Unconditional cash transfer and polarized support

Abstract

Can money buy support in established democracies? Existing research on the political effects of cash transfers yields mixed findings, in part because most district-level analyses of ‘conditional’ cash transfer fail to distinguish between recipients and nonrecipients. This article advances a polarization hypothesis based on rational choice and motivated reasoning theories: unconditional cash transfers increase political support for incumbents only among the co-partisan recipients while leaving opposition supporters unmoved. We test this hypothesis by examining a unique ‘unconditional’ COVID-19 relief voucher lottery in Taiwan in 2021, in which vouchers were randomly assigned based on the last two digits of citizens’ national identification numbers. Combining lottery outcomes with an original individual-level survey (n = 776), we show that voucher receipts increased presidential approval only among respondents aligned with the ruling party. These findings suggest that even under conditions most favourable to programmatic redistribution, the political effects of cash transfers remain polarized.

Keywords

Cash transfers political polarization natural experiment democratic accountability Taiwan

Introduction

Can money buy political support in established democracies? Research on clientelism has long emphasized that brokers tend to target loyal supporters rather than swing voters (Stokes et al., 2013), relying on comprehensive social networks and repeated exchanges to sustain electoral dominance (Magaloni, 2006). However, as societies modernize and populations grow, the costs of sustaining clientelist networks increase, making such strategies less efficient over time (Wang and Kruzman, 2007); parties then switch to programmatic strategies to attract and sustain political support (Kitschelt and Kselman, 2013).

Despite its theoretical importance, the political effect of programmatic cash transfers remains contested. By definition, programmatic policies do not condition eligibility on partisan affiliation or vote choice, nor do they allow governments to monitor individual vote choices under secret balloting. As a result, voters face little direct pressure to offer their votes or reveal their preferences in exchange for the cash benefit. Some studies nevertheless show that cash transfers increase electoral support for incumbents (Araújo, 2021; De La O and Ana, 2013; Zucco, 2013), while others report null effects (Imai et al., 2020).

This inconsistency may reflect persistent methodological challenges. Most existing studies examine ‘conditional’ cash transfers, which is defined as programmes that target specific subpopulations such as the poor or those meeting predefined criteria. Because eligibility is not random, recipients may differ systematically from nonrecipients in ways that have been politically relevant.

Moreover, previous analyses are typically conducted at the district level, making it difficult to disentangle whether observed electoral gains arise from recipients themselves, from co-partisans who support the policy ideologically, or from other voters exposed to the policy’s local implementation. These limitations complicate causal inference and likely obscure heterogeneity in individual-level responses.

To illustrate the problem, consider a party that consistently attracts support from young families due to its emphasis on social welfare. If, after winning office, the party introduces a universal child benefit, the party’s continued support from the young families cannot be cleanly attributed to the policy implemented after winning. Even when the policy is formally programmatic, prior partisan alignment confounds estimates of its political effect. District-level analyses further exacerbate this issue by conflating treatment effects with spillovers, credit-claiming by opposition actors (Bueno, 2018), or limited policy awareness among beneficiaries (Mettler, 2011).

This article addresses these challenges by exploiting a unique case of unconditional cash transfers in Taiwan. In 2021, the Taiwanese government implemented a series of COVID-19 relief voucher lotteries to stimulate economic activity. Vouchers, which were valued between approximately USD $8 and $30, were randomly assigned based on the last two digits of citizens’ national identification numbers. Because the assignment was both universal and random, voucher receipts were independent of all individuals’ socioeconomic characteristics and political preferences, creating a perfect natural experiment (Dunning, 2012). Vouchers were distributed directly by the central government, prominently labelled as such, and more than 90% of recipients redeemed them within one year, ensuring high policy salience and compliance.

Building on rational choice and motivated reasoning theories, we advance a polarization hypothesis: cash transfers in democracies do not generate uniform increases in political support. Instead, their effects are conditional on prior political alignment. Supporters of the ruling party, especially those who directly receive transfers, are expected to respond positively, while opposition supporters remain largely unmoved or may even backfire the benefit as wasteful or illegitimate. This within-party heterogeneity cannot be identified without individual-level data distinguishing recipients from nonrecipients.

To test this hypothesis, we conducted an original online survey of Taiwanese citizens in December 2021, shortly after the voucher lotteries concluded. The survey measured presidential approval and collected respondents’ last two national identification digits, allowing us to link individuals to lottery outcomes without relying on self-reported benefit receipt. Randomization checks confirm that voucher receipt is independent of respondents’ demographic and political characteristics. Our results show that winning vouchers increased presidential approval only among respondents aligned with the ruling Democratic Progressive Party. Among opposition supporters, voucher receipt had no statistically discernible effect.

