Timing Matters: How Unconditional Cash Transfer Frequency Shapes Dietary Choices Among Vulnerable Older Adults

ABSTRACT

Low‐income elderly consumers, often overlooked in studies of public financial assistance programs, may face challenges in managing resources to maintain dietary health. This study examines how the payment frequency of an unconditional cash transfer program impacts dietary behaviors among low‐income elderly populations in Colombia. Using nationally representative data from a federal financial assistance program, we implement a Multivalued Treatment Model with Propensity Score Matching and Augmented Inverse Probability Weighting to assess the effects of monthly versus bi‐monthly payments on food expenditures, dietary diversity, and diet quality. Results show that monthly payments enhance dietary diversity and diet quality by supporting healthier food purchases compared to bi‐monthly payments. Subgroup analyses revealed that older seniors and males benefit most from monthly payments. Mechanisms explored include consumption smoothing and subjective wealth perceptions, with findings suggesting that more frequent and predictable income flows may foster greater perceived financial stability. The study advances consumer behavior theory by linking payment frequency to dietary outcomes and offers practical implications for optimizing financial assistance programs. While bi‐monthly payments reduce administrative costs, monthly payments are recommended to improve dietary health. These findings contribute to understanding income timing's role in fostering healthier consumer behavior.

1. Introduction

Low‐income older adults represent a key population in economics, given their growing impact on consumption patterns, healthcare demand, and public policy (Price et al. 2000; United Nations 2019). Despite the increasing policy relevance of this group, there is a gap in empirical evidence on how income timing shapes consumer behavior, particularly among low‐income elderly populations. This study examines how the payment frequency of an unconditional cash transfer program influences dietary behaviors in this group. As individuals age, their nutritional needs change, making adequate nutrition critical for cognitive function, physical health, frailty and preventing malnutrition‐related complications (Norman et al. 2021). Healthier elderly individuals maintain independence, reduce healthcare costs, and sustain consumer demand over longer lifespans (Bloom and Canning 2008; Cylus et al. 2019; Santos and Cylus 2024). However, influencing individuals over 50 is challenging due to habitual choices driven by inertia and limited opportunities to increase income (Alós‐Ferrer et al. 2016). Older adults often adhere to established spending patterns, making it difficult to change behavior.

Low‐income levels are a primary barrier preventing vulnerable consumers from accessing a healthy diet (FAO 2023), leading to the rise of unconditional cash transfer programs globally. While these programs aim to alleviate poverty and improve consumer welfare, governments often modify payment schedules. For instance, Brazil and Argentina provide monthly payments, whereas Mexico and Colombia use bi‐monthly (once every 2‐months) schedules (ECLAC 2024). Low frequent payments like bi‐monthly or quarterly (once every 3‐months) are often favored by governments due to lower administrative costs and logistical simplicity. These changes in frequency can unintentionally affect consumer behavior, especially when resources are tight. De La Rosa et al. demonstrate that aligning financial benefits with familiar budgeting cycles, like monthly periods, enhances resource planning and intended use (2022). Therefore, for elderly consumers, monthly payments may support better budgeting, helping them prioritize nutritious food and maintain dietary quality.

While research on income stability and payment structures has extensively examined financial decision‐making (Thaler 1999), much of the research focuses on the general population, maternal or child populations, examining the size and conditions of cash transfer (Aizer et al. 2016; Parker and Todd 2017; Tiwari et al. 2016). The role of payment frequency in shaping food choices among low‐income elderly populations remains largely unexplored. Elderly consumers, in particular, may struggle to budget effectively over longer payment cycles due to cognitive aging and increased financial uncertainty. Existing evidence about payment frequency comes primarily from laboratory studies (De La Rosa and Tully 2022), offering limited insights into real‐world dietary behavior and nutrition outcomes. A key theoretical model used to explain the effect of payment frequency lies in its influence on subjective wealth perceptions. Higher payment frequencies reduce uncertainty about resource availability, allowing individuals to feel more financially secure throughout a period (De La Rosa and Tully 2022). Previous research on the delivery frequency of financial assistance programs has employed diverse designs and measured outcomes, providing insights into specific aspects of consumer behavior. However, these studies often lack a holistic approach to understanding overall nutritional behavior, instead focusing narrowly on specific food purchases or consumption patterns. For example, Aguila et al. found that monthly payments, as opposed to bi‐monthly payments, reduced food insecurity (2017). However, that study relied on hunger experience indicators rather than distinguishing between healthy and unhealthy choices or assessing dietary quality and diversity (Aguila et al. 2017). Additionally, Aguila et al. compared two programs with different payment rates. Basu et al. showed no significant improvements in nutritional indicators when consumers received weekly versus monthly food vouchers. This was likely due to restrictions on spending, which applied not only to the types of items that could be purchased but also to where the vouchers could be used (2019). Similarly, Banerjee, et al. found that consumers generally preferred monthly vouchers over weekly redemption, highlighting the influence of payment structure on consumer preferences (2024).

This study advances the understanding of how payment frequency in unconditional cash transfer (UCT) programs influences financial and dietary decisions among low‐income populations. While this research is grounded in the context of Colombia, its insights into subjective wealth perceptions and income cycles may extend to similar populations globally, where UCT programs may operate under comparable socio‐economic and institutional conditions. A key contribution is its empirical test of subjective wealth perceptions as a mechanism linking payment frequency to financial stability and consumer behavior. By extending theoretical insights into income cycles, particularly for low‐income elderly populations, it addresses a critical gap in health economics and consumer research: the impact of payment frequency in unconditional cash transfer programs on dietary behavior and nutrition among low‐income populations.

To our knowledge, this research is the first to empirically analyze payment frequency on nutrition using a nationally representative dataset from a government‐provided unconditional cash transfer program for the low‐income elderly. This study employs a Multivalued Treatment Model and Augmented Inverse Probability Weighting, which is a novel approach to performance propensity score matching, using doubly robust estimates to ensure unbiased results when either the propensity score or outcome model is correctly specified. This method allows for estimation of the effects across multiple levels of intervention, such as varying payment frequencies, while addressing selection bias. Despite its strengths, the multivalued treatment model has seen limited application in empirical research, making this study one of the first to apply it to consumer behavior and nutrition outcomes. The findings show that compared to bi‐monthly payments, monthly payments promote healthier dietary diversity and higher diet quality. To test the robustness of these results and explore underlying mechanisms, we conducted a subgroup analysis. The analysis reveals larger effects for individuals residing closer to food retailers, whereas bi‐monthly payments are associated with stronger effects among recipients facing longer travel distances to food outlets. These results suggest that monthly payments improve dietary outcomes more effectively overall, likely by enhancing subjective wealth perceptions. However, the subgroup differences highlight the potential value of tailoring payment policies geographically, such as implementing bi‐monthly payments or allowing benefits to roll over for up to 2 months in recipients facing longer travel distances to food outlets. Such an approach could optimize program effectiveness by addressing the unique challenges faced in these regions.

1.1. Dynamics of Nutrition Among Vulnerable Consumers

Elderly consumers' dietary decisions are shaped by income stability, resource allocation, preferences, and access to nutritious foods (Lo et al. 2012). While unconditional cash transfer programs are essential, they often overlook how payment frequency impacts budgeting and food purchasing behaviors. Often, elderly consumers struggle with budgeting correctly over longer payment cycles due to cognitive limitations such as memory loss, or present bias. Additionally, sex‐based differences further influence these outcomes (Alderman et al. 2003; Gonzalez et al. 2005). Payment frequency can also affect dietary outcomes in areas with limited access to nutritious food outlets.

Less frequent payments, such as bi‐monthly ones, may challenge consumers across all age groups to manage resources over extended periods. Memory‐related difficulties, prevalent among older adults, may exacerbate these challenges, leading to inconsistent dietary patterns and reliance on calorie‐dense, low‐nutrition foods as funds are depleted before the next payment cycle. Conversely, regular monthly disbursements may provide predictable cash flows, enabling smoother budgeting and more frequent purchases of fresh, nutritious foods. Additionally, present bias—the tendency to prioritize immediate rewards over future benefits—can increase with age, particularly among the oldest old, due to cognitive decline and a shorter perceived time horizon (Carstensen et al. 1999; Eppinger et al. 2018). This supports the hypothesis that monthly payments might improve diet quality more effectively than bi‐monthly payments, with the benefits increasing as seniors age due to reduced long‐term budgeting challenges.

