Awareness of need (perceived issue importance)

The first prerequisite for philanthropy is an awareness of need. In the charitable giving literature, to cite just a few examples, the perceived need for support has been linked with donations to homeless people (Lee & Farrell, Reference Lee and Farrell2003), international relief nonprofits (Cheung & Chan, Reference Cheung and Chan2000), and online charitable crowdfunding projects (Salido-Andres et al., Reference Salido-Andres, Rey-Garcia, Alvarez-Gonzalez and Vazquez-Casielles2021). This driver is unique in that needs may be tangible or intangible; needs may reside within, between, and outside people; needs originate from beneficiaries and fundraisers; and needs target potential donors to trigger the giving process (Bekkers & Wiepking, Reference Bekkers and Wiepking2011).

Relatedly, the awareness of an important public issue is a well-accepted driver of an individual citizen’s public policy attitudes and engagement. Such engagement can manifest in various forms such as voicing policy support or opposition (Zhang et al., Reference Zhang, Liu and Vedlitz2020b), electoral voting (Fournier et al., Reference Fournier, Blais, Nadeau, Gidengil and Nevitte2003), co-production (McLennan, Reference McLennan2020), or––as a few recent studies have documented––giving money to the government. For instance, Davidovic et al. (Reference Davidovic, Harring and Jagers2020) find that citizens holding stronger pro-environmental values are more willing to accept higher taxes to protect the environment; Zhang et al. (Reference Zhang, Liu and Vedlitz2020a) find that citizens priorly affected by flooding are more willing to accept higher water bills to help with local flood control; and Dvir et al. (Reference Dvir, Liu and Vedlitz2024) find that citizens more concerned with the state of local infrastructure are more willing to accept higher utility bills to support the construction of local infrastructure facilities. Meanwhile, particularly relevant to our study, Li et al. (Reference Li, Eckel, Grossman and Brown2011, Reference Li, Eckel, Grossman and Brown2015) learn through a lab experiment that people are more likely to make a one-time donation to a government agency if they believe that the agency serves an important cause.

Informed by and replicating the above knowledge, our first hypothesis is as follows:

Solicitation (government’s ask)

The second key external trigger is solicitation, which Bekkers and Wiepking (Reference Bekkers and Wiepking2011) frame as a tangible (e.g., a fundraising letter) or intangible (e.g., a verbal ask) interaction between people that originates from a beneficiary or fundraiser and targets a potential donor. In the charitable giving literature, it is generally agreed that “the most proximate cause of charitable gift is being asked to give” (Paxton, Reference Paxton2020, p. 547) and that “a large majority of all donation acts occurs in response to a solicitation” (Bekkers & Wiepking, Reference Bekkers and Wiepking2011, p. 931).

Of course, donation asks are not equally effective. In fact, there is clear evidence that many people try to avoid or reject solicitors (Andreoni et al., Reference Andreoni, Rao and Trachtman2017; DellaVigna et al., Reference DellaVigna, List and Malmendier2012; Exley & Petrie, Reference Exley and Petrie2018). What makes a donation ask successful, then? As Andreoni et al. (Reference Andreoni, Rao and Trachtman2017, p. 629) conclude, “the power of the ask derives from the emotional reaction it induces in a potential giver.” As a consequence, a donation ask tends to be more effective if it is in person and from a solicitor who has a good connection with the prospective donor (Gee & Meer, Reference Gee and Meer2020). Although this idea has been tested only in the nonprofit sector and not the public sector, there is room for us to conjecture that:

Material costs (requested donation amount)

Once a donation has been solicited, the material costs become salient. This driver is tangible, resides outside the donor, originates from the fundraiser, and affects the donor (Bekkers & Wiepking, Reference Bekkers and Wiepking2011). Indeed, “it is clear that giving money costs money” and thus “when the costs of a donation are lowered giving increases” (Bekkers & Wiepking, Reference Bekkers and Wiepking2011, p. 932). This idea has been tried out extensively in the charitable sector and corroborated by findings such as how a potential donor’s giving can be influenced by the tax deductibility level (Peloza & Steel, Reference Peloza and Steel2005), the donation matching rate (Karlan & List, Reference Karlan and List2007), and, perhaps most directly, the requested donation amount (Desmet & Feinberg, Reference Desmet and Feinberg2003). Hung et al.’s recent (Reference Hung, Lu and Hung2025) meta-analysis of this literature also confirms that donors are sensitive to the price of giving.