The political effect of cash transfer in democracies

If voters are homo economicus, they would support the party or candidate that maximizes their self-interest. From this perspective, cash transfers represent the most direct way for governments to improve voters’ pocketbooks, creating strong incentives for parties to distribute material benefits in exchange for political support. Both clientelist and programmatic strategies can be understood within this logic, differing primarily in whether benefits are conditional on vote choice.

However, voters are not purely self-interested actors (e.g. Wang, 2017), and the political consequences of cash transfers depend on how benefits are interpreted and attributed. In non-democratic settings, vote buying often backfires by undermining democratic norms and provoking electoral backlash (Carlin and Moseley, 2022). Even social welfare policies may generate negative reactions among groups that associate the state with repression or exclusion (Yang and Shen, 2021). These heterogeneous responses help to explain why brokers selectively target loyal supporters, who are perceived as more responsive to material inducements (Stokes et al., 2013). To sustain such exchanges, clientelist regimes must also invest heavily in monitoring mechanisms to ensure compliance (Rueda, 2017).

In established democracies, the ruling party could no longer monitor votes at the voting booth. Secret balloting allows voters to accept benefits without reciprocating electorally, and empirical evidence suggests that a substantial proportion of vote-buying recipients do exactly that (Wang and Kurzman, 2007). At the same time, competitive party systems foster long-term partisan attachments that shape how voters process political information (Aldrich, 1995). A large body of research shows that party identification structures information exposure and interpretation (Lau and Redlawsk, 2001; Lodge and Taber, 2005), influencing how voters assign credit and blame for government performance (Tilley and Hobolt, 2011).

Taken together, rational choice and motivated reasoning theories imply that the political effects of cash transfers in democracies should be conditional on prior political alignment. Supporters of the ruling party are more likely to view cash transfers as evidence of competent or benevolent governance, particularly when they directly receive the benefit. By contrast, opposition supporters may discount the same policy as fiscally irresponsible, inefficient, or electorally motivated. As a result, cash transfers may reinforce existing political divisions rather than generate broad-based electoral support.

This logic leads to a polarization hypothesis with within-party heterogeneity. Among supporters of the ruling party, recipients of cash transfers should respond more positively than nonrecipients, reflecting a direct material effect. Among supporters of opposition parties, however, receipt of the same benefit should have little or no effect on political support. The difference between recipients and nonrecipients within the ruling party thus captures the direct political effect of cash transfers, distinct from baseline partisan differences between political camps.

Testing this hypothesis requires individual-level data that distinguishes recipients from nonrecipients. Yet much of the existing literature relies on district-level analyses, often due to data limitations or concerns about self-reported benefit receipt. Studies of conditional cash transfers in Mexico, Brazil, Tanzania, and the United States, for example, primarily examine aggregate electoral outcomes (Croke, 2021; De La O and Ana, 2013; Kogan, 2021; Zucco, 2013). Even if ecological inference problems are set aside, such designs cannot disentangle whether observed effects reflect recipients’ self-interest, ideological support for redistribution, or motivated reasoning among co-partisans.

Moreover, conditional cash transfers pose additional challenges because eligibility is inherently non-random. Programmes typically target specific socioeconomic groups, whose political preferences may predate policy implementation. Even when transfers are formally programmatic, recipients may politicize or reinterpret them through partisan lenses, blurring causal inference (Kogan, 2021). These problems are further compounded when local governments or politicians claim credit for centrally funded programmes (Bueno, 2018), or when recipients fail to recognize that they have benefited from government policy at all (Mettler, 2011). In such cases, estimated effects often capture intent-to-treat rather than perceived treatment effects.

The unconditional COVID-19 relief voucher lotteries in Taiwan, 2021

Taiwan’s 2021 COVID-19 relief voucher lotteries provide an unusually well-suited setting for identifying the individual-level political effects of cash transfers. In response to economic disruption caused by the COVID-19 pandemic, the Taiwanese government first issued universal ‘Quintuple Stimulus Vouchers’ valued at NTD 5,000 (approximately USD 180). Subsequently, to support sectors disproportionately affected by public-health restrictions, the government implemented four additional voucher lotteries targeting the tourism, agriculture, sports and cultural industries.¹

Because budgetary constraints precluded universal distribution, the four additional vouchers valued between NTD 250 and NTD 1,000 (approximately USD 8–30) were assigned through public lotteries based on the last two digits of citizens’ national identification numbers. The drawings were conducted weekly in October 2021 by the Executive Yuan and relevant ministries and broadcast publicly via television and online platforms. Winners were notified by text message, and vouchers were made available in both digital and physical formats. Approximately 75% of Taiwanese citizens won at least one voucher (the full list is shown in Online Appendix Table A1), and more than 90% of winners redeemed them within one year.²

This policy design offers several methodological advantages. First, voucher assignment was purely random and universal in scope. The final two digits of Taiwanese national identification numbers are randomly generated during the birth registration, and lottery numbers were drawn independently of individual characteristics. As a result, voucher receipt is orthogonal to respondents’ socioeconomic background, political attitudes, and partisan affiliations. This randomization allows us to treat the voucher lotteries as a natural experiment (Dunning, 2012), an assumption which is empirically supported by balance tests presented later in the article.

