Sage Journals: Discover world-class research

Abstract

This study empirically examines the relative effectiveness of print and digital media in promoting self-funded retirement plans as a pathway to financial inclusion in India. Using data from the NABARD All India Financial Inclusion Survey (NAFIS) 2016–2017, and applying propensity score matching to mitigate selection bias, we compare how exposure to print and digital media influences household participation in voluntary retirement schemes. The findings reveal that for every 1,000 households, approximately 4.5 households that accessed financial information through digital media subscribed to a self-funded retirement plan, compared to only 2.5 households reached through print media. Nearly 80 per cent of Indian households remain without any retirement plan, with substantial state-level disparities in subscription patterns. By clearly distinguishing between traditional and digital communication channels, this study advances understanding of how information delivery shapes financial behaviour. The findings contribute to the policy discourse on effective communication for financial inclusion, offering guidance for prioritising digital outreach in national financial literacy initiatives.

JEL Codes: G53, D14, I38, O16, L82

Keywords

Retirement Planning Self-funded Plan Promotion Channels Print Media Digital Media

1. Introduction

Approximately 90 per cent of India’s workforce is employed in the informal sector (Murthy, 2019), often without legal contracts or provisions for financial support in their post-working years. Recent demographic trends highlight an increasing concern about financial dependency among the elderly. According to an estimate in 2023 (Mohan et al., 2024), an estimated 52 per cent of older individuals are entirely dependent on others, with an additional 18 per cent being partially dependent. These concerns are further corroborated by our analysis of the NABARD Financial Inclusion Survey (2016–2017), which indicates that approximately 80 per cent of rural households in India lack any form of retirement plan to ensure financial security during old age.

In India, there are multiple retirement planning options. However, only 19 per cent of the households in the country have access to any form of continuous post-retirement income (see Table 1). A significant proportion of these households comprise government employees, for whom pension benefits are a major advantage. Alarmingly, only 2.6 per cent of the households have subscribed to a self-financed retirement plan, underscoring the need to investigate the factors contributing to this low uptake. Existing studies have highlighted several potential barriers, including low income levels (Kumar et al., 2019), behavioural biases (Mahapatra et al., 2019) and deficiencies in financial literacy (Agarwal et al., 2015). Furthermore, at a macroeconomic level, the prevalence of informal employment and the limited scope of formal work opportunities substantially hinder the adoption of structured retirement savings mechanisms (Ginneken, 1996).

Table 1.

Status of Retirement Planning in Rural India 2016–2017

State	Households (in %) have a Retirement Plan either by Gov., Org., or Self-financed			No Retirement Plan
State	Social Security	Employment Pension	Self-Financed	No Retirement Plan
Rural India	85.9	12.0	2.6	81.1
Maharashtra	40.4	49.8	12.9	93.6
Uttar Pradesh	82.3	12.3	5.5	77.3
Kerala	82.6	18.0	5.4	66.5
Assam	76.4	19.1	4.5	92.9
Karnataka	93.9	2.9	3.2	69.3
Manipur	71.0	27.8	2.3	84.3
Uttarakhand	84.6	13.4	2.1	62.5
Tamil Nadu	89.7	8.9	1.7	82.4
Telangana	95.4	3.3	1.3	62.7
Andhra Pradesh	95.7	3.2	1.1	73.4
Chhattisgarh	93.6	5.4	1.0	86.2
Jharkhand	84.2	14.9	0.8	77.0
Madhya Pradesh	81.7	17.5	0.8	91.8
Odisha	85.2	14.5	0.5	71.5
Jammu and Kashmir	47.4	52.8	0.5	76.3
Gujarat	64.5	35.2	0.5	95.8
Punjab	82.8	16.8	0.4	83.0
Haryana	86.6	14.5	0.4	80.6
Bihar	85.2	14.5	0.3	83.5
West Bengal	82.6	17.1	0.2	92.7
Himachal Pradesh	62.0	38.6	0.2	67.8
Rajasthan	95.7	4.2	0.2	78.7

Source: Estimated by the authors using NAFIS 2016–2017.

This study seeks to identify and evaluate the most effective communication channel¹—between print and digital media—for disseminating information about self-funded retirement plans in India. Determining the most effective channel is essential for three key reasons. First, it strengthens outreach and financial inclusion, as the medium of communication critically influences individuals’ awareness, trust and engagement with formal financial products. Prior studies have shown that accessible and credible communication enhances financial literacy and inclusion outcomes (Cole et al., 2011; Lusardi & Mitchell, 2014). Second, it ensures the optimal use of financial resources, allowing institutions such as the Pension Fund Regulatory and Development Authority (PFRDA) and private pension providers to maximise outreach and enrolment for every rupee spent on promotion. Finally, at the academic level, this study contributes to the literature by linking communication channels to household financial behaviour, offering empirical insights into how media choice can shape participation in voluntary retirement schemes.

Information about self-funded retirement plans can reach households through multiple channels, including print media, digital media, financial professionals (such as bankers or advisers), social networks and other locally adapted mediums, such as community events or traditional art forms used for financial awareness campaigns. The present study relies on self-declared household responses regarding the source through which they received information about retirement plans. However, institutional evidence suggests that formal promotional efforts for such schemes are primarily concentrated in print and digital media, which together account for the largest share of communication and outreach budgets across financial sector awareness programmes (PFRDA, 2021; RBI, 2021). Hence, this study focuses on comparing the effectiveness of print and digital media, which represent the formal, policy-driven avenues of outreach.

Employing propensity score matching (PSM), the research evaluates the impact of households’ exposure to their self-declared preferred media channel on the likelihood of subscribing to a self-funded retirement plan. The treatment group comprises households that primarily rely on either print or digital media, while the control group includes those that utilise alternative channels. A comprehensive set of covariates is controlled to ensure that the estimated effect reflects the isolated impact of media exposure on subscription behaviour.

