Life Satisfaction and Incumbent Voting: Examining the Mediating Effect of Trust in Government

Abstract

An emerging body of evidence suggests a causal relationship between wellbeing and incumbent voting. However, the evidence is primarily founded upon established democracies with regular turnovers of power. Moreover, the mechanism underlying this relationship is still relatively unknown. Using the intricacies of the Malaysian political context and Malaysian data from the seventh wave of the World Values Survey (WVS), this study examines the mediating role of trust in government to explain the relationship between life satisfaction and incumbent voting. Notably, the Malaysian WVS was concluded two weeks before Malaysia’s fourteenth general election (GE14), which witnessed the end of the ruling coalition’s six-decade hold on power since independence and subsequently ushered in the country’s first-ever peaceful transfer of power. The empirical analysis indicated that the mediating role of trust in the relationship between life satisfaction and incumbent voting is supported. Further empirical analysis also showed that the mediating effect of trust was unique to the GE14 context compared to GE13, thus providing a better understanding of the role trust plays in the outcome of the election. The results provide valuable insights and implications in political science, especially for a nation emerging from its authoritarian state.

1 Introduction

Are people with higher levels of life satisfaction more likely to vote for the incumbent government? Building on earlier studies that have concluded that people’s wellbeing affects their intention to vote (e.g., Flavin & Keane, 2012; Weitz-Shapiro & Winters, 2011; Zhong & Chen, 2002), research has found that there is an electoral dividend for incumbent governments to focus on improving people’s wellbeing beyond ensuring a healthy economy (Ward, 2020; Ward et al., 2021).

Although the literature linking life satisfaction to voting is new (Ward et al., 2021), there is increasing evidence that the electoral fate of incumbent governments is connected to the citizens’ level of wellbeing, beyond the traditional macroeconomic indicators usually associated with the incumbent’s re-election chances. This phenomenon has been observed in Britain (Liberini et al., 2017), Europe (Ward, 2020), Latin America (Bravo, 2016) and the United States of America (USA) (Ward et al., 2021). However, the literature has primarily focused on established democracies with regular turnovers of power. In contrast, there is a paucity of literature examining this link in countries with dominant-party systems, where a single political party dominates consecutive elections, and turnovers of power are rare. Notably, Ng et al. (2017) had examined this link for the case of Malaysia, a country that had a 60-year uninterrupted rule by the National Front (Barisan Nasional, BN) coalition. The study found that life satisfaction can predict incumbent support, besides the entrenched ethnicity and rural–urban factors in the Malaysian context. However, shortly after in 2018, the Malaysian 14th General Election (GE14) resulted in the unexpected downfall of the BN coalition, bringing their hold on power to an abrupt end. Subsequently, the country witnessed its first-ever regime change and peaceful transfer of power from the BN to the opposition at the time, the Alliance of Hope (Pakatan Harapan, PH) coalition.

The first regime change in Malaysia’s authoritarian state has been studied through the lens of elite fragmentation,¹ whereby internal conflicts within the BN led to party defections and other institutional changes that eventually weakened the BN’s grip on power (Hazis, 2020; Welsh, 2018). For the commoner, there was electorate anger against the then-Prime Minister Najib Razak due to economic mismanagement and corruption scandals (Hutchinson, 2018; Rahman, 2018; Welsh, 2018). We argue that subjective wellbeing can be viewed as the common psychological pathway to electoral choice in these accounts. The increased economic vulnerabilities, corruption scandals and resentment against Najib (Hazis, 2020) arguably reduced the subjective wellbeing of the people, decreased the preference for status-quo and drove the voters against the incumbent government.

Therefore, using the particularities of the Malaysian electoral context, this study empirically examines the relationship between life satisfaction and incumbent voting in a country that experienced its first turnover of power after more than half a century since its independence. It contributes to the growing literature on life satisfaction and voting, responding to the call by Ward et al. (2021) to determine the link between life satisfaction and voting in other electoral contexts beyond the USA. Moreover, as the mechanism through which life satisfaction affects incumbent voting is still relatively unexplored in the literature, this study further extends the work of Ng et al. (2017) by examining whether life satisfaction affects incumbent voting through the mediating effect of the electorate’s trust in the government. In doing so, this study answers the call by Ward (2019) for more work to explain the current research findings, albeit in the Malaysian context.