Food access interacts with payment dynamics, particularly in areas with limited geographic access to food outlets, where bi‐monthly payments may incentivize bulk purchases to reduce transportation costs.

While convenient, bulk buying can compromise freshness and variety, affecting dietary quality. In contrast, monthly payments in areas with better food access allow for more frequent purchases of perishable, nutrient‐dense foods, supporting dietary quality. For consumers with limited access to food outlets, bi‐monthly payments may offer logistical convenience but at the expense of food diversity. This supports the hypothesis that monthly payments could be more effective in areas without limited access to food, while bi‐monthly payments may better align with logistical constraints in areas with limited access to food.

1.2. Optimizing Payment Frequencies in Unconditional Cash Transfer Program

Governments often adjust payment frequencies in cash transfer programs to optimize administrative efficiency and reduce transaction costs (Gentilini 2016). While consolidated disbursements achieve these goals, evidence suggests that payment frequency should be aligned with beneficiaries' budgeting cycles to ensure higher effectiveness (De La Rosa et al. 2022). Monthly payments are more aligned with traditional income schedules (De La Rosa et al. 2022), making it easier to plan budgets and manage resources effectively.

Research by De La Rosa et al. demonstrates that higher payment frequencies reduce uncertainty about resource availability, improving subjective wealth perceptions and encouraging more consistent spending patterns (De La Rosa and Tully 2022). Conversely, less frequent payment cycles, such as bi‐monthly disbursements, may disrupt established financial habits, particularly among older adults and exacerbating cognitive challenges, possibly leading to reliance on lower‐quality, calorie‐dense foods as funds are depleted before the next cycle, as have found in other studies (Harris‐Lagoudakis and Wich 2021; Hastings and Shapiro 2018).

Beyond cost and efficiency considerations, altering payment frequency may address critical logistical challenges faced by consumers, particularly those in remote or underserved areas (Aker et al. 2016). Fewer disbursements reduce the need for frequent travel to access funds, alleviating transportation costs and time constraints for vulnerable consumers (Aker et al. 2016). While bi‐monthly payments may enable bulk purchases in areas with limited geographic access to food outlets, consumers receiving monthly payments can also save funds for larger, infrequent shopping trips if needed. However, it is important to note that most cash transfer programs operate under a “use it or lose it” system, where recipients must claim or use their benefits within the payment cycle. Unused benefits are typically not saved—either partially or fully—for the next cycle which limits recipients' flexibility in managing their finances. These adjustments may still be impactful in contexts where physical or digital financial infrastructure remains limited, but the specific benefits of bi‐monthly payments depend on individual budgeting practices and access constraints.

1.3. Theoretical Background

Understanding how payment frequency influences consumption behavior requires integrating insights from consumption smoothing theory, liquidity constraints, and behavioral responses to income timing. While standard economic models predict that predictable income streams should not affect consumption choices, extensive empirical evidence documents systematic deviations from full consumption smoothing among low‐income households.

Under the life‐cycle hypothesis framework, individuals are assumed to smooth consumption over time, implying that predictable changes in income timing—such as monthly versus bi‐monthly payments—should not alter spending patterns (Modigliani 1985). However, low‐income households often fail to smooth consumption between income receipts, with expenditures on food and other necessities rising immediately after benefit receipt and declining thereafter, particularly among households with limited savings and liquidity (Mastrobuoni and Weinberg 2009; Shapiro 2005; Stephens 2003).

A growing literature therefore highlights payment frequency as a policy‐relevant design feature. Aguila et al. (2017), studying social programs in Mexico, show that more frequent payments are associated with improved food availability and healthcare use, whereas bi‐monthly payments facilitate larger purchases of durable goods. These findings indicate that payment frequency affects not only total spending but also the composition of expenditures, even when total transfer amounts are similar.

The above patterns can be understood through the lens of subjective wealth perceptions. Payment frequency may influence how individuals perceive their available resources, even when objective income remains constant. Frequent disbursements, such as monthly payments, reduce uncertainty about near‐term resource sufficiency and foster a sense of financial stability, increasing perceived subjective wealth (De La Rosa and Tully 2022). This perceived stability may encourage more consistent and intentional spending on essentials, particularly non‐storable goods such as nutritious food. Supporting this mechanism, recent evidence shows that higher subjective wealth perceptions are positively associated with improved dietary quality, independent of objective income levels (Rahal et al. 2023).

At the same time, subjective wealth perceptions might not operate unidirectionally. Large, infrequent payments may generate temporary increases in perceived financial slack, enabling households to accumulate cash balances and undertake purchases that are infeasible under more frequent payment schedules. Consistent with this mechanism, tax‐refund evidence shows that households—particularly those with low liquid assets—disproportionately time durable‐goods purchases around lump‐sum income events (JPMorgan Chase Institute 2020). Payment frequency may also interact with food perishability, as more frequent payments facilitate regular purchases of perishable foods (not durable goods)—which tend to be nutritionally healthier—while less frequent payments may encourage reliance on non‐perishable foods that can be stored over longer intervals and tend to be less nutritionally healthier.

These mechanisms imply that payment frequency might affect not only consumption levels but also its timing and composition. More frequent payments may promote nutritional investments by supporting access to healthy food, while less frequent payments may favor durable or semi‐durable purchases (e.g., ultra processed foods) when budgets are constrained. Guided by prior evidence, we hypothesize that monthly payments are associated with higher diet quality. Nevertheless, the relative effectiveness of payment frequency may depend also on contextual factors. For instance, in settings where beneficiaries face high fixed costs of shopping—such as long travel distances to grocery stores or limited market access—less frequent payments may be optimal by enabling fewer shopping trips with larger purchase volumes. In such contexts, bi‐monthly payments may reduce transaction costs and support bulk purchasing.

2. Methodology

This study examines the impact of payment frequency on key dimensions of spending and dietary behavior. Specifically, we analyze effects on food expenditures, food variety—including purchases of proteins, vegetables, and fruits—and consumption of ultra‐processed foods (UPFs) and sugar‐sweetened beverages (SSBs). We then assess overall diet quality using an adapted Diet Quality Index–International (DQI‐I). The analysis further examines heterogeneity by geographic access to food outlets and evaluates potential mechanisms underlying payment‐frequency effects on diet quality. Robustness checks are conducted to assess the stability of the findings and are reported in Appendix 2 due to space constraints.

Cash Transfer Program Context: Unconditional cash transfer programs specified for the elderly population living in poverty provide financial support to individuals who do not qualify for contributory pensions due to insufficient work history or contributions (Newson and Bourne 2011). Over 80 countries have implemented these programs (Newson and Bourne 2011), including Colombia's “Colombia Mayor”.

Launched in 2003, the Colombia Mayor program provides unconditional monetary support to elderly individuals living in poverty to alleviate financial hardship and improve overall well‐being. One of the program's primary goals is to enhance nutrition, particularly for vulnerable populations with limited resources (Ministerio del Trabajo 2019). Unlike conditional assistance programs, Colombia Mayor provides unconditional monetary support to those who do not qualify for contributory pensions due to insufficient work histories or contributions (Departamento Nacional de Planeación 2003).

Eligibility criteria for the program include being a Colombian citizen, residing in the country for at least 10 years, and having insufficient income to sustain oneself. Applicants must meet income thresholds: earning less than the minimum monthly salary, with individuals living alone limited to USD 142 (half the minimum monthly salary) or household income (adjusted for the number of household members) below USD 285. Income thresholds are assessed through the System of Identification of Social Program Beneficiaries (SISBEN IV, for its acronym in Spanish) used by the Colombia Government to determine eligibility for social welfare programs. It is a scoring system that classifies households into four main groups (A, B, C, and D) based on their income and living conditions to identify those most in need of social assistance (Castaneda 2010; Departamento Nacional de Planeacion 2020). Group A includes households in extreme poverty and is subdivided into five subgroups (A1 to A5). Group B corresponds to households in moderate poverty, consisting of seven subgroups (B1 to B7). Group C represents households in conditions of vulnerability and is divided into 18 subgroups (C1 to C18). Finally, Group D includes households that are neither poor nor vulnerable, classified into 21 subgroups (D1 to D21). Households in Groups A, B, and subgroup C1 are eligible for the program, as these represent the most socioeconomically disadvantaged populations (Income thresholds). Women qualify at age 54, and men at age 59, reflecting women's higher financial vulnerability in old age.