The analysis demonstrates that digital media is significantly more effective in facilitating household subscriptions to self-funded retirement plans. Households accessing information through digital platforms were twice as likely to subscribe, with 4.5 households per 1,000 opting for a plan, as compared to only 2.5 households per 1,000 relying on print media. This notable disparity highlights the increasing prominence of digital platforms in disseminating financial information and fostering proactive retirement planning. The findings underscore the potential benefits of re-allocating promotional efforts and resources towards digital channels, which are likely to yield higher returns in terms of new subscriptions, particularly in the context of an increasingly digitalised environment.

The remainder of the study is structured as follows: Section 2 details the data sources utilised in the analysis. Section 3 introduces the research framework, establishing the foundation for the research design used to examine the causal impact of print and digital media on household subscriptions to self-funded retirement plans. Section 4 discusses the key findings derived from the analysis, while Section 5 highlights the limitations of the study, which are essential for accurately interpreting the results. Finally, the study concludes with policy recommendations informed by the findings.

2. Data Sources

This study leverages the NABARD All India Financial Inclusion Survey (NAFIS) 2016–2017, a comprehensive dataset offering unit-level information on 40,327 households across 2,016 villages/wards in 245 districts, covering all 29 Indian states. Conducted using a multi-stage stratified random sampling design, the survey primarily focuses on Tier-III to Tier-VI centres, encompassing rural and semi-urban areas with populations under 50,000.²^,³

The survey gathered data through a structured questionnaire, covering financial instruments, household assets, income, savings, loans, insurance, pensions and risks. It aimed to capture financial inclusion and livelihood contexts during a period of economic growth (6.1–5.6 per cent in the first two quarters of 2017) and historically low inflation (1.5 per cent in June 2017). To address recall bias, reference periods varied but were standardised to 365 days preceding the survey date for most questions.

The authors accessed the NAFIS data following a formal request to NABARD.

3. Research Framework

To construct the research framework and examine the causal relationship between household exposure to specific media channels and their subscription to self-funded retirement plans, the analysis commenced with a review of cross-sectional descriptive statistics. This was subsequently followed by hypothesis testing to validate the observed associations and evaluate the impact of the selected media channels on household financial behaviour.

3.1 Examining Self-funded Retirement Plans

The adoption of self-funded retirement plans in India is alarmingly low, with only 2.5 per cent of households participating (Table 1). In contrast, 57 per cent of households in the United States contributed to 401(k) programmes, as of 2022 (Holden et al., 2024). Substantial state-level variations exist within India: Maharashtra leads with 12.9 per cent participation, while Bihar and West Bengal correspondingly report only 0.2 per cent.

A χ² test confirmed that these interstate differences are statistically significant, underscoring the need for region-specific strategies. Centralised schemes like the National Pension System (NPS) must account for state-specific social, cultural and economic factors to design effective awareness campaigns tailored to regional contexts.

To address high standard errors in the confidence intervals, we adjusted lower limits to zero where necessary, ensuring robust analysis. Despite these adjustments, p values remained below the significance threshold, allowing us to reject the null hypothesis. A z-test further validated these findings, showing significant deviations in state proportions from the national average, highlighting the need for tailored interventions. Detailed results are summarised in Table 2.

Table 2.

Summary of Hypothesis Tests

Sl No.	Hypothesis	Relevant Test	Outcome	Significance Level	Implication
1	Across states, the same proportion of households have self-funded retirement plan.	χ ²	Rejected	Significant	State-specific economic, social and cultural factors may influence households’ decisions.
2	State-level household self-funded retirement plan proportionate ownership is equal to the national average	z-test	Rejected	Significant	We need state-specific content to promote self-funded retirement plan.
3	There is either no or a positive association between self-funded retirement plan and household having no retirement plan	t-test	Not rejected	Not significant	Outcome is strange; however, reasons might be due to a large base effect, where a large number of households have no retirement plan across the states. However, at the same time, a very small proportion of households are able to afford a self-funded plan; hence data are not able to capture.
4	There is either no or a positive association between social security and households having no retirement plan	t-test	Rejected	Significant	A large percentage of households rely on social security as a supporting income and help in reducing households without any social security.
5	Households uniformly use all the channels	χ ²	Rejected	Significant	There is a significant variation across channels, which needs to be studied further to find an effective channel to use resources efficiently.
6	Print and digital media are equally effective	PSM	Rejected	Covariates are controlled	Digital media is more effective than print media, almost double. PFRDA, as well as private players, spent more resources on digital media.

Source: Prepared by the authors using NAFIS 2016–2017.

Given the inherent limitations of secondary data, it is not possible to explore all kinds of relationships. However, to continue to develop our framework, we examined correlations between the proportion of households with self-funded retirement plans and the decline in households without any plan. Although a positive correlation was observed, it was weak and statistically insignificant, likely due to pervasive income disparities. High-income households are more likely to adopt retirement plans, while the majority of India’s workforce, employed in the informal sector and earning marginally above subsistence levels, lacks the financial capacity to do so (Ghilarducci et al., 2022).

Financial literacy also plays a pivotal role in the low demand for self-funded retirement plans (Clark et al., 2012; Lusardi & Mitchelli, 2007). Limited access to financial services in low- and middle-income countries further constrains adoption (Barik & Lenka, 2022). In India, this reflects a transitional phase where adoption is growing but has yet to result in a proportional decline in households without any retirement plans (Yang & Ching, 2014).

Income remains the most critical determinant. Testing the hypothesis that social security availability negatively correlates with the proportion of households without retirement plans, we found a significant negative relationship. Over 50 per cent of Indian households earn less than ₹10,000 monthly (Jha & Basole, 2023), and 90 per cent of the workforce is employed in the informal sector, which severely limits retirement plan adoption (Murthy, 2019).