In the public sector, the cost sensitivity of citizens regarding monetary co-investment has been analyzed mostly by hypothetical willingness-to-pay experiments that appraise people’s willingness to accept a permanent tax increase or utility bill surcharge that would help fund government efforts to improve flood control (Zhang et al., Reference Zhang, Liu and Vedlitz2020a), clean energy (Jin et al., Reference Jin, Wan, Lin, Kuang and Ning2019; Mueller, Reference Mueller2013), overall infrastructure (Kotchen et al., Reference Kotchen, Turk and Leiserowitz2017), and so on. Unsurprisingly, these studies tend to find a negative effect of costs (i.e., the amount of tax or bill surcharge) on the probability of acceptance. As a result, our third hypothesis is as follows:

Altruism

Next, altruism is a mechanism that originates from a donor’s care for the tangible consequences of their gift for the beneficiaries (Bekkers & Wiepking, Reference Bekkers and Wiepking2011). Paxton (Reference Paxton2020, p. 547) defines altruism as “a desire to increase others’ well-being or the public good.” It is worth noting that altruism does not necessarily mean pure selflessness, as altruistic donors gain at least a “warm glow” feeling––a sense of satisfaction just from doing good (Andreoni et al., Reference Andreoni, Rao and Trachtman2017; Bekkers & Wiepking, Reference Bekkers and Wiepking2011; DellaVigna et al., Reference DellaVigna, List and Malmendier2012; Paxton, Reference Paxton2020; Slemrod & Kuchumova, Reference Slemrod and Kuchumova2023). Even as “impure altruists” (Bekkers & Wiepking, Reference Bekkers and Wiepking2011, p. 936), though, it is beyond debate that many people are willing to help total strangers even in anonymous settings (Andreoni et al., Reference Andreoni, Rao and Trachtman2017; Gee & Meer, Reference Gee and Meer2020).

The notion of altruism is more commonly studied in the nonprofit sector (e.g., Nowakowska, Reference Nowakowska2024), but it has also been grappled with in the public sector chiefly in conjunction with intertwined concepts like public service motivation (Piatak & Holt, Reference Piatak and Holt2020). A recent paper by Slemrod and Kuchumova (Reference Slemrod and Kuchumova2023) represents one of a few scholarly efforts so far to theorize about the effect of altruism (or at least the “warm glow” motivation) on citizens’ donations to their government. Although these scholars cannot directly measure individual citizens’ altruism, their theory is backed up by the indirect and suggestive fact that “in periods of popular wars the warm glow from gifts to government could be naturally elevated because of rise in patriotism and pride” (Slemrod & Kuchumova, Reference Slemrod and Kuchumova2023, p. 463). Thus, there is room for us to hypothesize that:

Efficacy (trust in government)

Lastly, the efficacy mechanism is an intangible driver located within the donor—a perception, generated by the fundraiser, that the donation will make a difference to the supported cause (Bekkers & Wiepking, Reference Bekkers and Wiepking2011). This perception is essential because donors may know about and have to choose among many worthy causes that need monetary support (Neumayr & Handy, Reference Neumayr and Handy2019). Accordingly, people tend to donate to organizations that they can trust to use the donated money well (Neumayr & Handy, Reference Neumayr and Handy2019).

Although this idea has been tested mostly in the nonprofit sector (e.g., Alhidari et al., Reference Alhidari, Veludo-de-Oliveira, Yousafzai and Yani-de-Soriano2018), it appears to also apply to trust in government––which Dvir et al. (Reference Dvir, Liu and Vedlitz2024, p. 2759) define as “individuals’ views of government ability to correctly and fairly execute its responsibilities.” Specifically, a few studies (Dvir et al., Reference Dvir, Liu and Vedlitz2024; Li et al., Reference Li, Eckel, Grossman and Brown2011; López-Pérez et al., Reference López-Pérez, Ramirez-Zamudio and Cruz-Martínez2024) have associated citizens’ trust in their government with their willingness to give to the government. It is worth noting how these studies operationalize trust differently, with Dvir et al. (Reference Dvir, Liu and Vedlitz2024) measuring a general level of trust in government, Li et al. (Reference Li, Eckel, Grossman and Brown2011) focusing on perceptions of the government’s quality of work, and López-Pérez et al. (Reference López-Pérez, Ramirez-Zamudio and Cruz-Martínez2024) considering perceptions of the government’s corruption. These differences reflect the emerging consensus in the literature on trust about the multidimensional nature of this concept (e.g., AbouAssi et al., Reference AbouAssi, Tran, Bowman and Johnston2024; Alhidari et al., Reference Alhidari, Veludo-de-Oliveira, Yousafzai and Yani-de-Soriano2018; Zhong et al., Reference Zhong, Su, Peng and Yang2017). Zhong et al.’s (Reference Zhong, Su, Peng and Yang2017) meta-analysis of management research on trust labels two main dimensions of trust: competence-based trust (the trustor’s beliefs about the trustee’s ability) versus goodwill-based trust (the trustor’s beliefs about the trustee’s integrity and benevolence). Correspondingly, it makes sense that citizens need to trust both the quality of work (Li et al., Reference Li, Eckel, Grossman and Brown2011) and the lack of corruption (López-Pérez et al., Reference López-Pérez, Ramirez-Zamudio and Cruz-Martínez2024) of their government in order to confidently entrust their donations to the government.