Second, the use of identification digits allows us to avoid reliance on self-reported benefit receipt. Motivated reasoning may lead opposition supporters to underreport or downplay government benefits that they received, while others may misremember policy details (Mettler, 2011). By asking respondents to report only the last two digits of their identification numbers, which information carries no identifying content, we can match individuals to lottery outcomes directly. Because the vast majority of winners redeemed their vouchers, the distinction between intent-to-treat and treatment effects is minimal, allowing for a more precise estimate of the political impact of cash transfers.

Third, the attribution of credit is unusually clear in this case. Voucher lotteries were administered and announced by the central government, distributed through government platforms and post offices, and labelled explicitly as government programmes. Local officials and opposition parties therefore had little opportunity to claim credit for the policy. This institutional clarity maximizes the likelihood that any observed political response reflects attribution to the ruling party rather than to intermediary actors or self-claimed brokers.

Fourth, Taiwan’s political context strengthens the theoretical interpretation of the results. Unlike many Western democracies, Taiwan lacks a well-understood left–right ideological dimension in mass politics. The major parties do not differ systematically in their stated positions on welfare policy (Shim, 2020), and most voters are unable to locate themselves meaningfully on a left–right scale (Hsiao et al., 2017; Jou, 2010). Instead, Taiwan politics gradually formed a single unification−independence dimension on issue voting and has polarized since its first presidential election in 2000 (Achen and Wang, 2017; Wang, 2019). As a result, Taiwanese voters rely heavily on partisan and elite cues when evaluating policies (Chin and Wang, 2024). This context reduces the likelihood that observed heterogeneity in responses to cash transfers reflects ideological disagreement over redistribution, instead highlighting the role of partisan motivated reasoning.

Finally, although the monetary value of individual vouchers was modest relative to average monthly income, their high redemption rate indicates that recipients perceived them as meaningful. Winning a voucher required active attention to lottery announcements and deliberate effort to claim and use the benefit. This salience distinguishes the vouchers from less visible forms of redistribution, such as tax credits, and increases the likelihood that recipients associated the benefit with the incumbent government.

Taken together, Taiwan’s 2021 voucher lotteries combine random assignment, high policy visibility, centralized implementation, and a partisan cue-driven political environment. These features make the case particularly well-suited for isolating the direct political effects of unconditional cash transfers in a consolidated democracy.

Data collection

To test the polarization hypothesis, this article conducted an original online survey in Taiwan shortly after the completion of the 2021 COVID-19 voucher lotteries. The survey was administered between 13 and 16 December 2021, prior to Taiwan’s national referendums on 18 December. Fielding the survey during this period likely increased respondents’ political attentiveness without restricting participation to highly engaged individuals.

A total of 910 respondents completed the survey through Pollcracylab,³ an academic survey platform operated by National Chengchi University. Rather than relying on opt-in panels, Pollcracylab recruited respondents from prior probability-based telephone and face-to-face surveys drawn from official household registration records. As a result, its sample diversity exceeds that of commonly used online platforms such as MTurk. All participants are verified individuals, reducing concerns about automated or fraudulent responses (Chmielewski and Kucker, 2020). The study was approved by the institutional review board of the author’s affiliated institution (#UNLV-2021-189) and was pre-registered on Open Science Framework (https://osf.io/8pr4j) before data collection.

Table 1 reports the sociodemographic characteristics of the sample. Compared to the Taiwanese population, the sample includes a higher proportion of male, younger, and more highly educated respondents.⁴ However, key political variables, including national identity and party identification, closely resemble those reported in contemporaneous nationally representative surveys, mitigating concerns about bias in political attitudes.

Table 1.

Background of the Pollcracylab respondents (n = 910).