We propagate this discussion using individual-level Malaysian data from the seventh wave of the World Values Survey (WVS7), undertaken shortly before GE14 in 2018. Given the short time lag between the conduct of the WVS7 and GE14, the survey responses can shed light on how respondents’ life satisfaction and trust in the government influenced their voting preference.

To better understand how the mediating effect of trust contributed to the first turnover of the federal government in the nation’s history, we extend our analysis by repeating the mediation analysis using data from the sixth wave of the WVS (WVS6). We leverage on the advantage that the WVS6 was, coincidentally, also undertaken less than a year before Malaysia’s thirteenth general election (GE13) in 2013 (Ng et al., 2017). By comparing the mediation analysis over these two time points, we show how the mediating effect of trust evolved over the 5-year gap between the two GEs, illuminating the loss of power by the BN arising from a deficit in trust due to a drop in the levels of life satisfaction.

1.1 Contextual Background of GE14

The seeds to the downfall of the BN in GE14 had been sown in the two prior general elections. In the 2008 general election (GE12), the opposition coalition deprived the BN of its customary two-thirds majority in parliament. In the 2013 general election (GE13), the BN won the simple majority of seats, although it had garnered less than 50% of the popular vote. The gains made by the opposition led some to believe that ethnic voting is no longer a significant factor in determining incumbent voting (Case, 2013; Ng et al., 2015; Pepinsky, 2009), although most of the political parties/coalitions draw their support from ethnic electoral bases.

Political analysts have attributed the spectacular downfall of the BN in GE14 to several factors, the chief of which was arguably the revelation of the 1Malaysia Development Berhad (1MDB) financial scandal that mired the incumbent prime minister Najib Razak (Wright & Hope, 2019). It involved USD681 million purportedly channelled into his bank account, which Najib claims were a Saudi royal family donation (Abdullah, 2019). The scandal was labelled by the former US attorney-general, Eric Holder, as ‘kleptocracy at its worst’. Najib also silenced any dissent from within his ranks by sacking several of his cabinet ministers who spoke up about the 1MDB scandal, including the deputy prime minister at the time, Muhyiddin Yassin (Teoh, 2015).

Despite being embroiled in the financial scandal, Najib, as the finance minister, implemented a deeply unpopular goods and services tax (GST) in April 2015, replacing the sales and services tax (SST). The GST was highly unpopular as it was a broad-based tax, shifting more burden to the electorate than the SST. This inevitably increased the economic vulnerabilities of the electorate (Welsh, 2018). FinanceAsia subsequently named Najib the worst finance minister in 2016 (Middleton, 2016).

Following his sacking, Muhyiddin joined forces with former prime minister Mahathir Mohamad to form the Malaysian United Indigenous Party (Bersatu). Bersatu subsequently joined the PH as a coalition member on 31 March 2017, 1 year before GE14. Mahathir proved to be the lynchpin that galvanised Malay support for the PH (Abdullah, 2019), propelling the coalition towards victory, where earlier attempts in GE12 and GE13 failed to unseat the ruling BN coalition.

As Malaysia’s longest-serving prime minister, Mahathir could be seen as the charismatic leader who galvanised support for the then-opposition PH, rallying the nation with a simple message—to overthrow the corrupt regime of Najib and the BN (Abdullah, 2019; Pillai & Williams, 1998; Williams et al., 2009). The election would eventually become a credibility contest between the two coalitions’ leaders (Abdullah, 2019). With Mahathir in the alliance, the Malays who were dissatisfied with Najib’s administration could eventually vote for the PH as Mahathir was someone ‘they had trust in’ (Abdullah, 2019, p. 532). Further, the economic and financial mismanagement by the Najib administration, coupled with the personal financial scandals, had also led to erosion of the economic wellbeing of the citizens, resulting in trust deficits in his government. Against this backdrop, this study seeks to empirically test the mediating effect of trust in government between life satisfaction and incumbent voting.

2 Literature Review and Conceptual Framework

2.1 Life Satisfaction and Incumbent Voting

A growing body of evidence suggests a causal link between life satisfaction and incumbent voting. Using panel data from the British Household Panel Survey, Liberini et al. (2017) were among the first to directly analyse the effect of wellbeing on incumbent support by augmenting standard voting models with wellbeing indicators. Their results suggest that the respondents’ voting intentions are sensitive to changes in their levels of life satisfaction. They further establish causality between life satisfaction and incumbent support by leveraging exogenous shocks to subjective wellbeing caused by widowhood. Ward (2020) built on the work of Liberini et al. (2017) by studying a 40-year panel of general elections in 15 European countries and found consistent evidence that individuals who were more satisfied with life were more likely to support a governing party. In the USA, Ward et al. (2021) evidenced that low levels of subjective wellbeing predicted anti-incumbent voting at the 2012 election and were predictive of Donald Trump’s 2016 electoral victory. The same results hold in Latin America, where the incumbent party is likely to win presidential elections when the average life satisfaction of Latin Americans is high (Bravo, 2016).

