief and peer-based measures, so that knowledge and confidence are reinforced at the moment of decision rather than in isolation from it.
Sequencing interventions in this order aligns program design with the empirical weight of each predictor and with the decision architecture set out in Figure 1.
Policy Implications Findings strengthen the case for the demand-side regulatory shift signaled by OJK Regulation No.
11/2024 and the SLIK reporting requirement effective July 2025.
Targeted enforcement should focus on platforms operating in industrial-estate WhatsApp ecosystems, and gender-disaggregated supervisory data should be made publicly available (Women's World Banking, 2.
Consistent with the finding that vulnerability is concentrated where financial stress and peer diffusion intersect, enforcement resources are likely to yield the greatest return when directed at unlicensed platforms circulating within industrial-estate WhatsApp ecosystems, and when demand-side protection under OJK Regulation No.
11/2024 is paired with structural measures Ai wage-adequacy monitoring and employer-provided liquidity facilities Ai that address the economic pressure the model identifies as the primary driver.
CONCLUSION AND RECOMMENDATIONS
This study set out to examine the predictive roles of digital financial literacy, financial stress, financial self-efficacy, and peer influence on predatory-lending vulnerability among female industrial workers in Indonesia Ai a population that the Introduction identified as both empirically under-studied and structurally exposed to the proliferation of pinjol ilegal and informal high-cost lenders.
The expectations articulated in the Introduction have been substantively realized in the Results and Discussion.
All four hypotheses derived from social cognitive theory (Bandura, 1.
and the consumer financial-vulnerability framework (O'Connor et al.
, 2.
were empirically supported in a PLS-SEM analysis of 378 female workers in Batam's integrated industrial estates.
Together, the four predictors explained 52.
4% of the variance in PLV, confirming the appropriateness of the integrated cognitiveAeaffectiveAepsychologicalAesocial model.
Beyond confirming the hypothesized relationships, the study's theoretical contribution is twofold: it develops the consumer financial-vulnerability framework from a taxonomy of risk dimensions into a process model in which push forces condition the operation of protective resources.
It establishes a boundary condition for social cognitive theory by showing that knowledge- and efficacy-based self-regulation weakens precisely where economic pressure is most acute.
Three substantive conclusions emerge.
First, financial stress is the dominant driver of PLV ( = .
, validating the phenomenon-gap argument that structural economic pressure Ai not merely ignorance Ai pushes industrial women toward exploitative credit.
Second, peer influence operates as a social amplifier ( = .
, confirming that the dense workplace and dormitory networks of industrial estates function as conduits for the diffusion of risky financial products.
Third, digital financial literacy ( = Ae.
and financial self-efficacy ( = Ae.
function as protective but partial buffers, supporting yet qualifying the policy assumption that education-only interventions can reduce predatorylending exposure.
Copyright A 2026.
The Author.
This is an open-access article under the CCAeBY-SA license.
Journal of Economics and Management Vol.
4 No.
September 2026 | E-ISSN: 2987-7407 | DOI: https://doi.
org/10.
70716/ecoma.
The conclusions above should be read within the studyAos limitations, which are distinct from its future research agenda.
The cross-sectional design restricts causal claims to theoretically grounded associations.
self-reported borrowing Ai particularly from illegal platforms Ai is likely understated by social-desirability bias.
the single-region scope in Batam and the deliberate exclusion of male workers bound the population to which the estimates generalize.
and the predatory-lending-vulnerability measure, although rigorously validated, was adapted from international scales rather than developed indigenously.
These boundaries qualify the strength, not the direction, of the conclusions drawn.
Looking forward, the prospects for the development of these findings span three Empirically, the model invites longitudinal replication across other Indonesian industrial belts (Bekasi.
Tangerang.
Surabaya.
Cikaran.
and comparative testing in other Southeast Asian manufacturing economies.
Methodologically, future research can extend the present framework by incorporating mediatedAemoderated PLS-SEM, list-experiment techniques to overcome social desirability bias, and netnographic analysis of pinjolrelated digital communities.
In practical terms, the application prospects are immediate:
the findings can directly inform workplace-based financial-wellness curricula codesigned by OJK, the Ministry of Manpower, and industrial-estate operators, as well as employer-sponsored salary-on-demand and emergency-savings facilities that relieve liquidity pressure at the source.
The same evidence base can underpin peer-led "financial buddy" programs that convert workplace networks into protective infrastructure, as well as gender-sensitive supervisory data architectures aligned with OJK Regulation No.
11/2024 and the SLIK reporting requirement.
By translating the rank-ordered effect sizes into rank-ordered policy priorities, the study advances a research-to-policy pathway that is both gender-responsive and structurally informed Ai contributing to the realization of SDG 5 and SDG 8 in Indonesia's digital economy.
REFERENCES