Gender	Male	553 (60.8%)
Gender	Female	357 (39.2%)
Age	20−29	158 (17.4%)
	30−39	278 (30.5%)
	40−49	245 (26.9%)
	50−59	172 (18.9%)
	60 upward	57 (6.2%)
Education	Senior High	60 (6.6%)
	Junior College	118 (13.0%)
	College	379 (41.6%)
	Graduate School	353 (38.8%)
Taiwanese identity	Taiwanese	547 (60.1%)
	Both Taiwanese and Chinese	348 (38.2%)
	Chinese	9 (1.0%)
	Others	6 (0.7%)
Party identification	Kuomintang	158 (17.4%)
	Democratic Progressive Party	223 (24.5%)
	Taiwan People’s Party	124 (13.6%)
	Non-partisan and others	405 (44.5%)

All respondents firstly received an invitation email from Pollcracylab, indicating that they were invited to participate in a survey titled ‘How do people discuss politics in daily life.’ They were informed that there were at most 40 items, the survey had passed the Institutional Review Board, and they could skip any item, but they must reach the last page to receive compensation (a gift card worth NTD 50) distributed by Pollcracylab directly.

Respondents were first asked a series of questions unrelated to the present study, including items on information consumption and attitudes toward cross-strait relations. Presidential approval was measured before questions about identity, partisanship, and demographics, to minimize post-treatment priming effects (Klar et al., 2020). Approval of President Tsai Ing-wen was measured by the question “Are you satisfied with the performance of President Tsai Ing-Wen? Very satisfied (+2, 14.8%), Satisfied (+1, 26.0%), Unsatisfied (−1, 23.6%), Very unsatisfied (−2, 35.3%), and no response (0, 0.2%).”

After completing political and demographic questions, respondents were asked to report the last two digits of their national identification numbers at the end of the survey. The question wording is: “In the end, we would like to invite you to provide the last two digits of your national identification number. We cannot identify your personal information through the last two digits, and Pollcracylab will not provide any identifiable information to the researchers. The two digits you provide will be only for research purposes.” The question explicitly stated that the digits could not be used to identify individuals and would be used solely for research purposes. No mention was made of the voucher lotteries, reducing the likelihood that respondents inferred the purpose of the item. Because the question appeared at the end of the survey and responses could not be revised, it could not influence earlier answers.

Since this question was asked at the end of the survey, it would not serve as a heuristic cue to influence respondents’ answers in other questions, given the fact that they were not allowed to change their answer retrospectively in the online questionnaire. There is no difference in the id reporting rate regarding age (T-test, p = 0.21) or gender (Chi-square test, p = 0.84), but those who reported the numbers have a slightly higher level of education (3.2) compared to those who did not (3.0) (T-test, p = 0.03), probably because of their higher numeric literacy. We do not think that this bias may influence overall inferences, and we rule out this possible effect from the level of education in the ‘Robustness check’ section.

We then matched the last two digits of respondents’ national ID to the list of voucher lottery, which is organized by an open-source website.⁵ The full list of the winning numbers can be found in the Online Appendix Table A1. In Online Appendix Table A1, twenty-five numbers (25%) failed to win any lottery. Among our respondents who provided their two digits (776 in 910, 85.3%), 560 won at least one lottery (72.1%), and 216 did not win any lottery (27.8%), which is close to the theoretical proportion. This number also suggests the high level of representativeness of our data. If voucher lottery losers tended to hide their identification numbers because of losing, we would obtain a higher winning percentage in our sample, which is not true in our case. We then create two independent variables: Winner (0/1; 1 means winning any lottery, and 0 otherwise); and Vouchers (0 to 4; number of vouchers won in the lotteries).

There is no relationship between Winner and age (T-test, p = 0.87), gender (Chi-square test, p = 0.70), and level of education (T-test, p = 0.51), and there is no relationship between Vouchers and age, gender, and level of education (Pearson’s correlation, p = 0.27, 0.87, and 0.53, respectively). The zero correlations suggest that the experimental design is indeed a natural experiment.

Taken together, the survey design and matching procedure allow us to identify the individual-level political effects of unconditional cash transfers while minimizing reporting bias and confounding.

Results

Table 2 shows seven ordinary least squares (OLS) regression models explaining the presidential approval rate after the government implemented the voucher lottery for unconditional cash transfer. Model 1 reports the average treatment effect of winning any voucher. The coefficient is positive but not statistically significant, indicating that unconditional cash transfers, on average, did not increase presidential approval. This null result masks substantial heterogeneity, however, which becomes apparent once partisan and identity-based interactions are introduced.

Table 2.

Interactive ordinary least squares on explaining the presidential approval rate by winning vouchers and partisan/national identity.