Initially, payments (in cash) were disbursed monthly, amounting up to $75,000 Colombian Pesos‐COP ($18.93 USD) nationally, with a higher level of $120,000 COP ($30.59 USD) in Bogotá (All USD values are adjusted to 2023). The Colombia Mayor transfer represents roughly 23% of the average monthly income among program eligible older adults and about 7% of the statutory minimum wage in 2023. Payments were collected in person at designated cash payment points, requiring beneficiaries to travel to a physical location.

The beneficiaries must claim their payments within the designated period; otherwise, the payment for that cycle is forfeited and does not carry over to the next payment cycle. However, since 2019, beneficiaries in 63 municipalities receive bi‐monthly payments due to geographic distribution challenges (Ministerio del Trabajo 2018). Because of this, studies should control at least for key geographic and administrative factors that predict the assignment of payment frequency—such as weak financial infrastructure, long travel distances, and low population density. These constraints can be proxied in the analysis using walking distance to financial service providers (to capture access to payment infrastructure), rural versus urban classification (as a proxy for remoteness and population sparsity), and municipality fixed effects (to absorb unobserved, time‐invariant characteristics such as administrative capacity, transportation barriers, and security conditions, that often do not change in a short period). Including these controls helps address potential selection bias stemming from the non‐random assignment of bi‐monthly payments across municipalities. Each transaction carries a fee of approximately $0.87 USD per beneficiary payment (Ministerio del Trabajo 2018). The program covered 1.7 million beneficiaries in 2020, with expenditures growing from $195 million to $430 million USD between 2011 and 2020, representing 0.2% of Colombia's GDP. By 2020, program coverage reached 21% of the elderly population aged 65 and older, up from 12% in 2011 (ECLAC 2024).

2.1. Data

This research utilizes data from the Colombian National Quality of Life Survey (ECV), a nationally representative cross‐sectional survey conducted annually by Colombia's National Administrative Department of Statistics. The ECV, part of the World Bank's Living Standards Measurement Study initiative, provides comprehensive demographic, socioeconomic, and consumer behavior data (Departame and nto Administrativo Nacional de Estadística 2021; Grosh et al. 1998). The dataset spans 2020 to 2022 and includes detailed information on food expenditures, as well as dietary behaviors, enabling a rigorous analysis of payment frequency impacts.

The ECV covers approximately 250,000 individuals annually, representing both rural and urban areas. For this study, the sample was restricted to individuals aged 54 and older for women and 59 and older for men, living below the poverty line as defined by the SISBEN classification, and who directly reported information on nutritional indicators. At our request, the Colombia's National Administrative Department of Statistics provided the SISBEN information using ECV responses, consistent with previous studies (Hessel et al. 2018). The SISBEN information provided was only available as categorical groups (A, B, C1, C2, etc.). Observations from Bogotá were excluded because beneficiaries in that place receive higher payment rates. That approach ensures a more homogeneous comparison among beneficiaries with similar payment rates, resulting in a final sample of 55,602 individuals over 3 years, including 22,368 program beneficiaries (2020 = 7,387, 2021 = 7,582, 2022 = 7399).

2.2. Outcomes

Our primary outcome is the Diet Quality Index, it is adapted from the Diet Quality Index‐International (DQI‐I) is a comprehensive measure designed for global monitoring and cross‐national comparisons of diet quality (Kim et al. 2003), and measures dietary quality and variety using a score ranging from 0 to 20, with higher scores indicating better nutrition. It includes two components: variety across five food groups (meat/poultry/fish/egg, dairy/beans, grains, fruits, and vegetables) and variety among six protein sources (meat, poultry, fish, dairy, beans, and eggs) (Appendix 1). The secondary outcomes includes weekly household: food expenditure, binary indicators for protein, vegetable, and fruit purchases, daily UPF consumption (packaged foods such as chips, fried plantains, cheese puffs, mixed snacks, doughnuts, or pork rinds at least once per day), and daily SSB consumption as binary (intake of sugary drinks, including sodas, soft drinks, processed fruit juice drinks, sweetened tea, and powdered refreshments, at least once per day) indicators as well. Self‐reported income satisfaction, used as a proxy for Subjective Wealth Perceptions, was measured using participant overall satisfaction with their current income on a scale from 0 to 10, where 0 indicates completely dissatisfied and 10 indicates completely satisfied.

2.3. Control

The analysis includes demographic controls (age, sex, race/ethnicity, geographic area (rural and urban), survey year, and municipality of residence), marital status, beneficiary education level, total household expenditure (weekly household food and non‐food expenditures), presence of children ≤ 17 years old, potential caregivers (women between 18 and 50 years old), and household size. To account for supply‐side factors, controls include walking distance in minutes to food stores, public markets, and financial service providers, as well as a binary variable if there was a receipt of COVID‐19‐related financial assistance by the government (Table 1).

TABLE 1.

Characteristics of the participants.

Variables	No program (n = 33,234)		Monthly payment (n = 20,724)		Bi‐monthly payment (n = 1644)
	%	n	%	n	%	n
Sex (males & females)
Male	51.29	17,045	51.72	10,718	53.59	881
Age—mean (SD)	65.20	(7.73)	73.00	(8.02)	72.25	(8.11)
Race
Indigenous	10.62	3528	6.81	1412	20.38	335
Black	10.06	3345	9.29	1925	20.38	335
Non‐indigenous/black (eg. white/caucasian)	79.32	26,361	83.90	17,387	59.25	974
Geographical area (urban & rural)
Urban	45.59	15,151	38.87	8055	33.64	553
Education level of the beneficiary
None	17.77	5900	29.89	6193	33.23	546
Primary	55.53	18,439	61.71	12,783	59.77	982
Secondary	23.63	7845	8.06	1670	6.51	107
University or more	3.07	1021	0.34	70	0.49	8
Marital status (with partner/married & no partner/married)
With partner/married	44.69	14,852	41.68	8638	43.13	709
Child in the house (< 17 years; yes & not)
Yes	25.80	8576	19.80	4103	20.74	341
Potential caregiver (women 18–50 years; yes & not)
Yes	25.62	8514	19.73	4088	16.24	267
Household size—mean (SD)	2.47	(1.58)	2.44	(1.57)	2.33	(1.49)
Distance to food stores (minutes walking)—mean (SD)	14.13	(22.91)	15.31	(23.95)	15.69	(27.52)
Distance to public markets (minutes walking)—mean (SD)	21.65	(30.64)	23.11	(32.88)	18.75	(33.22)
Distance to financial service providers (minutes walking)—mean (SD)	20.80	(29.87)	22.96	−32.44	19.99	(19.99)
Monetary aid during the COVID‐19 pandemic (yes & not)
Yes	15.97	5307	18.21	3773	16.67	274

Note: Table 1 reports the characteristics of the total population included in the sample. For the continuous variable, means are reported, and standard deviations are reported in parentheses. For categorical variables, percentages are reported. No program: people eligible by age and income but not in the program are the reference group.

2.4. Empirical Strategy

This study leverages data from the Colombian ECV from 2020 to 2022, a post‐policy change period when some beneficiaries transitioned to bi‐monthly payments. The descriptive analysis examines the outcome distributions across the three groups (non‐beneficiaries eligible for the program, program participants receiving monthly payments, and those receiving bi‐monthly payments) to identify differences in resource allocation and dietary behavior. Control variables such as demographic and socioeconomic characteristics, geographic context, and household composition are included to account for factors influencing payment frequency and dietary behaviors.

Beneficiaries receiving monthly or bi‐monthly payments may differ systematically from non‐beneficiaries or each other. Such differences—arising from factors such as socioeconomic status, household composition, or geographic disparities—can introduce selection bias. A simple unadjusted comparison of the groups would therefore yield biased estimates of the impact of payment frequency on expenditure patterns.