3.2 Sources of Financial Information

The NAFIS survey asked households about their sources of information for banking and financial products, including print media, digital media, social networks, agents and others. Although households could select multiple sources, the frequency of selections declined significantly after the primary source, with only 0.6 per cent identifying all six options. This pattern highlights households’ prioritisation of influential information sources.

For this study, the first reported source was treated as the primary one, reflecting the most impactful medium influencing households’ financial decisions (Thaler & Sunstein, 2021). Cross-verification across multiple financial products validated the reliability of this approach. Social networks strongly influence microfinance adoption (Banerjee et al., 2013), while agents and professionals are key for products like fixed deposits and mutual funds. Analysing the proportion of households aware of specific financial products and their sources (Table 3) confirmed the consistency of the first reported source.

Across financial products, print media emerged as the dominant source, followed by social networks and digital media. In recent years, however, digital platforms have gained prominence, particularly in online financial transactions. For self-funded retirement plans, approximately 60 per cent of households identified print media as their primary source, followed by 20 per cent citing social networks and a similar proportion using digital media.

Table 3.

Households’ Information Sources for Financial Products

Financial Products	Print Media	Digital Media	Social Networks	Professional	Others
Bank deposits like FD and RD	49.2	14.3	31.6	3.4	1.4
Credit/loan from bank	50.7	14.1	32.3	2.2	0.6
Credit card/Kisan credit card	55.1	13.6	29.6	1.3	0.4
Govt. deposits being run through post office like KVP/NSC, etc.	49.6	15.4	31.4	2.5	1.1
Credit/ loan from a microfinance institution/SHG/JLG, etc.	39.6	15.7	40.9	2.5	1.3
Crop insurance	46.9	14.9	33.2	2.7	2.3
Other insurance	48.2	15.2	33	2.3	1.3
Shares/stocks/bonds/mutual funds/SIP	55.7	11.9	28	3	1.4
Mobile/internet banking	57.7	16.6	22.3	2.4	1
Mobile wallets (Paytm, Freecharge, Mpesa, etc.)	63.8	18.7	15.3	1.9	0.3
Pension/provident fund	46.9	13.6	36.2	2.2	1.2
Self-funded retirement plan	56.5	19.6	22	1.4	0.5

Source: Estimated by the authors using NAFIS 2016–2017.

A χ2 test revealed significant variation in households’ access to in- formation about self-funded retirement plans across different channels, emphasising the distinct characteristics of each source. This insight supports targeted resource allocation to maximise impact.

While social networks play a meaningful role in disseminating financial information—often ranking as the second most prominent or, in some cases, the dominant informal source (see Table 3)—they are inherently unstructured, context-dependent and difficult to quantify within a formal analytical framework. Their inclusion would also require a detailed behavioural or network analysis that extends beyond the scope and word constraints of the present article. Similarly, professional agents, such as bankers, insurance advisers or finance professionals, though influential in promoting financial products, operate primarily through personalised advisory rather than through mass communication mechanisms. Since the study’s objective is to compare formal communication channels—those that disseminate information at scale and are directly shaped by institutional outreach strategies—the professional agent channel is excluded from the empirical analysis. Therefore, the exclusion of both social networks and professional agents does not imply their ineffectiveness; rather, it reflects the focus of the study on policy-driven, resource-intensive communication channels—specifically print and digital media—that can be systematically evaluated and optimised for improving financial inclusion.

3.3 Research Design

This study examines the impact of print and digital media on households’ likelihood of subscribing to self-funded retirement plans. Recognising the limitations of cross-sectional data in establishing causality, we employ statistical matching, a robust method widely used in social science research to evaluate interventions or policies. By matching households based on a similarity function derived from comprehensive covariates, we minimise biases and enhance the robustness of our findings (Thoemmes & Kim, 2011).

Statistical matching offers flexibility over methods like instrumental variable (IV) estimation, particularly in capturing non-linear relationships and addressing limited overlap in covariate distributions. PSM, a cornerstone of our analysis, calculates the probability of a household being exposed to print or digital media based on pre-treatment characteristics such as income, assets and demographics. This score enables the creation of matched pairs of media-exposed (treated) and non-exposed (control) households with similar attributes, significantly reducing bias and improving the accuracy of comparisons (Rosenbaum & Rubin, 1983).

The analysis unfolds in two stages. First, we estimate the effect of print media on subscriptions to self-funded retirement plans. In the second stage, using a consistent set of covariates, we evaluate the impact of digital media, allowing for a comparative assessment of both channels. Throughout the study, the term ‘preferable media’ is used, with specific references to print or digital media as required for clarity.

p (X) = P r {D = 1 | X} = E {D | X} .

(1)

In this context, in Equation (1), D = 0,1 represents a binary variable indicating whether a household used the preferable media (D = 1) or not (D = 0). X denotes a multi-dimensional vector of pre-treatment characteristics, encompassing both media related and household attributes. As demonstrated by Rosenbaum and Rubin (1983), if the allocation of media exposure is random within groups defined by specific characteristics (X), it is equally random within groups defined by the propensity score, p(X).

Subsequently, the effect of preferable media can be estimated as follows:

τ \equiv E {W_{1 i} - W_{} 01 | D_{i} = 1},

(2)

= E {E {W_{1 i} - W_{} 0 i | D_{i}_{} = 1, p (X i)}},

(3)

= E {E {W_{1 i} | D_{i}_{} = 1, p (X_{i})} - E {W_{0 i} | D_{i}_{} = 0, p (X_{i})} | D_{i} = 1} .

(4)

The subscript i in Equations (2), (3) and (4) denotes the ith household. The variable W_i represents the potential outcome, capturing the household’s subscription decision under two counterfactual scenarios: one where the household is exposed to the preferable media and another where it is not exposed.