	DV: Presidential approval rate (−2 to +2)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Winning voucher (0/1)	0.182 (0.123)	0.195* (0.100)	−0.037 (0.163)	−0.034 (0.163)	0.241*** (0.107)	−0.005 (0.353)	−0.010 (0.153)
Taiwan identity (0/1)		1.832*** (0.092)	1.563*** (0.175)	1.589*** (0.176)
Winning × Taiwan ID			0.372* (0.206)	0.365* (0.206)
Pan-green supporter (0/1)					1.514*** (0.096)	1.166*** (0.181)	1.159*** (0.182)
Winning × pan-green						0.481** (0.213)	0.483** (0.214)
Age, gender, education				YES			YES
Intercept	−0.468*** (0.104)	−1.613*** (0.103)	−1.444*** (0.138)	−1.398*** (0.363)	−1.259*** (0.104)	−1.078*** (0.131)	−1.540*** (0.388)
n	776	776	776	776	776	776	776
Adjusted R-squared	0.002	0.338	0.340	0.341	0.244	0.248	0.247

p < 0.1. **p < 0.05. ***p < 0.01.

Models 2−4 examine heterogeneity by national identity. Taiwanese identifiers exhibit substantially higher baseline approval of the president, as expected. More importantly, the interaction between voucher receipt and Taiwanese identity is positive and statistically significant across specifications (p = 0.072 and 0.078 in Models 3 and 4, respectively). Among Taiwanese identifiers, winning a voucher increased presidential approval by approximately 0.35 points on the approval scale, while voucher receipt had no discernible effect among respondents who did not identify exclusively as Taiwanese.

Models 5−7 replicate this analysis using partisan identification instead of national identity. Consistent with the polarization hypothesis, pan-green supporters who won the voucher display a strong positive response. Winning a voucher increased presidential approval among pan-green respondents by approximately 0.48 points, a substantively meaningful effect. In contrast, voucher receipts had no statistically significant effect among non–pan-green respondents.

Unsurprisingly, in Table 1, Model 2 and Model 5 show that national and partisan identification contribute to a considerable proportion of explanatory power. Meanwhile, the increasing adjusted R-squared values between Models 2 and 3 and between Models 5 and 6 suggest that the interaction term between winning vouchers and group identity still contributes to additional explanatory power in explaining the variation in presidential approval rate.

To further illustrate the interactive effect in the regression models, Figure 1 shows the simulated heterogenous treatment effect from Models 4 and 7 in Table 2. In Table 2 and Figure 1, all other variables are controlled at the mean value, and the error bars represent the 95% of the simulated results. Both simulations show that winning vouchers significantly increased the presidential approval rate among pan-green supporters and Taiwanese identifiers. Meanwhile, winning itself has no effect among non-pan-green voters and dual identifiers.

Figure 1.

Simulated effect from Models 4 and 7 in Table 2.

In Table 3, we replaced winning any voucher (as a binary variable) in Table 2 with the number of vouchers (0 to 4) that respondents could have won. This replacement can further examine the psychological mechanism behind the shift and verify the robustness of the results in Table 2. The results basically follow the same pattern as those in Table 2. Similarly, by comparing the adjusted R- squared values between Models 2 and 3 and between Models 5 and 6, we show that the interaction terms between the number of vouchers won and group identity provide extra explanatory power on the presidential approval rates.

Table 3.

Interactive ordinary least squares on explaining presidential approval rate by number of vouchers won and identity.

	DV: Presidential approval rate (−2 to +2)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Number of vouchers won (0–4)	−0.014 (0.059)	0.037 (0.048)	−0.054 (0.072)	−0.055 (0.072)	0.027 (0.051)	−0.081 (0.069)	−0.083 (0.069)
Taiwan identity (0/1)		1.835*** (0.093)	1.652*** (0.143)	1.675*** (0.144)
Number of vouchers × Taiwan ID			0.162* (0.097)	0.160* (0.097)
Pan-green supporter (0/1)					1.509*** (0.096)	1.246*** (0.149)	1.237*** (0.149)
Number of vouchers × pan-green						0.238** (0.103)	0.241** (0.103)
Age, gender, education				YES			YES
Intercept	−0.320*** (0.085)	−1.515*** (0.092)	−1.408*** (0.112)	−1.357*** (0.359)	−1.113*** (0.090)	−0.998*** (0.105)	−1.496*** (0.383)
n	776	776	776	776	776	776	776
Adjusted R-squared	−0.001	0.335	0.337	0.338	0.240	0.244	0.243

p < 0.1. **p < 0.05. ***p < 0.01.

Figure 2 illustrates the simulated interactive effect between the number of vouchers and identity on the presidential approval rate, controlling all other variables at their means. In Figure 2, the bold lines are the expected values of two groups, while the shaded areas are the 95% distribution of the expected values. Both plots show that Taiwanese respondents became more polarized on the presidential approval rate once they won more vouchers. This polarization clearly illustrates the motivated reasoning behind the effect of nonconditional cash transfer.

Figure 2.

Simulated effect from Models 4 and 7 in Table 3.