While binary Propensity Score Matching (PSM) is often used to adjust for selection bias, it assumes a binary treatment (treated vs. control). This assumption is not appropriate for the current study, where there are three distinct treatment levels: non‐beneficiaries, monthly beneficiaries, and bi‐monthly beneficiaries. Binary approaches cannot account for heterogeneity in treatment effects across multiple treatment groups. One alternative approach is to run separate PSM models for monthly beneficiaries versus non‐beneficiaries and bi‐monthly beneficiaries versus non‐beneficiaries. However, this method presents several challenges: it assumes comparability of non‐beneficiaries across both models, which may not hold due to differences in covariate distributions; it introduces bias from differences in covariate balance and matching quality between the models; and it fails to account for heterogeneity in treatment effects, as monthly and bi‐monthly beneficiaries are not directly compared within the same framework. These limitations make this approach less suitable for analyzing the impact of payment frequency.

To address this, and estimate the impact of payment frequency, we employ a Multivalued Treatment Model (MTM) using Multinomial Propensity Score Matching (MPSM) with inverse probability weighting (Brown et al. 2020; Cattaneo 2010; Mccaffrey et al. 2013). This method assigns a propensity score to each individual based on the likelihood of receiving one of the three treatment levels (non‐beneficiary, monthly beneficiary, or bi‐monthly beneficiary), given their observed characteristics, such as geographic location, household composition, and socioeconomic status. These propensity scores are then used to match individuals across the three groups, ensuring that each group is comparable in terms of these characteristics. By improving balance on observed covariates, this approach reduces bias arising from observable differences across groups; however, it cannot account for unobserved factors. Thus, the MTM should be interpreted as estimating effects conditional on observed characteristics. For example, in this study, an individual from a rural municipality receiving bi‐monthly payments would be matched to a similar individual in either the monthly payment group and the non‐beneficiary group, ensuring comparability based on shared baseline characteristics. This method is particularly suited for assessing programs with multiple treatment categories, as it allows for balanced comparisons across groups, enabling a more accurate assessment of the impact of the program on the outcomes. The MTM framework explicitly addresses the non‐random nature of treatment assignment, selection bias, and treatment heterogeneity. In this case, the categories included in the model are non‐beneficiaries, monthly beneficiaries, and bi‐monthly beneficiaries. Including non‐program participants establishes a baseline for comparison, avoids selection bias, and highlights the impact of unconditional cash transfer programs. Including non‐participants helps account for the influence of factors that may determine treatment assignment as well.

To accurately estimate the impact of payment frequency on the outcomes, the analysis incorporates a comprehensive set of control variables. These controls account for potential confounding factors that could influence nutritional patterns and that the observed effects are attributable to payment frequency rather than external or pre‐existing differences across groups. For example, demographic variables address differences in consumption preferences, resource access, and local contexts. Socioeconomic controls capture financial and literacy‐related influences on spending. Household composition factors reflect intra‐household dynamics impacting expenditure and nutrition. Supply‐side variables, such as distances to food stores, markets, financial service providers, and receipt of COVID‐19 aid, account for external accessibility and financial support.

The model also adjusts for the fact that beneficiaries in 63 municipalities receive bi‐monthly payments due to geographic distribution challenges by including controls for municipality of residence, rural versus urban classification, and supply‐side factors. These variables account for regional disparities in access to resources and logistical constraints that influence payment frequency.

2.5. Model Specification

The empirical model is designed to estimate the effect of payment frequency on food expenditures, food variety (protein, vegetable, and fruit purchases, as well as daily UPF and SSB consumption), diet quality score, while accounting for the multivalued treatment structure. The model incorporates Augmented Inverse Probability Weighting (AIPW) with robust standard errors to ensure unbiased estimates. This doubly robust method combines propensity score matching with outcome modeling, providing valid results as long as either the propensity score or the outcome model is correctly specified. The application of the MPSM and AIPW approaches is novel in consumer and health economic research, particularly in the context of unconditional cash transfer programs.

The outcome model is specified as an OLS regression, food expenditure (with log‐transformed), diet quality score, protein, vegetable, and fruit purchases, as well as daily UPF and SSB consumption, as dependent variables. We also re‐estimated the models using a standardized (z‐score) version of the diet quality score outcome; full results are reported in Appendix 4.

The model includes the treatment indicator (payment frequency) and control variables as predictors. The estimation approach follows Mccaffrey et al. (2013), who applied multivalued treatment models in evaluating three adolescent substance abuse programs (Mccaffrey et al. 2013). Similarly, this study examines the effect of payment frequency by using non‐beneficiaries as the reference group and comparing them to monthly and bi‐monthly beneficiaries. Below we present the general equation of the model:

$${\text{FoodOutcome}}_{it}=\beta 1·\text{Payment}{\text{Frequency}}_{it}+\beta 2·{\mathrm{X}}_{it}+{\alpha }_{\mathrm{m}}+{\text{Year}}_t+{ϵ}_{it}$$ (1)

Where FoodOutcome _it : denotes food‐related outcomes for individual i at time t. Specifically, outcomes include log‐transformed food expenditure, food variety (purchases of protein, vegetables, and fruits, as well as daily consumption of UPF and SSB) and diet quality score. Payment Frequency _it : Treatment indicator (T = 0: Non‐beneficiary, T = 1: Monthly, T = 2: Bi‐monthly). X _it : Vector of control variables (e.g., demographic, socioeconomic, and household characteristics). α _m: Municipality fixed effects. Year _t : Year‐level fixed effects. ϵ _it : Error term. Control variables are included in the propensity score model through Inverse Probability Weighting to balance treatment groups and in the outcome model to adjust for residual confounding. This combination ensures double robustness, providing unbiased estimates of the impact of payment frequency on food outcomes. The AIPW method enhances the reliability of the results by combining weighting and outcome regression. The quality of the propensity score matching was assessed to ensure the model's reliability and balance across treatment groups (Figure 1).

FIGURE 1.

Matching quality for the multivalued treatment model. The balance plot provides a visual representation of the propensity score distributions across treatment groups before and after weighting.

Although we analyze multiple outcomes, these outcomes capture distinct dimensions of dietary behavior and are estimated in separate regression models. Diet quality is designated as primary outcome, while additional measures are treated as secondary and exploratory. Accordingly, results are interpreted outcome by outcome rather than as a joint hypothesis‐testing exercise. As a robustness check, we verified that inference for the primary outcome (diet quality) was unchanged under a Bonferroni adjustment.

2.5.1. Exploring Mechanisms Underlying Payment Frequency Effects on Diet Quality

Our empirical analysis is guided by two testable implications derived from standard consumption models.

First, under the life‐cycle hypothesis with consumption smoothing, predictable income timing should not affect diet quality once total income is held constant. Accordingly, we test whether estimated payment‐frequency effects on diet quality attenuate or disappear over time by estimating year‐specific effects within the multivalued treatment framework. Persistent differences across payment schedules would be inconsistent with full consumption smoothing.

Second, if payment frequency affects perceived financial security, more frequent payments may influence food choices even when objective income is unchanged. We therefore test whether payment‐frequency effects on diet quality vary systematically with subjective wealth perceptions, measured using a self‐reported income satisfaction scale (0–10). These analyses are implemented within the same multivalued treatment framework used for the main estimates.

2.5.2. Heterogeneity in the Impact of Payment Frequency on Nutrition

We examine heterogeneity in the effects of payment frequency by geographic access to food outlets. Geographic access is measured using walking time to the nearest food store or public market, with households located 20 min or more away classified as having more limited access, consistent with prior distance‐based measures of food availability (Jiao et al. 2012; Rasmussen et al. 2019). This analysis assesses whether payment frequency interacts with spatial constraints that affect shopping costs and purchasing strategies. In settings with longer travel distances, less frequent payments may facilitate bulk purchasing and reduce transportation costs, whereas in settings with closer access, more frequent payments may better support regular purchases of perishable foods. Heterogeneity estimates are obtained within the same multivalued treatment framework used in the main analysis.

3. Results

Table 1 presents descriptive characteristics of eligible older adults by program participation and payment frequency. Program beneficiaries are older, less educated, and more likely to reside in rural areas than non‐participants. Monthly and bi‐monthly beneficiaries display similar demographic profiles. Average distances to food outlets, public markets, and financial service providers are comparable across groups.

3.1. Matching Quality for the Multivalued Treatment Model

The results confirm that the propensity score model achieves a high level of balance across groups, supporting the validity of subsequent analyses using AIPW and the outcome model. In the unweighted sample, propensity score distributions differ across monthly beneficiaries, bi‐monthly beneficiaries, and non‐participants, indicating meaningful differences in observed characteristics. After weighting, the distributions align closely across groups, suggesting that the propensity score model effectively balances observed covariates and improves comparability for the subsequent AIPW analysis.