Equation (2) defines the effect of preferred media as the expected difference in the subscription decision for the ith household when exposed to preferred media, compared to the counterfactual scenario in which the household lacks such exposure.

Equation (3) extends this concept by calculating the expected exposure effect across the distribution of the propensity score. The propensity score represents the likelihood of a household being exposed to preferable media based on its pre-treatment characteristics.

Equation (4) refines this further, representing the exposure effect as the expected difference in the subscription decision for the ith household when exposed to the preferable media versus without exposure. Both scenarios are evaluated under the same distribution of the probability of exposure to the preferable media.

To apply PSM effectively, two key assumptions must be satisfied: the balancing hypothesis and the ‘unconfoundedness’ assumption. These assumptions are critical for Equation (4) to hold, as they ensure that treatment assignment is independent of the potential outcomes, conditional on the observed covariates. The subsequent section will explore the process of achieving covariate balancing.

3.3.1 Covariate Balancing

Estimating the impact of preferable media on household subscription to self-funded retirement plans is challenging due to potential differences in unobserved characteristics between subscribing and non-subscribing households. For instance, households using preferred media often include older individuals (Nimrod, 2017), those with higher education, income and wealth (Scherer & Naab, 2009) and greater financial literacy (Ugargol et al., 2023), all of which can pre-dispose them to subscribe (Murari et al., 2021; Rooj & Sengupta, 2025). Thus, observed associations may reflect these characteristics rather than media exposure itself.

PSM addresses these concerns by creating comparable counterfactual groups based on similar characteristics (Gertler et al., 2016; Rosenbaum & Rubin, 1983). To ensure accurate matching, we included media-related covariates such as ownership of TVs, mobile phones and housing conditions. These factors strongly influence media usage, especially in contexts like India, where infrastructure constraints can limit access to digital media (a comprehensive list of variables used for the analysis is presented in Table 4).

Table 4.

Variable Description and Summary Statistics

Variable	Definition	Obs	Mean	Std Dev
OutcomeSelf PensionPlan	Households reported to have subscribed (or receiving) self-funded retirement plan: 1 Yes; 0 No	33,801	0.006	0.078
PriMedia	Print media as a preferable source of information about various banking or financial products or services: 1 Print media; 0 Other sources excluding digital media	34,389	0.515	0.500
DigiMedia	Digital media as a preferable source of information about various banking or financial products or services: 1 Digital media; 0 Other sources excluding print media	22,622	0.262	0.440
MPCY	Monthly per capita income	40,327	2,492	4,410
HeadGender	Gender of the head of the household: male and female	40,327	1.1	0.3
HeadAge	Age of the head of the household	40,327	48	13
HeadEdu5	Education of the head of the household	40,327	2.4	1.1
HeadOccu	Occupation of the head of the household	40,327	2.3	1.5
SRG	Household combined social and religious identity	40,327	3.378	2.365
RationCard	Government-recognised household economic status	40,327	1.563	0.619
Vehicle	Household ownership of a personal vehicle	40,327	1.374	0.558
MobilePhone	Household ownership of basic (landline) or mobile phone	40,327	0.868	0.339
TV	Household ownership of TV (any type)	40,327	0.587	0.492
House	Household status (owned or rented) and the condition (katcha or pakka) of the house	40,327	3.169	1.766
HHDebt	Household have any outstanding debt	40,327	0.771	0.420
KisanCreditCard	Any member of the household has a valid Kisan Credit Card	40,327	0.129	0.335
InflationPerception	Household perception about inflation that it increases the cost of living: 0 Don’t know; 1 Yes (True); 2 No (False)	40,327	1.154	0.476
NoInterestSaving	Experience satisfaction in spending money rather than saving it for the long term: 1 Completely agree, 2 partially agree and disagree, 3 Nor agree or disagree, 4 Completely disagree	40,327	2.275	1.190
Emergency Fund3Months	Household set aside emergency or rainy-day funds that would cover your expenses for 3 months	40,327	0.244	0.430
Sector	Household location: 1 Rural, 2 Urban	40,327	1.140	0.347
StateRe	Clubbed the states by regions	40,327	3.371	1.782

Source: Prepared by the authors using NAFIS 2016–2017.

We adopted a matching strategy where households exposed to preferable media were paired with those of similar propensity scores but not subscribing to self-funded plans. The propensity scores, derived from a probit regression model, were based on household and media characteristics. By balancing observed covariates between treatment (media-exposed) and control groups, we attributed differences in subscription likelihood to media exposure (Caliendo & Kopeinig, 2008).

Two set-ups were tested for robustness. In one, the control group included all other media channels. In the other, dominant media types (e.g., digital when evaluating print) were excluded from the control group to assess the ‘extant effect’. While both set-ups were robust, results for the excluded media approach are reported, with full-sample results available upon request. This strategy isolates each media type’s unique impact and highlights its specific contribution. Descriptive statistics reveal significant differences between households exposed to preferable media and those not exposed. Treated households generally have higher incomes, better housing, greater financial literacy and more media access (e.g., TV and mobile phones). These characteristics were balanced in the matched sample (refer to Table 5), satisfying the assumption of ‘unconfoundedness’ (Becker & Ichino, 2002).

Table 5.