Taken together, the results provide consistent evidence that unconditional cash transfers do not generate uniform political support. Instead, their effects are concentrated among individuals who already share partisan or identity-based alignment with the ruling party.

Robustness check

In Tables 2 and 3, the dependent variable in the regression model is coded as a continuous variable (approval rate, −2 to +2). However, it could be possible that the distribution does not follow the assumption of linear continuity. Therefore, in the Online Appendix Tables A2 and A3, ordinal regression is used for analysis. In Online Appendix Tables A2 and A3, the interaction terms between Winner, Voucher, Party, and Taiwan are all statistically significant, and the results are the same as Tables 2 and 3.

We further examine whether results depend on how treatment intensity is measured. Online Appendix Table A4 replaces the binary and count measures of voucher receipt with the total monetary value of vouchers won, which ranges from NTD 0 to 2.35k. Although the interaction terms are no longer statistically significant at conventional levels (p = 0.157, 0.157, 0.104, and 0.102 for Models 1, 2, 3, and 4, respectively), their signs remain consistent with the main results. This attenuation likely reflects how recipients perceive the treatment: respondents are more likely to remember winning a lottery than to recall or aggregate the precise monetary value of multiple vouchers, especially when vouchers are earmarked for different uses. This psychological mechanism follows the hot cognition hypothesis (Lodge and Taber, 2005) that people tend to update their effect toward specific topics in their mind, but will forget the details after the update; and this hypothesis may also explain the difference between Tables 2 (Winning any voucher) and 3 (Number of vouchers).

To further probe this interpretation, Online Appendix Table A5 examines nonlinearity by estimating the effects of all 16 possible combinations (four vouchers won/not won) of voucher receipt across the four lotteries.⁶ The analysis is conducted by an ordinal regression model, controlling for gender, age, and education. The results reveal no monotonic relationship between total voucher value and presidential approval. Among pan-green supporters, small wins are associated with positive evaluations of the president, while larger cumulative winnings do not produce additional gains. Among opposition supporters, receiving more vouchers is associated with weakly negative evaluations. These patterns are consistent with a motivated reasoning account in which affective updating occurs without precise tracking of the material magnitude.

Another alternative explanation for the main result is that people who paid more attention to politics would be more responsive to the cash transfer, rather than group identity. To falsify this hypothesis, variable Taiwan and its interaction with Voucher are replaced with the level of education in Models 1 and 2 in Table 3, assuming that the level of education could be a proxy for attention to politics.⁷ Both OLS regressions show that the interactions between the level of education and Voucher are not statistically significant (p = 0.561 and 0.595, respectively). Therefore, the attention hypothesis is not supported.

Taken together, these robustness checks reinforce the main conclusion: the political effects of unconditional cash transfers are polarized and conditional on prior political alignment, not on modelling choices, alternative treatment measures, or general political attentiveness.

Discussion

This article examines whether unconditional cash transfers generate political support in established democracies. Leveraging a rare natural experiment created by Taiwan’s 2021 COVID-19 relief voucher lotteries, we show that direct transfers do not produce uniform political returns. Instead, their effects are sharply polarized: only individuals who already align with the ruling party respond positively, while opposition supporters remain unmoved.

These findings contribute to the literature in several ways. First, they clarify why prior research on the electoral effects of cash transfers has yielded mixed results. When treatment effects are estimated at the aggregate level, heterogeneous individual responses are likely to cancel out, producing null or inconsistent findings. By distinguishing recipients from nonrecipients at the individual level, this study demonstrates that cash transfers can matter politically, but only within partisan in-groups.

Second, the results refine our understanding of programmatic redistribution in democracies. Even when transfers are unconditional, randomly assigned, highly salient, and implemented in a consolidated democracy, their political effects resemble those typically associated with clientelism. Rather than persuading undecided or opposition voters, transfers reinforce existing loyalties. This pattern suggests that the distinction between clientelist and programmatic linkages may be less about policy design and more about voter psychology; specifically, the result speaks directly to the role of motivated reasoning in attributing credit for government action.

Third, this study highlights a micro-level mechanism underlying strategic targeting in distributive politics. Previous research shows that political brokers disproportionately target loyal supporters because they expect higher returns per unit of resource spent (e.g. Magaloni, 2006; Stokes et al., 2013). Our findings provide individual-level evidence consistent with this logic: even small, randomly assigned benefits are interpreted through partisan lenses, leading co-partisans to reward incumbents while opponents discount or reinterpret the same benefits. This psychological asymmetry helps to explain why distributive strategies persist despite weak average effects.