As posited by Ward et al. (2021), this phenomenon can be viewed through the lens of the affect-as-information theory (Schwarz, 2011), whereby people use their feelings as a source of information that guides their decision-making. This dependence on feelings becomes accentuated when people find it difficult to make political vote choices due to their limited grasp of complicated political and economic matters (Campbell et al., 1960). In addition, voters’ decisions may also be subject to the “availability heuristic” (Tversky & Kahneman, 1973). For instance, voters base their voting decisions on easily recallable pieces of information, such as their recent life satisfaction, rather than a complete evaluation of the entire incumbent performance (Ward, 2020). Taken together, whether voters vote for the incumbent is partly dependent on how they feel their life has been under the incumbent’s administration.

Despite the increasing evidence in happiness studies that link life satisfaction to incumbent voting, the causal mechanism through which it occurs is rarely discussed or empirically tested. Therefore, using the intricacies of the Malaysian context, this study empirically tests whether life satisfaction influences incumbent voting through trust in government. In doing so, we contribute to the growing interest in examining why life satisfaction affects voting behaviour (Ward, 2019).

2.2 Drivers of Vote Choice in Malaysia

Life satisfaction has also been shown to influence voting preferences in Malaysia (Ng et al., 2017). However, there are at least four contextual motivations in determining vote choice in Malaysia. First, during BN’s term in power, the rapid economic growth that Malaysia enjoyed during the first 40 years of independence improved the livelihood of most Malaysians. This track record of economic prosperity and the ensuing political and ethnic stability enabled BN to claim credit for it (Mutalib, 2000). Therefore, to preserve the status quo, Malaysians continued to vote for the BN to keep the current state of affairs to ensure the country’s continued growth (Ng et al., 2017). This outcome is consistent with the literature on retrospective voting, which suggests that the electoral success of incumbent governments is closely linked to the voters’ evaluation of incumbent performance (Healy & Malhotra, 2013). These studies have typically focused on the economic performance, concluding that good economic conditions translate into incumbent support (see Ashworth, 2012; Healy & Malhotra, 2013; Lewis-Beck & Stegmaier, 2013 for a review).

Second, the ethnic makeup of the Malaysian population has historically led voters to vote along racial lines (Welsh, 2014). During the long reign of the BN coalition, its main component parties appealed to the different ethnic groups for support. The United Malays National Organisation (UMNO) appealed to the Malays, the Malaysian Chinese Association (MCA) to the Chinese, and the Malaysian Indian Congress (MIC) the Indian community. Subsequently, the BN had managed to “co-opt most of the middle ground of politics” (Weiss, 2014, p. 4), allowing it to fare better than the opposition in ethnically mixed constituencies. It is helpful to note that the consociational nature of the BN coalition to govern a plural society such as Malaysia can be traced back to its struggle for independence from British rule in 1957 (Hutchinson, 2018). In particular, the “British preference for politically moderate and multi-ethnic leadership…emerged as the main interlocutor in negotiations regarding the timing, form, and duration of the independence process” (p. 584). Therefore, the foundation of Malaysian politics was based on the unity and cooperation of the major ethnic groups that led to Malaysia’s independence.

Third, the rural–urban differences have also been found to give rise to a relatively consistent voting pattern since GE12, with urban voters having a significant propensity to vote anti-establishment whereas rural voters continue to vote for the incumbent BN (Ng et al., 2015, 2020).

Fourth, the dominance of the BN coalition rule has given rise to clientelism and political patronage afforded by their incumbency advantage of government resources, to the extent that voters expect developmental promises and resources to be a given for their continued support of the incumbent BN regime (Weiss, 2014; Welsh, 2014). This was further perpetuated with the intertwining of the business establishment with UMNO, especially during Mahathir Mohamad’s first stint as prime minister (Balasubramaniam, 2006).