Hence, our last two hypotheses are as follows:

Research setting and data source

While most research on citizen co-investment has been based in Western, democratic, and developed countries, the present article examines citizen co-investment in Vietnam––a little-studied, Southeast-Asian, single-party, and developing country. Besides the well-recognized benefits of replication––a key tenet of the scientific method that has often been inadequate in public and nonprofit management research (Helmig et al., Reference Helmig, Spraul and Tremp2012; Pedersen & Stritch, Reference Pedersen and Stritch2018), Vietnam provides a valuable setting for our investigation of citizens’ donations to local governments to improve local infrastructure owing to the critical public infrastructure needs in the country. This is not only because Vietnam’s developing economy has been hindered by poor infrastructure capacity (Dang & Pheng, Reference Dang and Pheng2015), but also because Vietnam’s infrastructure is geographically vulnerable to some of the most dangerous climate-change threats (e.g., sea-level rise) in the world (Balboni, Reference Balboni2025). Facing such serious problems, Vietnamese local authorities in the last decade have been requesting citizens to donate resources to help build and repair roads and bridges in pursuit of infrastructure upgrades and economic development goals (Thanh & Pham, Reference Thanh and Pham2019). Moreover, in a single-party state, the government and its affiliated organizations are the primary channels for civic action, making donations to the government a more common and substantial form of participation than in nations with large, independent nonprofit sectors (Nguyen & Doan, Reference Nguyen, Doan, Wiepking and Handy2015). Additionally, while philanthropy and mutual assistance are deeply embedded in Vietnamese culture and tradition with a high rate of domestic giving, the monetary sums are often small and suggesting significant untapped potential (Nguyen & Doan, Reference Nguyen, Doan, Wiepking and Handy2015).

To study this phenomenon, our article relies on the 2023 data of the Vietnam Provincial Governance and Public Administration Performance Index (PAPI) project. PAPI is a citizen-participatory, public-policy-monitoring project funded and implemented every year mainly by the United Nations Development Programme (UNDP) and the Centre for Community Support and Development Studies (CECODES)––a local nongovernmental research organization “working in promoting civil society, strengthening its voice and its engagement with the state” (CECODES, 2024). As one of the largest citizen feedback collection projects in Vietnam, PAPI plays “an instrumental role in shifting central and local governments’ mindsets towards an openness to external reviews and citizens’ candid feedback on governance and public administration performance” and “assists central and local governments in identifying governance challenges and anticipating where central and local governments should focus their energy and resources” (CECODES et al., 2024, p. 1).

After two pilot rounds to validate and fine-tune survey methods, first in three provinces in 2009 and then in 30 provinces in 2010, since 2011 PAPI has been conducted every year in all 63 provinces in Vietnam, randomly sampling and surveying thousands of citizens annually via a rigorous sampling method to ensure they are representative of the Vietnamese population (CECODES et al., 2024). Specifically, PAPI uses a multistage, clustered-sampling strategy each year that randomly picks a couple of districts within each of the 63 provinces, a couple of communes within each of the selected districts, a couple of villages within each of the selected communes, and lastly about 20 adult citizens within each of the selected villages––the smallest public administrative tier in Vietnam (although communes are the smallest tier with a People’s Committee––the official local administrative authority). Despite PAPI’s nationally representative sampling and high response rate, we cannot completely rule out a non-response bias where non-respondents may differ somehow from respondents (e.g., non-respondents may be less civically engaged or harder to reach).

Our study relies on data from the 2023 PAPI project, which had an approximately 81% response rate and, besides many other public affairs topics, surveyed a random subset of about 9,000 Vietnamese citizens about their perceptions of and willingness to help improve local roads. Table 1 highlights the representativeness of this sample by comparing some demographic variables between our sample data and the latest Vietnamese census data, showing the closeness of the sample to the actual demographic characteristics of the Vietnamese population.

Table 1.

National representativeness of some demographic characteristics

Table 1. Long description

The table consists of three columns. The first column lists demographic characteristics. The second column contains data for Our 2023 sample where N equals 9,423. The third column contains data for the 2019 census where N equals 96,208,984.