The findings also carry broader implications for democratic accountability. If unconditional cash transfers primarily mobilize in-group supporters, their capacity to reduce polarization or broaden electoral coalitions may be limited. In highly polarized environments, redistribution may entrench partisan divisions rather than mitigate them. This dynamic suggests that the political effectiveness of welfare policies depends not only on their economic design but also on the structure of partisan competition and the prevalence of affective polarization.

Several limitations point to directions for future research. The vouchers examined here were relatively modest in value, and their effects may differ when transfers are larger or more frequent. In addition, Taiwan’s party system is characterized by strong partisan cues and weak ideological structuring (Wang, 2019), which may amplify motivated reasoning. Future studies could examine whether similar patterns emerge in multiparty systems with clearer ideological distinctions or lower levels of affective polarization (e.g. Dalton, 2008). Experimental designs that vary both the magnitude and framing of transfers may further illuminate how voters translate material benefits into political evaluations.

In sum, this article shows that money can still matter politically in democracies − but not in the way conventional models often assume. Unconditional cash transfers do not buy broad support; instead, they deepen existing political divides by reinforcing partisan loyalty. Understanding the political consequences of redistribution therefore requires integrating economic incentives with psychological mechanisms of political judgment.

Supplemental Material

sj-csv-2-ips-10.1177_01925121261447222 – Supplemental material for Unconditional cash transfer and polarized support

Supplemental material, sj-csv-2-ips-10.1177_01925121261447222 for Unconditional cash transfer and polarized support by Austin Horng-En Wang in International Political Science Review

Supplemental Material

sj-docx-1-ips-10.1177_01925121261447222 – Supplemental material for Unconditional cash transfer and polarized support

Supplemental material, sj-docx-1-ips-10.1177_01925121261447222 for Unconditional cash transfer and polarized support by Austin Horng-En Wang in International Political Science Review

Footnotes

Appendix

Appendix Table A5.

Interactive ordinal regression on explaining the presidential approval rate by the 16 groups of total amounts of monetary reward received and partisanship (n =776).

Group				Prize won ($NTD)	Number	Ordinal regression coefficient for approval
Travel	Sports	Food	ArtsFun	Prize won ($NTD)	Number	Prize won	Green	P × G
0	0	0	0	0	216	Base	1.493***	base
0	0	1	0	250	53	−0.021	-	1.808***
0	1	0	0	500	55	0.480	-	−0.276
0	0	0	1	600	184	0.028	-	0.679*
0	1	1	0	750	17	−1.217	-	2.497**
0	0	1	1	850	31	−0.327	-	0.883
1	0	0	0	1000	43	0.046	-	−0.27
0	1	0	1	1100	52	−0.032	-	0.825
1	0	1	0	1250	27	0.441	-	0.659
0	1	1	1	1350	16	−2.061*	-	2.526**
1	1	0	0	1500	11	−0.196	-	−0.069
1	0	0	1	1600	23	−0.405	-	1.150
1	1	1	0	1750	0	-	-	-
1	0	1	1	1850	0	-	-	-
1	1	0	1	2100	37	0.249	-	0.027
1	1	1	1	2350	11	−2.057*	-	1.001
ID not reported				-	134	-	-	-

p < 0.1. **p < 0.05. ***p < 0.01.

Acknowledgements

I am grateful to the reviewers for their valuable comments, which substantially improved the rigour of this article and strengthened its engagement with the literature. I also thank Peggy Chen and Wei-Ting Yen for their comments and assistance during the writing process. Any remaining errors are my own responsibility.

ORCID iD

Austin Horng-En Wang

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article’s data collection was primarily supported by the Department of Political Science at the University of Nevada, Las Vegas.

Declaration of conflicting interests

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

Supplemental material

Supplemental material for this article is available online.

Notes

Author biography

Austin Horng-En Wang is an Associate Professor in the Department of Political Science at the University of Nevada, Las Vegas. He received Ph.D. in Political Science from Duke University.

References

Achen

Christopher

Wang

Teh-Yu

(2017) The Taiwan Voter. Ann Arbor, Michigan: University of Michigan Press, 325.

Aldrich

John H

(1995) Why Parties?: The Origin and Transformation of Political Parties in America. Chicago, Illinois: University of Chicago Press.

Araújo

Victor

(2021) Do Anti-Poverty Policies Sway Voters? Evidence from a meta-analysis of conditional cash transfers. Research & Politics 8(1): 2053168021991715.

Bueno

Natália S

(2018) Bypassing the Enemy: Distributive politics, credit claiming, and nonstate organizations in Brazil. Comparative Political Studies 51(3): 304–340.

Carlin

Ryan E

Moseley

Mason W

(2022) When Clientelism Backfires: Vote buying, democratic attitudes, and electoral retaliation in Latin America. Political Research Quarterly 75(3): 766–781.