Covariate Balancing

List of Variables	Preferable Media: Print (Treatment Group). From the Control Group, Digital Media Excluded.						Preferable Media: Digital (Treatment Group). From the Control Group, Print Media Excluded.
	Probit Model		Matched Sample				Probit Model		Matched Sample
	Coef	P>z	Exposed	Unexposed	t-test	p > t	Coef	p > z	Exposed	Unexposed	t-test	p > t
	Coef	P>z	Treated	Control	t-test	p > t	Coef	p > z	Treated	Control	t-test	p > t
A	B	C	D	E	F	G	H	I	J	K	L	M
MPCY	−3.03E−06	0.103	2,594	2,617	−0.420	0.674	4.68E–07	0.855	2,631	2,628	0.030	0.974
Ration card type: 1 BPL
2 APL	0.040	0.017	0.473	0.475	−0.220	0.827	0.067	0.003	0.449	0.445	0.430	0.668
3 Antyodaya	0.173	0.000	0.073	0.071	0.850	0.394	0.113	0.009	0.062	0.061	0.310	0.754
House status and condition: 1 Kutcha owned
2 Kutcha rented	−0.051	0.371	0.016	0.017	−0.520	0.604	0.029	0.681	0.020	0.021	−0.120	0.903
3 Semi-pucca owned	0.105	0.000	0.223	0.214	1.980	0.306	0.129	0.000	0.250	0.248	0.220	0.828
4 Semi-pucca rented	0.160	0.002	0.026	0.025	0.840	0.399	0.035	0.606	0.024	0.024	−0.100	0.919
5 Pucca owned	0.034	0.108	0.401	0.407	−1.070	0.284	0.057	0.042	0.366	0.363	0.280	0.776
6 Pucca rented	0.079	0.089	0.039	0.040	−0.490	0.627	0.085	0.161	0.035	0.037	−0.520	0.601
Personal vehicle: 1 None
2 Two-wheeler	0.089	0.000	0.343	0.338	0.890	0.376	0.041	0.096	0.322	0.316	0.570	0.568
3 Car	0.258	0.000	0.051	0.053	−0.530	0.598	0.004	0.954	0.037	0.036	0.180	0.853
Gender head: 1 Male
2 Female	−0.164	0.000	0.076	0.074	0.580	0.560	−0.192	0.000	0.082	0.082	0.030	0.975
Head age: 1 ≤25
2 >25 and ≤35	0.017	0.711	0.193	0.192	0.050	0.962	−0.047	0.438	0.192	0.194	−0.220	0.829
3 >35 and ≤45	0.091	0.047	0.308	0.305	0.550	0.584	0.030	0.621	0.319	0.320	−0.030	0.978
4 >45 and ≤60	0.114	0.013	0.366	0.365	0.100	0.923	−0.008	0.900	0.363	0.361	0.200	0.841
5 >60	0.097	0.057	0.107	0.109	−0.730	0.463	−0.070	0.302	0.098	0.097	0.140	0.892
Socio-religious group: 1 Hindu OBC
2 High caste Hindu	0.085	0.000	0.216	0.211	0.870	0.382	0.115	0.000	0.187	0.186	0.170	0.867
3 Hindu SC	−0.091	0.000	0.162	0.159	0.510	0.611	0.024	0.453	0.163	0.166	−0.420	0.672
4 Hindu ST	−0.175	0.000	0.106	0.101	1.280	0.201	−0.105	0.003	0.107	0.110	−0.440	0.660
5 Muslim general	0.003	0.949	0.036	0.038	−0.810	0.420	0.157	0.004	0.045	0.043	0.310	0.756
6 Muslim OBC, SC and ST	0.030	0.413	0.057	0.059	−0.530	0.596	0.039	0.446	0.048	0.046	0.270	0.784
7 Other minority general	0.199	0.000	0.034	0.036	−0.980	0.327	0.077	0.343	0.022	0.023	−0.110	0.912
8 Other minority SC, ST and OBC	−0.140	0.000	0.101	0.106	−1.430	0.154	0.007	0.842	0.177	0.177	0.080	0.938
Head edu.: 1 Illiterate
2 Primary	0.093	0.000	0.223	0.228	−0.950	0.342	0.057	0.043	0.261	0.261	0.000	0.998
3 Junior and secondary	0.218	0.000	0.392	0.388	0.630	0.526	0.041	0.136	0.361	0.358	0.350	0.724
4 Senior secondary	0.288	0.000	0.093	0.091	0.470	0.637	0.076	0.085	0.083	0.083	0.130	0.897
5 Grad and above	0.415	0.000	0.079	0.081	−0.680	0.498	0.196	0.000	0.064	0.064	0.050	0.958
Head occup.: 1 Business
2 Agri and non-agri labour	−0.042	0.023	0.292	0.292	0.060	0.955	0.000	0.992	0.339	0.341	−0.150	0.882
3 Household work	0.027	0.449	0.064	0.067	−1.280	0.201	0.000	0.994	0.060	0.059	0.080	0.939
4 Rent, remit and pension	0.054	0.513	0.010	0.010	0.230	0.822	0.030	0.778	0.009	0.009	0.260	0.794
5 Wages and salaried	0.048	0.046	0.173	0.166	1.570	0.117	0.006	0.855	0.157	0.155	0.260	0.793
6 Others	−0.008	0.895	0.016	0.016	−0.570	0.567	−0.030	0.728	0.014	0.014	−0.100	0.923
TV: 0: No, 1: Yes	0.034	0.072	0.613	0.606	1.360	0.174	0.278	0.000	0.690	0.682	0.880	0.380
Mobile or landline: 0: No, 1: Yes	0.289	0.000	0.915	0.914	0.360	0.716	0.158	0.000	0.897	0.894	0.440	0.657
HH debt: 0: No, 1: Yes	−0.041	0.026	0.777	0.777	−0.020	0.988	−0.024	0.319	0.780	0.778	0.180	0.856
Any insurance: 0: No, 1: Yes	0.103	0.000	0.257	0.239	3.470	0.023	0.085	0.001	0.239	0.233	0.650	0.515
Kisan cr. card: 0: No, 1: Yes	0.142	0.000	0.154	0.156	−0.360	0.722	0.080	0.026	0.109	0.105	0.800	0.422
Understand inflation: 0 Don’t know
1 True	0.423	0.000	0.765	0.765	0.120	0.902	0.328	0.000	0.761	0.761	0.050	0.957
2 False	0.459	0.000	0.208	0.206	0.570	0.567	0.337	0.000	0.201	0.199	0.280	0.779
Spending than saving in the long-run: 1 Completely agree
2 Partially agree and disagree	−0.044	0.021	0.297	0.290	1.240	0.214	0.083	0.001	0.376	0.376	0.030	0.980
3 Nor agree or disagree	0.006	0.859	0.075	0.072	0.910	0.363	−0.070	0.088	0.069	0.069	0.150	0.878
4 Completely disagree	0.094	0.000	0.297	0.286	1.900	0.180	−0.029	0.298	0.221	0.216	0.540	0.590
Emergency fund for 3 Months: 0: No, 1: Yes	0.292	0.000	0.292	0.284	1.600	0.109	0.008	0.740	0.218	0.215	0.360	0.717
Sector: 1: Rural, 2: Urban	0.193	0.000	0.167	0.162	1.270	0.203	0.075	0.018	0.144	0.142	0.380	0.706
State region: 1 North
2 Central	−0.303	0.000	0.137	0.134	0.570	0.572	−0.405	0.000	0.089	0.090	−0.200	0.839
3 East	−0.307	0.000	0.182	0.183	−0.300	0.762	−0.196	0.000	0.165	0.167	−0.190	0.848
4 West	−0.204	0.000	0.142	0.124	4.430	0.660	0.024	0.543	0.143	0.139	0.620	0.533
5 North-East	−0.567	0.000	0.139	0.146	−1.720	0.266	−0.107	0.005	0.254	0.256	−0.330	0.743
6 South	−0.597	0.000	0.126	0.127	−0.460	0.649	−0.322	0.000	0.141	0.148	−1.060	0.289
_Cons	−0.720	0.000					−1.252	0.000
Sample size	28,721						19,115