* Female: 52.2 percent in the 2023 sample versus 50.2 percent in the 2019 census.

* Kinh ethnicity: 84.2 percent in the 2023 sample versus 85.3 percent in the 2019 census.

* Age 65 or older: 7.7 percent in the 2023 sample versus 7.6 percent in the 2019 census.

* Above highschool-level education among adult citizens: 17.3 percent in the 2023 sample versus 16.1 percent in the 2019 census.

* Poor household: 4.8 percent in the 2023 sample versus 3.8 percent in the 2019 census.

Variables

Table 2 provides the descriptive statistics of this study’s variables. Our only dependent variable is Willingness to Donate, a binary variable that equals 1 if a citizen answered “yes” (or 0 otherwise) to this question: “In the next year, if there is a decision to invest in building or repairing public roads in your locality, would you be willing to donate [randomly assigned amount of money] to your local government to help fund the project?” It is noteworthy that 77% of the surveyees answered “yes” to this question.

Table 2.

Descriptive statistics (N = 9,423)

Table 2. Long description

The table contains five columns: Variables, Mean, S D, Min, and Max. Key data points include:

* Willingness to donate: Mean 0.77, S D 0.42, Min 0, Max 1.

* Perceived issue importance: Mean 0.27, S D 0.44, Min 0, Max 1.

* Government solicitation categories (None, Village chief, Village party secretary, Commune committee chair, Elected village chief) all show a Mean of 0.20 and S D of 0.40.

* Requested donation amounts range from 100,000 V N D to 2,000,000 V N D, each with a Mean of 0.10 and S D of 0.30.

* Altruistic: Mean 0.71, S D 0.45.

* Perceived government quality: Mean 82.72, S D 19.79, Min 0, Max 100.

* Perceived government corruption: Mean 0.49, S D 0.50.

* Poor household: Mean 0.05, S D 0.22.

* Education level: Below highschool (Mean 0.58), Highschool graduate (Mean 0.25), Above highschool (Mean 0.17).

* Female: Mean 0.52, S D 0.50.

* Party member: Mean 0.12, S D 0.33.

To test our first hypothesis, our first independent variable is Perceived Issue Importance, a binary variable that equals 1 if a citizen answered “roads” (or 0 otherwise) to this open-ended question (which, importantly, was asked before the above question): “In the next year, if there is a decision to invest in building or repairing a type of public infrastructure in your locality, what do you think should be prioritized?” “Roads” was the most common answer to this question, followed by community centers, playgrounds, water systems, and so on. We acknowledge that capturing the first-to-mind response from an open-ended question may not be identical to asking about the importance of roads directly, which poses a measurement limitation.

To test our second hypothesis, our next set of independent variables include five binary variables that correspond to different conceivable sources (or lack) of local government pressure: personally requested by None (the control condition and reference category) to donate, by the Village’s Chief, by the Village’s Party Secretary, by the Commune’s People-Committee Chair, or by the Village’s Chief whom Citizen Elected. Among these options, the last category is the type of local government leader that an average Vietnamese citizen should have the closest connection with (compared with a generic village’s chief whom the citizen may not have chosen to elect, a village’s party secretary––who can be elected only by members of the Communist Party of Vietnam (CPV), or a commune’s People-Committee chair––who can be elected only by the local People’s Council (i.e., the local government’s legislative branch)). These five hypothetical conditions, which citizens were asked to imagine, were randomly assigned to the citizens to strengthen internal validity by ensuring that results would not be confounded by selection biases.

To test our third hypothesis, our next set of independent variables includes ten binary variables that correspond to ten different requested donation amounts ranging from 100,000 VND (the reference category) to 2,000,000 VND. These amounts were also randomly assigned to the citizens to bolster internal validity. Similar to previous willingness-to-pay experiments in the public sector (e.g., Jin et al., Reference Jin, Wan, Lin, Kuang and Ning2019; Kotchen et al., Reference Kotchen, Turk and Leiserowitz2017; Mueller, Reference Mueller2013; Zhang et al., Reference Zhang, Liu and Vedlitz2020a), each citizen was asked whether he/she would be willing to give the randomly assigned amount to support the proposed public good/service improvement (i.e., local public roads in this case).

To test our fourth hypothesis, our next independent variable is Altruistic, a binary variable that equals 1 if a citizen answered “yes” (or 0 otherwise) to this question: “To help alleviate poverty, the government sometimes transfers funds from richer provinces to poorer provinces. Would you be willing to pay more in taxes to pay for transfers to a poor province outside of your own?” This measure was adapted for our research context based on more common general measures of altruism such as “I have sympathy for people who are less fortunate than I am” or “I try to give generously to those in need” (Piatak & Holt, Reference Piatak and Holt2020). It is interesting that 71% of the surveyees showed an altruistic tendency here.