Chin

Wang

(2024) The Effect of Elite Cues on Policy Attitudes through the Lens of Affective Polarization in Taiwan: The potent force of animosity. Asian Survey 64(5): 730–755.

Chmielewski

Michael

Kucker

Sarah C

(2020) An MTurk Crisis? Shifts in data quality and the impact on study results. Social Psychological and Personality Science 11(4): 464–473.

Croke

Kevin

(2021) The Impact of Health Programs on Political Opinion: Evidence from Malaria control in Tanzania. The Journal of Politics 83(1): 340–353.

Dalton

Russell J

(2008) The Quantity and the Quality of Party Systems: Party system polarization, its measurement, and its consequences. Comparative Political Studies 41(7): 899–920.

10.

De La

Ana

(2013) Do Conditional Cash Transfers Affect Electoral Behavior? Evidence from a randomized experiment in Mexico. American Journal of Political Science 57(1): 1–14.

11.

Dunning

Thad

(2012) Natural Experiments in the Social Sciences: A Design-Based Approach. Cambridge: Cambridge University Press.

12.

Hsiao

Cheng

Achen

(2017) Political Left and Right in Taiwan. In Achen

Christopher H

Wang

(eds) The Taiwan Voter. Ann Arbor, Michigan: University of Michigan Press, 198.

13.

Imai

Kosuke

King

Gary

Rivera

C Velasco

(2020) Do Nonpartisan Programmatic Policies Have Partisan Electoral Effects? Evidence from two large-scale experiments. The Journal of Politics 82(2): 714–730.

14.

Jou

(2010) The Heuristic Value of the Left−Right Schema in East Asia. International Political Science Review 31(3): 366–394.

15.

Kitschelt

Herbert

Kselman

Daniel M

(2013) Economic Development, Democratic Experience, and Political Parties’ Linkage Strategies. Comparative Political Studies 46(11): 1453–1484.

16.

Klar

Samara

Leeper

Thomas

Robison

Joshua

(2020) Studying Identities with Experiments: Weighing the risk of posttreatment bias against priming effects. Journal of Experimental Political Science 7(1): 56–60.

17.

Kogan

Vladimir

(2021) Do Welfare Benefits Pay Electoral Dividends? Evidence from the National Food Stamp Program rollout. The Journal of Politics 83(1): 58–70.

18.

Lau

Richard R

Redlawsk

David P

(2001) Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making. American Journal of Political Science 45(4): 951–971.

19.

Lodge

Milton

Taber

Charles S

(2005) The Automaticity of Affect for Political Leaders, Groups, and Issues: An experimental test of the hot cognition hypothesis. Political Psychology 26(3): 455–482.

20.

Magaloni

Beatriz

(2006) Voting for Autocracy: Hegemonic Party Survival and Its Demise in Mexico. Cambridge: Cambridge University Press.

21.

Mettler

Suzanne

(2011) The Submerged State: How Invisible Government Policies Undermine American Democracy. Chicago, Illinois: University of Chicago Press.

22.

Rueda

Miguel R

(2017) Small Aggregates, Big Manipulation: Vote buying enforcement and collective monitoring. American Journal of Political Science 61(1): 163–177.

23.

Shim

(2020) Left Is Right and Right Is Left? Partisan difference on social welfare and particularistic benefits in Japan, South Korea and Taiwan. Journal of International and Comparative Social Policy 36(1): 25–41.

24.

Stokes

Susan C

Dunning

Thad

Nazareno

Marcelo

Brusco

Valeria

(2013) Brokers, Voters, and Clientelism: The Puzzle of Distributive Politics. Cambridge: Cambridge University Press.

25.

Tilley

James

Hobolt

Sara B

(2011) Is the Government to Blame? An experimental test of how partisanship shapes perceptions of performance and responsibility. The Journal of Politics 73(2): 316–330.

26.

Wang

Austin Horng-En

(2017) Patience as the Rational Foundation of Sociotropic Voting. Electoral Studies 50: 15–25.

27.

Wang

Austin Horng-En

(2019) The Myth of Polarization Among Taiwanese Voters: The missing middle. Journal of East Asian Studies 19(3): 275–287.

28.

Wang

Chin-shou

Kurzman

Charles

(2007) Dilemmas of Electoral Clientelism: Taiwan, 1993. International Political Science Review 28(2): 225–245.

29.

Yang

Wenhui

Shen

Xiaoxiao

(2021) Can Social Welfare Buy Mass Loyalty? Governance 34(4): 1213–1233.

30.

Zucco

Cesar

Jr (2013) When Payouts Pay Off: Conditional cash transfers and voting behavior in Brazil 2002–10. American Journal of Political Science 57(4): 810–822.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.16 MB

0.00 MB

0.02 MB