3.2. Payment Frequency on Food Expenditures

Descriptive analysis indicates that non‐participants (n = 33,234) spent slightly more on food ($10.37 USD, SD = 13.23) than monthly beneficiaries (n = 20,724; $9.62 USD, SD = 11.30) and bi‐monthly beneficiaries (n = 1644; $8.75 USD, SD = 11.30). These differences suggest that non‐program participants, who may have different socioeconomic characteristics than program beneficiaries, allocate slightly more toward food expenditures.

However, the results from the MTM (n = 52,630) indicate that participation in the Colombia Mayor program is associated with higher food expenditures relative to non‐participants. Compared to eligible non‐beneficiaries, monthly payment recipients spent 0.584% points more on food items (SE = 0.043, p < 0.01), while bi‐monthly recipients spent 0.643% points more (SE = 0.119, p < 0.01). These increases reflect the positive income elasticity of demand for food among low‐income older adults, indicating that additional income from the program translates into higher food spending regardless of payment frequency.

When comparing spending differences between bi‐monthly and monthly beneficiaries relative to non‐program participants, a non‐statistically significant difference was observed (β = 5.8%‐points for bi‐monthly, SE = 12.1%, p > 0.10). Importantly, higher food expenditure does not necessarily reflect better or healthier food choices. In the next results, we further explore the influence of payment frequency on food variety purchasing, with a focus on nutritional patterns and the consumption of healthy versus unhealthy foods.

3.3. Payment Frequency and Food Variety

This subsection examines how payment frequency influences food purchasing patterns among older adults. We analyze weekly purchases of proteins, vegetables, and fruits, as well as daily consumption of UPFs and SSBs. Unlike the previous subsection, which focused on total food expenditures, this analysis disaggregates spending by food category to assess differences in dietary composition across non‐beneficiaries, monthly beneficiaries, and bi‐monthly beneficiaries.

The unadjusted descriptive analysis reveals differences in food purchasing behaviors across the three groups. For protein purchases, the proportion of weekly purchase was highest among monthly beneficiaries (mean = 0.62, SD = 0.48) compared to non‐program participants (mean = 0.60, SD = 0.49) and bi‐monthly beneficiaries (mean = 0.58, SD = 0.49). Similarly, vegetable purchases were slightly higher for monthly beneficiaries (mean = 0.12, SD = 0.33) than non‐program participants and bi‐monthly beneficiaries, both with a mean of 0.10 (SD = 0.30 for both groups). In contrast, fruit purchases showed a decline among bi‐monthly beneficiaries (mean = 0.07, SD = 0.25) compared to monthly beneficiaries (mean = 0.13, SD = 0.34) and non‐program participants (mean = 0.12, SD = 0.33). For less nutritious food categories, daily UPF consumption remained low across all groups (mean = 0.02, SD = 0.12 for both non‐program participants, mean = 0.01, SD = 0.10 for monthly beneficiaries, and mean = 0.02, SD = 0.12 for bi‐monthly beneficiaries). Similarly, daily SSB consumption was slightly lower among monthly and bi‐monthly beneficiaries (mean = 0.05, SD = 0.22, and SD = 0.23, respectively) compared to non‐program participants (mean = 0.07, SD = 0.26).

The MTM results provide insights into how payment frequency influences food variety among elderly consumers, with effect sizes contextualized against the mean among program beneficiaries (Table 2).

TABLE 2.

Effect of the payment frequency on food variety among elderly: Multivalued treatment model.

	(1)	(2)	(3)	(4)	(5)
Payment frequency	Protein purchase (weekly)	Vegetables purchase (weekly)	Fruits purchase (weekly)	Ultra processed foods (daily)	Sugar sweetened beverages (daily)
Non‐program beneficiary	(Ref)	(Ref)	(Ref)	(Ref)	(Ref)
Monthly payment	0.050***	0.026***	0.007**	−0.004***	−0.009***
	(0.005)	(0.003)	(0.004)	(0.001)	(0.003)
Bi‐monthly payment	0.064***	0.004	−0.035***	0.005	0.001
	(0.014)	(0.010)	(0.011)	(0.005)	(0.009)
Bi‐monthly versus Monthly payment	0.014	−0.022**	−0.042***	0.010**	0.011
	(0.014)	(0.010)	(0.011)	(0.005)	(0.009)
Overall mean—program beneficiary	0.621	0.122	0.125	0.010	0.053
Observations	52,630	51,037	51,037	52,630	52,630

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Augmented Inverse Probability Weighting. The full set of control variables presented in descriptive Table 1 were used for matching and adjusting the Multivalued Treatment Models. The outcome models are specified as Linear Probability Models. Separate models were run for each outcome variable.

For protein purchases, monthly payments increased the likelihood of purchasing proteins by 5.0%‐points (p < 0.01) relative to non‐program participants, while bi‐monthly payments showed a slightly larger increase of 6.4%‐points (< 0.01). The difference in protein purchases between bi‐monthly and monthly beneficiaries is not statistically significant (Table 2).

For vegetable purchases, monthly payments led to a 2.6%‐point increase (p < 0.01) compared to non‐program participants, while bi‐monthly payments showed no significant effect. When comparing bi‐monthly to monthly payments, relative to non‐program participants, bi‐monthly payments were associated with a 2.2%‐point decrease (p < 0.05). Relative to the mean proportion (program beneficiaries) of vegetable purchases (12.2%), this decrease represents a significant decline of 18%.

For fruit purchases, monthly payments increased the likelihood of fruit purchases by 0.7%‐points (p < 0.05) compared to non‐program participants, while bi‐monthly payments significantly decreased the likelihood by 3.5%‐points (p < 0.01). The direct comparison showed that bi‐monthly payments reduced fruit purchases by 4.2%‐points (p < 0.01) compared to monthly payments. Against the mean fruit purchase proportion (12.5%), this represents a substantial decrease of 33.6%.

For UPFs, monthly payments reduced daily UPF consumption by 0.4%‐points (p < 0.01) compared to non‐program participants, while bi‐monthly payments had no significant effect. Comparing bi‐monthly to monthly payments, bi‐monthly payments were a 1%‐point increase in UPF consumption (p < 0.05). For SSBs, monthly payments reduced daily consumption by 0.9%‐points (p < 0.01) compared to non‐program participants, while bi‐monthly payments showed no significant effect. The comparison between bi‐monthly and monthly payments revealed no significant difference.

Overall, the results indicate that payment frequency affects the composition of food purchases. Monthly payments are associated with higher purchases of vegetables and fruits and lower consumption of UPF and SSBs, whereas bi‐monthly payments show weaker or adverse effects for these food categories. Differences across food groups suggest reallocation rather than uniform improvements in consumption, motivating an assessment of overall dietary balance using the Diet Quality Index in the next subsection.

3.4. Payment Frequency and Diet Quality

This subsection estimates the impact of payment frequency on diet quality among older adults using the Diet Quality Index, a multidimensional measure designed to assess nutritional quality by incorporating multiple food groups and is tailored for global monitoring and cross‐national comparisons of diet quality of dietary balance across food groups (Kim et al. 2003), allowing assessment of whether diet quality differs by payment frequency (see Appendix 1).

In the unadjusted analysis, the diet quality score had a mean of 6.01 (SD = 5.32) for non‐program participants. Monthly beneficiaries reported a slightly higher mean score of 6.14 (SD = 5.31), while bi‐monthly beneficiaries exhibited the lowest mean score of 5.10 (SD = 4.70).

Estimates from the MTM (n = 51,024) indicate that payment frequency affects the overall diet quality. Relative to eligible non‐participants, monthly beneficiaries experience a 0.458‐point increase in the Diet Quality Index (SE = 0.049, p < 0.01), while the effect for bi‐monthly beneficiaries is not statistically different from zero (β = −0.122, SE = 0.146, p > 0.10). Comparing payment schedules directly, bi‐monthly payments cause a 0.580‐point lower diet quality score relative to monthly payments (SE = 0.148, p < 0.01). Relative to the mean diet quality score among beneficiaries (6.068), this corresponds to a 9.6% reduction.

While these aggregate estimates document a clear effect of payment frequency on overall diet quality, they may mask meaningful contextual variation. We therefore examine heterogeneity by geographic access to food outlets.