Source: Estimated by the authors using NAFIS 2016–2017.

After matching, t-tests confirmed no significant differences in covariates between treatment and control groups, reinforcing the robustness of the approach. We used nearest neighbour and Kernel matching to estimate the average treatment effect on the treated, as these methods address limitations like missing data and sensitivity to radius parameters (Austin, 2014). Full methodological details and supplementary results are available upon request.

4. Results

4.1 Findings

This section presents the treatment effects estimated using matching methods to assess the influence of print and digital media on household subscriptions to self-funded retirement plans.

The results (Table 6) reveal a significant positive effect of both print and digital media. The analysis is divided into two sets as follows:

Table 6.

Results of PSM (Effects of Preferable Media on the Subscription of Self-funded Retirement Plan)

Media Type	Treated	Control	Average Treatment Effect on Treated (ATT)	SE	t-value	Average Treatment Effect on Treated (ATT)/1,000 Households	% Change
Set 1
Panel A: Nearest neighbour matching
Print media (digital media excluded from the control group)	14,035	14,686	0.002	0.001	1.84	2.5
Digital (print media excluded from the control group)	5,080	14,035	0.005	0.002	2.64	4.5	84.7
Panel B: Kernel matching
Print media (digital media excluded from the control group)	14,686	14,035	0.003	0.001	2.75	2.7
Digital (print media excluded from the control group)	5,080	14,035	0.004	0.001	2.83	3.9	44.8
Set 2
Panel C: Nearest neighbour matching
Print media (digital media included in the control group)	14,686	19,115	0.002	0.001	1.42	1.8
Digital (print media included in the control group)	5,080	28,721	0.003	0.002	1.59	2.8	55.7
Panel D: Kernel matching
Print media (digital media included in the control group)	14,686	19,115	0.002	0.001	2.06	1.9
Digital (print media included in the control group)	5,080	28,721	0.002	0.001	1.66	2.2	14.8

Source: Estimated by the authors using NAFIS 2016–2017.

Set 1: The control group excludes digital media when evaluating print media and vice versa.

Set 2: Both media channels are included in the control group when analysing each as a treatment.

In Set 1 (panel A), using the nearest neighbour method, for every 1,000 households surveyed, approximately 2.5 accessed information via print media to subscribe to a self-funded retirement plan. For digital media, this figure is 4.5 households, an effect about 85 per cent higher, indicating digital media’s greater influence.

In Set 2, where both media channels are included in the control group, print media influenced 2 households per 1,000, while the effect of digital media was about 50 per cent higher. The lower effect size in Set 2 may stem from increased competition among media channels or complementary effects when print and digital media are jointly considered.

The kernel matching algorithm corroborates these findings, with treatment effects similar to those from nearest neighbour matching but slightly lower for digital media (14 percentage points). Kernel matching also yielded smaller standard errors, suggesting more precise estimates.

Overall, the results reinforce that digital media plays a more prominent role than print media in influencing households’ financial decisions regarding self-funded retirement plans, even as print media retains importance in certain contexts.

4.2 Robustness Test

To validate the causality analysis suggesting that digital media is more influential than print media, we conducted Rosenbaum’s test, which evaluates the potential bias introduced by unobserved confounders. This test calculates the bounds of hypothetical treatment effects within a specified range, comparing these bounds to the observed treatment effect from PSM. If the observed effect falls within the bounds, the estimate is robust, suggesting a true causal relationship. Conversely, if it falls outside, unobserved confounders may significantly influence the treatment effect, raising questions about its validity.

Given the binary nature of the treatment and outcome variables, we used the ‘Rbounds’ command in Stata to perform the test, assuming that the propensity score follows a Gamma distribution. Gamma values ranging from 1 to 2, in intervals of 0.2, were used to assess sensitivity. The results (Table 7) showed statistically significant estimates for both print and digital media. For instance, a Gamma value of 1.2 indicates that households using print or digital media are 1.2 times more likely to access financial information as compared to others, attributable to unobserved covariates among households with similar matched characteristics.