To test our last two hypotheses, our next two independent variables are Perceived Government Quality and Perceived Government Corruption. Perceived Government Quality is a continuous variable ranging from 0 to 100 and corresponding with each citizen’s rating of the quality of work by their commune’s People’s Committee––the official governmental body in charge of implementing local public administrative functions in Vietnam. As Table 2 indicates, the average rating is 82.72 out of 100. Meanwhile, Perceived Government Corruption is a binary variable that equals 1 if a citizen agreed (or 0 otherwise) that local government employees sometimes engaged in bribery or embezzlement. It is striking that 49% of the surveyees reported knowing about local government corruption, which aligns with longstanding concerns about severe corruption in Vietnam (e.g., see Tromme, Reference Tromme2016).

Finally, following prior research in both the nonprofit and public sectors, we include several individual demographic characteristics as control variables. To start with, income is obviously an imperative factor as it allows a donor to afford donating (Paxton, Reference Paxton2020). We control for income using the binary variable Poor Household, which equals 1 if a citizen’s household income met (or 0 otherwise) the poverty standard (i.e., about 1,500,000 VND or below per person per month in 2023) (Kim, Reference Kim2024). Next, the level of education is also an influential predictor of donations because education may boost prosocial motivation, awareness of need, and income itself (Paxton, Reference Paxton2020). We measure formal education with three binary variables: Below Highschool (the reference category), Highschool Graduate, and Above Highschool. Another factor is gender differences in giving, with most studies noticing that women are more likely to donate probably due to their stronger prosocial values (De Wit & Bekkers, Reference De Wit and Bekkers2016). We control for gender with the binary variable Female. Lastly, political affiliation and ideology, especially regarding the appropriate size and roles of the public sector, may also shape charitable contributions (Dvir et al., Reference Dvir, Liu and Vedlitz2024; Paarlberg et al., Reference Paarlberg, Nesbit, Clerkin and Christensen2019). We control for political affiliation using the binary variable Party Member, which equals 1 for CPV members (or 0 otherwise)––the only legal and, therefore, monopolistic political party in this one-party state.

Analytical methods

We use probit regression, a maximum-likelihood method of estimation suitable for binary outcome variables, to examine the above variables in six regression models as reported in Table 3. The first five models examine our five types of drivers separately, while the last model includes all of the variables in order to test all of the hypotheses together. We also control for village fixed effects in all of the models so that we can compare citizens living in the same village and thereby increase comparability and internal validity, as different localities in Vietnam may have different public infrastructure conditions, construction costs, cultures, and other potentially confounding characteristics (Balboni, Reference Balboni2025). As reported in the Appendix, we also conduct a series of supplemental analyses to bolster confidence in and add nuances to the findings from our main regression models. These include using cluster-robust standard errors to account for the clustered sampling design, a linear probability model for easier interpretation, an analysis to mitigate common source bias, and a comparison of stated donation willingness versus actual recent donation behaviors to assess the extent of hypothetical bias.

Table 3.

Main results

Table 3. Long description

A table titled Main results presenting Probit model coefficients for Willingness to donate across six numbered columns from 1 to 6.

Variables and key findings include:

* Perceived issue importance: Shows a positive significant effect in model 1 (0.344) and model 6 (0.334).

* Government solicitation: Using None as a reference category, Village's chief whom citizen elected shows a significant positive effect in model 2 (0.127) and model 6 (0.119).

* Requested donation amount: Using 100,000 V N D as a reference category, all higher amounts from 200,000 to 2,000,000 V N D show increasingly negative significant coefficients, reaching -1.804 for 2,000,000 V N D in model 6.

* Altruistic: Shows a strong positive significant effect in model 4 (0.669) and model 6 (0.713).

* Perceived government quality: Shows a small but significant positive effect (0.006) in models 5 and 6.

* Demographic controls: Poor household and Female variables consistently show negative significant effects across all models. Education level (Highschool and Above) and Party member status consistently show positive significant effects.

Bottom rows indicate Village fixed effects are included for all models. Pseudo-R2 values range from 0.11 to 0.24, with model 6 having the highest at 0.24. Total observations for all models are 9,423. Standard errors are provided in parentheses below coefficients. Significance levels are marked with asterisks: three for p less than 0.01, two for p less than 0.05, and one for p less than 0.1.

Notes: Probit models. Standard errors in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.1.