3.5. Heterogeneity in the Impact of Payment Frequency on Nutrition

Geographic access to food Geographic access to food outlets moderates the effect of payment frequency on diet quality. Among beneficiaries residing within 20 min walking distance to the nearest food store or public market, bi‐monthly payments are associated with a significantly lower diet quality score relative to monthly payments (β = −0.662, SE = 0.191, p < 0.01; n = 42,643). In contrast, among beneficiaries residing 20 min or more walking distance from the nearest food outlet, bi‐monthly payments are associated with a higher diet quality score relative to monthly payments (β = 0.754, SE = 0.352, p < 0.05; n = 8381). In Appendix 3, we present the results for food expenditure and food variety outcomes.

3.6. Potential Mechanisms Underlying Payment Frequency Effects on Diet Quality

Year‐specific estimates provide limited support for full consumption smoothing. Relative to monthly payments, bi‐monthly payments are associated with a 0.597‐point lower diet quality score in 2020 (SE = 0.256, p < 0.05), no statistically significant difference in 2021 (β = −0.050, SE = 0.167, p > 0.10), and a larger 0.823‐point reduction in 2022 (SE = 0.297, p < 0.01). Rather than converging over time, differences in diet quality by payment frequency persist and re‐emerge in later periods.

Consistent with the subjective wealth mechanism, bi‐monthly beneficiaries report lower income satisfaction than monthly beneficiaries, with a difference of 0.246‐points on a 0–10 scale (SE = 0.109, p < 0.05). This pattern indicates that payment frequency is associated with differences in perceived financial security, which may contribute to observed differences in diet quality across payment schedules.

4. Discussion

This research demonstrates that payment frequency significantly impacts consumer behavior among low‐income elderly populations in Colombia. Monthly payments improved dietary diversity and overall diet quality, encouraging purchases of nutrient‐dense foods while reducing consumption of UPF and SSBs. In contrast, bi‐monthly payments often led to less nutritious purchasing patterns, reflecting challenges in resource management. The effects varied across contexts, individuals residing closer to food retailers benefiting more from monthly payments.

For food expenditures, although bi‐monthly payments resulted in a larger increase compared to monthly payments, this difference was not statistically significant. Moreover, findings from food variety models suggest that the additional food expenditure associated with bi‐monthly payments may be disproportionately allocated to less nutritious options, such as UPFs.

Regarding dietary diversity, monthly payments facilitated more purchases of nutrient‐dense foods, vegetables, and fruits, while reducing consumption of UPFs and SSBs. These results support theories suggesting that frequent payments may alleviate psychological scarcity and enable better access to perishable, healthier food options (Mullainathan and Shafir 2013). In contrast, bi‐monthly payments reduced the variety of healthy food purchases and increased less nutritious options, potentially reflecting the challenges of managing resources over longer intervals. This finding aligns with previous research suggesting that less frequent payments increase expenditure on temptation goods, as individuals may struggle to resist impulsive purchases under extended budgeting periods (White and Basu 2016).

A broader evaluation of diet quality further reinforced the advantages of monthly payments. Beneficiaries receiving monthly disbursements demonstrated significantly higher Diet Quality Index scores, reflecting healthier and more diverse nutritional patterns. This finding aligns with prior studies emphasizing the role of regular income in promoting healthier behaviors (Hastings and Shapiro 2018; Shapiro 2005). Conversely, the reduced diet quality observed among bi‐monthly beneficiaries highlights the challenges of extended budgeting cycles, which may amplify financial strain and hinder the prioritization of nutrient‐rich foods.

The magnitude of our estimates is consistent with prior studies. In our sample of low‐income older adults (mean ≈ 6‐points), diet quality levels are lower than those typically observed in general populations from middle‐ and high‐income settings (9.7–13‐points) (Kim et al. 2003; Tur et al. 2005), as expected given socioeconomic differences. On the original scale, bi‐monthly beneficiaries exhibited a 9.6% lower diet quality score relative to monthly beneficiaries. This effect size is comparable to changes observed in other cash transfer‐related nutrition outcomes; for example, systematic evidence shows that participation in cash transfer programs is associated with an approximately 11.5% increase in the probability of consuming diverse food groups and a 6.7% increase in animal‐source food consumption, both of which are commonly used indicators of improved diet quality (Manley et al. 2022). To strengthen comparability, we also examined standardized (z‐score) diet quality outcome. The standardized impacts (bi‐monthly: −0.109 SD) mirror the original‐scale results in sign, significance, and payment‐frequency pattern (Appendix 4), reinforcing that the observed differences reflect substantive variation in dietary diversity rather than scaling artifacts.

Geographic context also played a pivotal role, with monthly payments proving more effective in areas with food access, where beneficiaries might have better access to diverse and nutritious foods. Conversely, bi‐monthly payments appeared to mitigate transportation and logistical constraints in areas with limited access to food, offering modest advantages in these settings. These findings align with previous research on the unique challenges and adaptations required in areas with limited geographic access to food outlets (Walker et al. 2010). The mechanisms underlying these effects were explored through the lenses of two consumer behavior theories: the Life‐Cycle Hypothesis and Subjective Wealth Perceptions. The life‐cycle hypothesis serves as a conceptual benchmark; however, persistent differences in diet quality by payment frequency suggest incomplete consumption smoothing, although our cross‐sectional elderly sample may limit our ability to fully test this hypothesis. In contrast, subjective wealth perceptions provide empirical support, with more frequent payments associated with healthier food purchasing. Although these patterns may also reflect differences in food perishability, as more frequent payments can facilitate regular purchases of perishable, nutritionally healthier foods, while less frequent payments may encourage reliance on non‐perishable foods (e.g., UPF) that can be stored over longer intervals.

4.1. Practical and Policy Implications

The findings have significant implications for consumer research and policy. Monthly payments emerge as a more effective strategy for improving dietary outcomes, emphasizing the importance of income timing in unconditional cash transfer program design. Policymakers should consider the benefits of frequent payments to support vulnerable populations in achieving healthier dietary behaviors. Tailoring interventions to the specific needs of subgroups, such as older adults in areas with limited access to food outlets, could significantly enhance program effectiveness. For example, in areas with limited access to food outlets, where bi‐monthly payments have shown greater nutritional impact, policies allowing the full or partial rollover of unused benefits could provide beneficiaries with more flexibility to save for future purchases, reducing the constraints of a “use it or lose it” system.

The findings underscore the importance of aligning payment schedules with household budgeting cycles (De La Rosa et al. 2022). Although governments often favor less frequent payments to reduce transaction costs, monthly disbursements—consistent with common budgeting periods—appear more effective. Payment frequency interacts with transfer size: more frequent payments relax short‐term liquidity constraints, but only if amounts remain sufficient to meet immediate needs without constraining lumpy expenditures. Our results indicate that monthly payments are more cost‐effective in improving diet quality, with a cost of $3.53 per diet quality point compared to $5.86 under bi‐monthly payments, yielding an ICER of $1.50. In standardized terms, monthly payments increase diet quality by 0.109 standard deviations, corresponding to approximately $0.80 per 0.1 SD. While bi‐monthly payments reduce administrative costs, the larger nutritional gains from monthly disbursements justify the additional expense, highlighting the need to jointly consider payment frequency and transfer size in program design. The lower transactional fees associated with bi‐monthly payments may seem advantageous initially; however, the greater improvements in diet quality achieved with monthly disbursements justify the additional cost. Policymakers should carefully weigh these trade‐offs when designing cash transfer programs, as monthly payments yield more efficient outcomes per dollar spent.

This research provides robust evidence but is not without limitations. First, the reliance on cross‐sectional data may challenge causal inferences regarding the mechanisms driving observed effects due to time‐variant factors. However, the study employed advanced econometric techniques to enhance reliability and validity. The use of Multinomial Propensity Score Matching ensured balanced groups across multiple categories, addressing potential selection bias. Additionally, the application of Augmented Inverse Probability Weighting with robust standard errors provided doubly robust estimates, ensuring unbiased results as long as either the propensity score or outcome model was correctly specified. However, because these approaches rely on selection on observables, they may not fully eliminate the possibility of bias from unobserved factors or from potential policy endogeneity in the allocation of monthly versus bi‐monthly payments.