Table 7.

Rosenbaum Bounds for Propensity Score

Gamma	sig+	sig−	t-hat+	t-hat−	CI+	CI−
Print media (N = 28,721 matched pairs)
1	0.000	0.000	0.512	0.512	0.510	0.514
1.2	0.000	0.000	0.498	0.526	0.496	0.528
1.4	0.000	0.000	0.486	0.538	0.484	0.540
1.6	0.000	0.000	0.476	0.548	0.474	0.550
1.8	0.000	0.000	0.467	0.557	0.465	0.559
2	0.000	0.000	0.460	0.564	0.458	0.566
Digital media (N = 19,115 matched pairs)
1	0.000	0.000	0.264	0.264	0.263	0.266
1.2	0.000	0.000	0.256	0.272	0.254	0.274
1.4	0.000	0.000	0.249	0.280	0.247	0.281
1.6	0.000	0.000	0.243	0.286	0.241	0.287
1.8	0.000	0.000	0.237	0.291	0.236	0.293
2	0.000	0.000	0.233	0.296	0.231	0.298

Source: Estimated by the authors using NAFIS 2016–2017.

Notes: gamma −log odds of differential assignment due to unobserved factors; sig+: upper bound significance level; sig−: lower bound significance level; t-hat+: upper bound Hodges–Lehmann point estimate; t-hat−: lower bound Hodges–Lehmann point estimate; CI+: upper bound confidence interval (a = 0.95). CI−: lower bound confidence interval (a = 0.95).

These findings suggest that while a significant association exists between media use and household subscription to self-funded retirement plans, unobserved factors introduce potential bias, thereby complicating causal interpretation. Thus, while the results strengthen the evidence of association, causal claims should be made cautiously, acknowledging the limitations posed by unobserved variables.

5. Limitations of the Study

This study faces several limitations. First, it employs the same set of covariates for both print and digital media channels to control for confounding effects. While these determinants are generally relevant to self-funded retirement plan subscriptions, they may not fully capture factors unique to each channel, potentially introducing omitted variable bias. Moreover, the study is also not able to differentiate between various modes within print and digital media due to data constraints.

The data used are somewhat outdated, predating significant reforms to the NPS since 2014. Reforms such as the introduction of the Atal Pension Yojana for the unorganised sector, enhanced flexibility in entry and exit rules, tax incentives, increased equity allocation options, portability and the ability to select fund managers could significantly influence household decisions. As these reforms are not reflected in the data, the treatment effects may be under-estimated. Additionally, findings are not fully generalisable to all of India, as the survey primarily covers rural and semi-urban areas (Tier-III to Tier-VI centres), though these account for about 70 per cent of the population.

The study estimates a higher impact of digital media than print media but cannot explore the mechanisms underlying the greater influence of digital media on financial decisions. Factors such as social networks and professional agents, which may also serve as effective channels for information dissemination, have not been analysed in detail due to space constraints. Exploring these channels would require separate, in-depth investigations.

Finally, ambiguity in classifying media exposure—such as whether households accessed newspapers via mobile phones or computers and whether this was recorded as digital or print media—may affect the accuracy of media classification in the analysis.

6. Conclusions and Policy Implications

India, a country of vast size, faces significant challenges in disseminating financial information across its diverse population. Traditional methods of communication, particularly print media, have high costs associated with them, making broad outreach difficult. However, with the advent of consumer-friendly technologies, digital media presents a cost-effective alternative for disseminating important financial information, particularly in rural areas. In our empirical analysis, we found that digital media is almost twice as effective as print media in reaching households and encouraging them to subscribe to self-funded retirement plans. Using PSM to address selection bias, we estimated that for every 1,000 households, approximately 4.5 households accessed financial information through digital platforms and subsequently subscribed to self-funded retirement plans, as compared to just 2.5 households using print media.

Furthermore, when comparing rural India’s self-funded retirement plan adoption rates at the national level with state-level estimations, we observed significant differences. Although these findings warrant further statistical exploration, they provide valuable policy insights. Specifically, besides allocating more resources to digital media, it is critical that content be made available in all Indian languages and tailored to local dialects and colloquial expressions to ensure effective outreach. For example, while the PFRDA’s YouTube channel primarily offers content in English and Hindi, this limits accessibility for large segments of the population who communicate in regional languages. This linguistic mismatch may partly explain the relatively low engagement with official digital platforms, as compared to the growing popularity of financial influencers (finfluencers) who create regionally customised, relatable and language-specific content. Recognising the rising influence of such digital communicators, future research could systematically examine how finfluencers complement or substitute formal institutional outreach, especially in promoting financial literacy and retirement planning among younger and rural populations. Clearly, a more inclusive and evidence-driven digital outreach strategy is necessary to expand participation in self-funded retirement plans.

The findings of this study carry important implications for policymakers. First, there is a clear need to allocate greater financial resources to digital media platforms. Second, emerging technologies should be leveraged not only to maximise the efficiency of limited financial resources but also to effectively reach India’s largest demographic group, ‘Gen-Z’, who are more likely to engage with digital content. Policy efforts should focus on enhancing digital literacy, ensuring that digital tools are accessible across rural India, and tailoring financial content to meet the diverse linguistic and cultural needs of the population.

Footnotes

Acknowledgements

The authors gratefully acknowledge Mr Vinay Jadhav and Mr Ashutosh Kumar at NABARD for providing access to the NABARD-NAFIS data, and the anonymous reviewers for their detailed and constructive feedback that helped improve the clarity and readability of the manuscript.