Second, self‐reported dietary data may be subject to recall bias; however, this concern is mitigated by focusing on weekly food expenditures and purchasing proportions, which are less prone to recall error than longer reference periods. We further restrict the sample to respondents reporting nutritional information, though food variety may not perfectly reflect individual consumption. Future research should incorporate direct intake measures. Third, while unobserved factors such as dietary preferences or psychological traits may influence outcomes, extensive controls and balance diagnostics indicate strong overlap in observed characteristics across treatment groups. Robustness checks, including falsification tests and a complementary DiD analysis of SSB consumption, support the main findings. The survey does not measure cognitive or mobility limitations; although we partially address this using distance to financial service providers, future work should incorporate direct measures of functional status and exploit longitudinal data to assess longer‐run effects across programs and settings.

Another limitation is the lack of supermarket food purchase‐data, which prevents a assessment of nutritional quality. Future research could address this using purchase datasets; however, these often lack socio‐economic and demographic information present in our dataset, which is crucial for understanding broader patterns among vulnerable seniors. Concerns that cash transfer programs could lead to price increases by shopkeepers, reducing their benefits, are unlikely to affect our estimates, as monthly and bi‐monthly payments should react similarly. Lastly, although we included a variable for COVID‐related financial assistance, we did not explicitly model broader pandemic‐induced disruptions. Since the dataset spans 2020–2022, COVID‐19 may have affected food prices, availability, or consumer behaviors. However, these shocks likely impacted both monthly and bi‐monthly groups similarly. Moreover, our empirical strategy includes municipality and year fixed effects, along with subgroup analyses, which help mitigate concerns about confounding from time‐ and location‐specific shocks. Future research could more explicitly model pandemic‐related effects using longitudinal or high‐frequency panel data to better capture longer‐term behavioral adjustments.

In conclusion, this research advances the understanding of how payment frequency shapes spending and dietary behaviors among vulnerable consumers. By integrating insights from theoretical frameworks and addressing heterogeneity across geographic subgroups, the findings contribute to a deeper understanding of consumer decision‐making. These insights offer practical evidence for designing unconditional cash transfer programs that enhance well‐being and foster healthier nutritional patterns.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix 1. Diet Quality Index

 

Diet quality component	Grouping of diet quality component	Scoring criteria	Score
Variety—food groups	5 food groups: Meat/poultry/fish/egg, dairy/beans, grains, fruits, and vegetables	Each food group awarded 0 or 3 pts. 3 points awarded if at least 1 item from that group was got	0–15
Variety—protein sources	6 sources: Meat, poultry, fish, dairy, beans, eggs	3 or more sources got: 5 pts 2 sources got: 3 pts 1 source got: 1 pts 0 sources got: 0 pts	0–5

Note: Table adapted from Kim et al. 2003). This indicator will use the information for purchasing group food at least once per week by the household.

Appendix 2. Robustness Checks

To ensure the validity and reliability of the findings presented in earlier sections, this section conducts a series of robustness checks designed to address potential biases and strengthen the evidence base. These checks are crucial for advancing consumer research and refining the design of unconditional cash transfer programs. Four key approaches are employed.

Outcome Falsification Test

The outcome falsification test focuses on educational expenditure, an unlikely priority for older adults due to their immediate needs such as nutrition, healthcare, and daily living expenses. For individuals aged 50 and older, the opportunity cost of pursuing education increases, diminishing the likelihood of long‐term educational investments. This test hypothesizes that payment frequency should have no significant impact on educational spending within this population. The absence of any relationship would further support the robustness of the findings.

Population Falsification Test

This test examines whether the observed effects of payment frequency extend to a population that is income‐ineligible (In SISBEN‐Group D; more than 2,000,000 COP or USD 453.13) for unconditional cash transfer programs but meets the age criteria. For this group, which is not directly affected by program payments, no significant differences are expected in nutritional indicators (e.g., diet quality, protein and vegetable consumption) or food expenditure between municipalities implementing monthly and bi‐monthly payment structures. The absence of effects in this falsified population provides confidence that the identified differences in prior studies are specific to program beneficiaries.

Urban‐Rural Robustness Check

This test evaluates whether the effects of payment frequency on diet quality are driven by geographic context rather than the payment frequency itself, as bi‐monthly payments were originally adopted to address logistical challenges in cash distribution. If location drives the results, the hypothesis suggests that monthly payments will perform better in urban areas due to better resource access, while bi‐monthly payments will be more effective in rural areas. Conversely, if location has no significant influence, monthly payments are expected to outperform bi‐monthly payments in both rural and urban areas.

Using the ECV dataset from 2020 to 2022, in this section, we implement a series of falsification checks—including outcome falsification, population falsification, and urban‐rural heterogeneity tests—employing the same analytical procedures used in earlier sections. Each analysis compares treatment groups, with outcomes modeled using an OLS model.

Difference‐in‐Differences Robustness Check

To reinforce our main findings, we implement a difference‐in‐differences (DiD) robustness check leveraging an extended outcome variable—sugar‐sweetened beverage (SSB) consumption—available in ECV from 2018 to 2022. While dietary quality and other nutrition indicators are only observed post‐policy (2020–2022), SSB consumption allows for a pre‐post comparison across beneficiaries in municipalities exposed to monthly versus bi‐monthly transfers. The policy was legislated in December 2018 and implemented in early 2019.

This approach estimates the Average Treatment Effect on the Treated (ATT). While our primary analysis relies on a multivalued treatment model estimating the Average Treatment Effect (ATE), we do not treat DiD as a substitute, but as a complementary test of consistency in sign and significance under alternative identifying assumptions. Given that ATT and ATE reflect distinct estimands—ATT focusing on those exposed to the intervention and ATE generalizing to the full eligible population—differences in magnitude are expected and do not undermine robustness when effects are statistically and directionally consistent. We examine three outcomes derived from the SSB frequency question.

• Daily SSB consumption: Defined as 1 if the respondent consumes sugary drinks at least once per day (e.g., two or more times per day; once per day), and 0 otherwise. As in our main AIPW analysis, we expect minimal behavioral response among high‐frequency consumers due to stable preferences and behavioral inertia, especially in older populations.

• Any SSB consumption (Prevalence): Defined as 1 if the individual consumes SSBs at least once per week, and 0 otherwise. This includes both daily and occasional users. We hypothesize that most of the behavioral response, if present, will be concentrated among occasional or “light” consumers, whose weaker preferences may make them more responsive to liquidity changes induced by payment frequency.

• SSB frequency: To further test this mechanism, we re‐estimate the model using the SSB frequency outcome. When the analysis includes all frequency categories (1 = Not consumption, 2 = Less than once per week, 3 = Once a week, 4 = Two or three times per week, 5 = Four to six times per week, 6 = Every day of the week (once per day), and 7 = Every day of the week (two or more times per day))—including daily consumers—we expect a muted average effect due to the limited responsiveness of heavy consumers. However, when restricting the sample to exclude daily consumers, we anticipate detecting statistically significant changes in consumption frequency between monthly and bi‐monthly recipients.

This behavioral pattern aligns with established insights from consumer theory and behavioral economics: light users tend to be more elastic and responsive to nudges or liquidity constraints, whereas habitual consumers exhibit greater resistance to change. By distinguishing between these subgroups, the DiD framework may provide additional support for our central finding—that more frequent cash transfers lead to healthier consumption behaviors.

Results

Overall, these robustness checks provide confidence in the reliability of the primary findings. The results indicate that the effect of payment frequency on diet quality is attributable to the program's intended mechanisms and is not confounded by unrelated population or outcome factors. For outcome falsification, examining educational spending as an unrelated outcome revealed no significant differences between monthly and bi‐monthly beneficiaries (0.025, SE = 0.024, p > 0.10), consistent with expectations for older adults prioritizing immediate needs over long‐term educational investments (Table A1).

TABLE A1.

Robustness checks: Outcome falsification.

Monthly versus Bi‐Monthly payment	Outcome falsification: (Beneficiary education)
Bi‐monthly versus Monthly payment	0.025
	(0.024)
Observations	52,630

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Augmented Inverse Probability Weighting. The full set of control variables presented in descriptive Table 1 were used for matching and adjusting the Multivalued Treatment Models. The outcome models are specified as OLS Models.

Results from population falsification analyses show no significant differences between monthly and bi‐monthly payment groups for age‐eligible but income‐ineligible individuals in food expenditure (0.035, SE = 0.175, p > 0.10), vegetable purchases (−0.007, SE = 0.011, p > 0.10), protein purchases (0.024, SE = 0.019, p > 0.10), and diet quality (−0.261, SE = 0.182, p > 0.10). These findings confirm that the observed effects are specific to income‐eligible beneficiaries (Table A2).

TABLE A2.

Robustness checks: Population falsification.

Payment frequency	Population falsification: Age‐eligible, income‐ineligible (diet quality)	Population falsification: Age‐eligible, income‐ineligible (protein purchase)	Population falsification: Age‐eligible, income‐ineligible (vegetables purchase)	Population falsification: Age‐eligible, income‐ineligible (expenditure on food log)
Monthly payment	(Ref)	(Ref)	(Ref)	(Ref)
Bi‐monthly	−0.261	0.024	−0.007	0.035
	(0.182)	(0.019)	(0.011)	(0.175)
Observations	12,196	12,480	12,323	12,480

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Augmented Inverse Probability Weighting. The full set of control variables presented in descriptive Table 1 were used for matching and adjusting the Multivalued Treatment Models. The outcome models are specified as OLS Models. Separate models were run for each subsample.

The urban‐rural robustness check identifies that the effects of payment frequency on diet quality are slightly similar across urban and rural settings, with monthly payments yielding better dietary outcomes. In urban areas, beneficiaries receiving bi‐monthly payments had significantly lower diet quality compared to those receiving monthly payments (−0.608, SE = 0.330, p < 0.10). Similarly, in rural areas, bi‐monthly payments led to a greater reduction in diet quality compared to monthly payments (−0.692, SE = 0.163, p < 0.01).

Table A3 presents results from a DiD framework that leverages pre‐ and post‐policy variation in SSB consumption, comparing monthly versus bi‐monthly payments, to assess the robustness of our main findings. Consistent with theoretical expectations and our primary results from the multivalued treatment model, we find no statistically significant change in daily SSB consumption—a pattern indicative of behavioral inertia or stable long‐term preferences among habitual elderly consumers. In contrast, monthly payments are associated with a 10.9% point reduction in the probability of any SSB consumption in the past week, primarily driven by reductions among occasional consumers. This is further supported by a decline in weekly consumption frequency among non‐daily users. These findings are consistent with a liquidity‐based mechanism, wherein occasional consumers—who exhibit higher price and income elasticity—adjust their behavior in response to the timing of transfers. While the DiD framework estimates the ATT and our main analysis focuses on the ATE, the consistency in direction and statistical significance across estimands reinforces the internal validity of our findings.

TABLE A3.

Robustness checks: Difference‐in‐differences framework.

Payment frequency	Sugar‐sweetened beverages (daily)	Sugar‐sweetened beverages (last week)	Sugar‐sweetened beverages frequency (with daily consumers)	Sugar‐sweetened beverages frequency (without daily consumers)
Monthly payment	(Ref)	(Ref)	(Ref)	(Ref)
Bi‐monthly	0.037	0.109**	0.086	0.379***
	(0.037)	(0.052)	(0.222)	(0.097)
Observations	37,775	37,775	37,775	35,363

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Each column reports estimates from separate Difference‐in‐Differences OLS models evaluating the effect of bi‐monthly versus monthly cash transfer payments on sugar‐sweetened beverage (SSB) consumption. The dependent variables capture different dimensions of SSB intake: daily consumption, any consumption in the past week, weekly frequency among daily and non‐daily consumers. All models control for covariates and include municipality and year fixed effects. The full set of control variables are presented in descriptive Table 1. Separate models were run for each outcome.

Although the availability of only one pre‐policy year (2018) limits our ability to formally test the parallel trends assumption, the observed consumption patterns offer suggestive support for its plausibility. From 2018 to 2022, we observe a steady decline in the overall prevalence of weekly SSB consumption among older adult beneficiaries, falling from 46.8% to 34.4%. This downward trend is consistent with broader secular influences, such as national health campaigns, shifting preferences, or age‐related reductions in sugar intake (Figure A1).

FIGURE A1.

Weekly SSB consumption trends by payment frequency (2018–2022). The top panel displays the mean probability of weekly sugar‐sweetened beverage (SSB) consumption among all beneficiaries from 2018 to 2022. The bottom panel disaggregates this trend by payment frequency (monthly vs. bi‐monthly). The vertical dashed line marks the policy implementation in 2019.

Among recipients in municipalities with monthly payments, the declining trend begins prior to the reform and continues smoothly thereafter, reinforcing the interpretation of a broader behavioral shift. In contrast, beneficiaries in municipalities assigned to bi‐monthly payments display a distinct break in trend: despite starting at a lower consumption level in 2018, their weekly SSB consumption rises sharply in 2019—the year the policy was implemented—and subsequently levels off at a persistently higher level than their monthly counterparts. This post‐reform divergence, occurring against a backdrop of declining national trends, suggests that the shift in payment frequency may have disrupted otherwise similar behavioral trajectories.

While we cannot formally verify parallel trends due to the single pre‐treatment period, the opposing post‐reform trends between monthly and bi‐monthly groups—combined with a shared declining baseline in 2018—offer visual and descriptive evidence consistent with the identifying assumptions of the DiD framework. We therefore interpret the DiD estimates as a robustness check rather than a primary identification strategy. Nonetheless, the divergence in trends reinforces our main findings and strengthens the case for a causal interpretation: less frequent transfers appear to undermine reductions in SSB consumption, particularly among more elastic, light consumers.

Appendix 3. Heterogeneous Effects of Payment Frequency on Food Groups Among the Elderly: Multivalued Treatment Model

	(1)	(2)	(3)	(4)	(5)	(6)
Monthly versus Bi‐Monthly payment	Expenditure on food (log)	Ultra processed foods	Sugar sweetened beverages	Proteinpurchase	Vegetables purchase	Fruits purchase
	Long distance: 20+ minutes walking distance to nearest food store
Non‐program beneficiary	(Ref)	(Ref)	(Ref)	(Ref)	(Ref)	(Ref)
Bi‐monthly versus Monthly payment	0.522*	0.012	0.042**	0.075**	−0.017	−0.021
	(0.282)	(0.010)	(0.016)	(0.031)	(0.024)	(0.026)
Observations	8629	8629	8629	8629	8383	8383
	Short distance: < 20 min walking distance to nearest food store
Non‐program beneficiary	(Ref)	(Ref)	(Ref)	(Ref)	(Ref)	(Ref)
Bi‐monthly versus Monthly payment	−0.070	0.006	0.004	0.002	−0.021*	−0.044***
	(0.136)	(0.005)	0.010	(0.016)	(0.012)	(0.013)
Observations	44,001	44,001	44,001	44,001	42,654	42,654

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Augmented Inverse Probability Weighting. The full set of control variables presented in descriptive Table 1 were used for matching and adjusting the Multivalued Treatment Models. The outcome models are specified as OLS Models. Separate models were run for each subsample. The results presented in Appendix 4 reflect the differences in the outcomes between bi‐monthly and monthly beneficiaries, adjusted for non‐program participants as the reference group.

Appendix 4. Effect of Payment Frequency on Standardized Diet Quality Among Older Adults: Multivalued Treatment Model

	(1)
Payment frequency	Standardized diet quality score
Bi‐monthly versus Monthly payment	−0.109***
	(0.027)
Observations	51,024

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Augmented Inverse Probability Weighting. The full set of control variables presented in descriptive Table 1 were used for matching and adjusting the Multivalued Treatment Models. The outcome models are specified as OLS Models. The results presented in Appendix 4 reflect the differences in the effect of the Colombia Mayor program on the standardized diet quality score between bi‐monthly and monthly beneficiaries, adjusted for non‐program participants as the reference group.

Data Availability Statement

The data that support the findings of this study are available in Departamento Administrativo Nacional de Estadística (DANE) at https://www.dane.gov.co/index.php/estadisticas‐por‐tema/pobreza‐y‐condiciones‐de‐vida/calidad‐de‐vida‐ecv. These data were derived from the following resources available in the public domain: —Encuesta nacional de calidad de vida, https://www.dane.gov.co/index.php/estadisticas‐por‐tema/pobreza‐y‐condiciones‐de‐vida/calidad‐de‐vida‐ecv.