Data Availability

The datasets used and/or analysed during the current study can be accessed from NABARD while writing at these e-mail IDs: dear@nabard.org; ashutosh.kumar@nabard.org; dear.admin@nabard.org.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Ethical Approval

This article does not contain any studies with direct human participants by any of the authors.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Notes

References

Agarwal

, Amromin

, Ben-David

, Chomsisengphet

, & Evanoff

D. D.

(2015). Financial literacy and financial planning: Evidence from India. Journal of Housing Economics, 27, 4–21.

Austin

P. C.

(2014). A comparison of 12 algorithms for matching on the propensity score. Statistics in Medicine, 33(6), 1057–1069.

Banerjee

, Chandrasekhar

A. G.

, Duflo

, & Jackson

M. O.

(2013). The diffusion of microfinance. Science, 341(6144), 1236498.

Barik

, & Lenka

S. K.

(2022). Demand-side and supply-side determinants of financial inclusion in Indian states: Evidence from post-liberalization period. Emerging Economy Studies, 8(1), 7–25.

Becker

S. O.

, & Ichino

(2002). Estimation of average treatment effects based on propensity scores. The Stata Journal, 2(4), 358–377.

Caliendo

, & Kopeinig

(2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys, 22(1), 31–72.

Clark

R. L.

, Morrill

M. S.

, & Allen

S. G.

(2012). The role of financial literacy in determining retirement plans. Economic Inquiry, 50(4), 851–866.

Cole

, Sampson

, & Zia

(2011). Prices or knowledge? What drives demand for financial services in emerging markets? The Journal of Finance, 66(6), 1933–1967.

Gertler

P. J.

, Martinez

, Premand

, Rawlings

L. B.

, & Vermeersch

C. M.

(2016). Impact evaluation in practice. World Bank Publications.

10.

Ghilarducci

, Radpour

, & Webb

(2022). Retirement plan wealth inequality: Measurement and trends. Journal of Pension Economics & Finance, 21(1), 119–139.

11.

Ginneken

W. V.

(1996). Social security for the informal sector: Issues, options and tasks ahead. Technical report, International Labour Organization.

12.

Holden

, Bass

, & Copeland

(2024). 401(k) plan asset allocation, account balances, and loan activity in 2022. ICI Research Perspective, 30(3), 1–24.

13.

Jha

, & Basole

(2023). Labour incomes in India: A comparison of two national household surveys. The Indian Journal of Labour Economics, 66(1), 181–201.

14.

Kumar

, Shukla

G. P.

, & Sharma

(2019). Analysis of key barriers in retirement planning: An approach based on interpretive structural modeling. Journal of Modelling in Management, 14(4), 972–986.

15.

Lusardi

, & Mitchell

O. S.

(2014). The economic importance of financial literacy: Theory and evidence. Journal of Economic Literature, 52(1), 5–44.

16.

Lusardi

, & Mitchelli

O. S.

(2007). Financial literacy and retirement preparedness: Evidence and implications for financial education: The problems are serious, and remedies are not simple. Business Economics, 42, 35–44.

17.

Mahapatra

M. S.

, Raveendran

, & De

(2019). Building a model on influence of behavioural and cognitive factors on personal financial planning: A study among Indian households. Global Business Review, 20(4), 996–1009.

18.

Mohan

, Jothikaran

T. A. J.

, Gudi

, & Ashok

(2024). Growing old in india: Mapping health and social welfare policy responses on aging. In Handbook of aging, health and public policy: Perspectives from Asia (pp. 1–16). Springer.

19.

Murari

, Shukla

, & Adhikari

(2021). Do psychological social and financial perceptions of post-retirement life and demography influence the retirement planning behaviour? International Journal of Social Economics, 48(11), 1545–1566.

20.

Murthy

S. R.

(2019). Measuring informal economy in India: Indian experience. In Seventh IMF statistical forum, Washington, DC.

21.

Nimrod

(2017). Older audiences in the digital media environment. Information, Communication & Society, 20(2), 233–249.

22.

Pension Fund Regulatory and Development, Authority (PFRDA). (2021). Annual report 2020–21. Government of India.

23.

Reserve Bank of India. (2021). Annual report 2020–21. Government of India.

24.

Rooj

, & Sengupta

(2025). Contributory pension scheme and formal retirement savings: Is there a trade-off? Evidence from India’s Atal Pension Yojana using copula regression methodology. Review of Economics of the Household, 23(1), 225–244.

25.

Rosenbaum

P. R.

, & Rubin

D. B.

(1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41–55.

26.

Scherer

, & Naab

T. K.

(2009). Money does matter. In Media choice (pp. 84–97). Routledge.

27.

Thaler

R. H.

, & Sunstein

C. R.

(2021). Nudge: The final edition. Yale University Press.

28.

Thoemmes

F. J.

, & Kim

E. S.

(2011). A systematic review of propensity score methods in the social sciences. Multivariate Behavioral Research, 46(1), 90–118.

29.

Ugargol

A. P.

, Kashyap

G. C.

, & Shaw

(2023). Exploring the influence of work, retirement, and pensions on the self-reported health of older adults in India: Evidence from the LASI and beyond. In Rajan

S. I.

(Ed.), Handbook of aging, health and public policy: Perspectives from Asia (pp. 1–28). Springer.

30.

Yang

, & Ching

A. T.

(2014). Dynamics of consumer adoption of financial innovation: The case of ATM cards. Management Science, 60(4), 903–922.

Identifying Effective Communication Channels for Financial Inclusion: Evidence from Self-funded Retirement Plans

Abstract

Keywords

1. Introduction

2. Data Sources

3. Research Framework

3.1 Examining Self-funded Retirement Plans

3.2 Sources of Financial Information

3.3 Research Design

3.3.1 Covariate Balancing

4. Results

4.1 Findings

4.2 Robustness Test

5. Limitations of the Study

6. Conclusions and Policy Implications

Footnotes

Acknowledgements

Data Availability

Declaration of Conflicting Interests

Ethical Approval

Funding

Notes

References