Since GE14, the political outcomes are now less certain than before. The ever-changing Malaysian political dynamics continue to intrigue political researchers post-GE14 (Hutchinson, 2018). Subsequent shifts in political alignment and allegiance have resulted in no coalition or party having complete dominance as before. This potentially signals the beginning of Malaysia’s move away from being an authoritarian state where more frequent transfer of power could occur. Subsequently, besides the factors above that influence Malaysians’ voting preferences, this study examines if there are any electoral dividends to gain from voters’ life satisfaction in the context of a nation emerging from its authoritarian state and if trust has a significant role in facilitating that dividend.

2.3 Hypothesis Development

In political support research, life satisfaction has been found to predict political trust at the country level (Veenhoven, 2009). At the individual level, life satisfaction predicts democratic satisfaction (Esaiasson et al., 2020; Hooghe, 2012); it also moderates the negative impact of perceived corruption on political participation (Zheng et al., 2017). However, political system support researchers rarely provide theoretical justifications for why life satisfaction is causally linked to perceived political performance. The commonly used social success and wellbeing theory primarily focus on social trust as the outcome of interest rather than political system support (Delhey & Newton, 2003). Very few wellbeing research studies target political fate outcomes.

In a recent study, Esaiasson et al. (2020) claim that citizens behave as if they have signed a psychological happiness contract with the government and will subsequently judge the government according to their happiness level. This contract asserts that citizens will hold their government accountable for delivering conditions that allow them to reach a satisfactory level of wellbeing. Building on this psychological happiness contract, we also expect that when citizens are called upon to judge if they trust the government, they will evaluate whether their life is satisfied enough to merit trust.

In any political system, citizens are generally able to withstand a certain level of dysfunction and underperformance, which are regarded as reasonable and unavoidable (Bélanger, 2017). However, when they perceive a significant deterioration in the quality of performance by the incumbent, citizen dissatisfaction towards the incumbent will grow, subsequently generating feelings of distrust and low confidence (Newton & Norris, 2000). These feelings of distrust will have consequences for vote choice. Specifically, political distrust gives rise to anti-incumbent behaviour. Those with lower trust in government are more likely to vote for non-incumbents (Citrin & Green, 1986; Dalton & Weldon, 2005; Hetherington, 1999). This pattern is magnified in political systems where voting is mandatory (Bélanger, 2017). Therefore, we propose the following hypotheses:

H1

There is a positive relationship between life satisfaction and trust in government.

H2

There is a positive relationship between trust in government and incumbent voting.

The Hirschman (1970) framework helps explain this phenomenon. Under this framework, when the level of dissatisfaction exceeds a certain threshold, distrustful citizens either abstain from voting or express their discontent by voting for a challenger party.² The former happens when distrust is too high to the extent that people believe that the political system has broken down, rendering political parties irrelevant. In the latter context, the citizens’ distrust of the incumbent is expressed by abandoning the incumbent and voting for the challenger party. This happens because citizens who distrust the government perceive it as ‘incapable of addressing their concerns, and hence they may welcome other options’ (Hetherington, 1999). In other words, the challenger party acts as a vehicle for voters to channel their dissatisfaction.

To conclude, based on the theoretical underpinnings developed above, we predict that trust in government will serve as a mediator between life satisfaction and incumbent voting, particularly in the contextual premise specific to citizens’ discontent of the incumbent. Context is defined as “situational or environmental constraints” (Johns, 2017, p. 577) that can affect the occurrence and meaning that can shape behaviour (see Johns, 2006, 2017 for a recent review). Here, contextual effects can explain the underlying behaviour (Johns, 2006; Whetten, 2009). Thus, we hypothesise that higher levels of life satisfaction will predict higher levels of trust in government, which will, in turn, predict a higher probability of a respondent voting for the incumbent party (i.e., BN). This gives rise to the following hypothesis:

H3

Trust mediates the relationship between life satisfaction and voting for the incumbent ruling party.

Figure 1 provides the conceptual model used to test the above three hypotheses. Specifically, it displays the relationship between life satisfaction and incumbent voting, with trust in government as a mediator. In Fig. 1, a, b and c represent the weights of the causal effects of life satisfaction (LS) on trust (TRUST), trust on voting (VOTE), and the direct effect of life satisfaction on voting, respectively. Mediation is said to occur if the indirect effect, measured by the product of a and b (i.e., ab), is significant.

Fig. 1

Conceptual model

To show how trust’s mediating effect helps one understand the GE14 outcome, we apply the conceptual framework depicted in Fig. 1 to data from WVS6 and WVS7, respectively, and compare the subsequent mediation results. WVS7 was concluded about two weeks before GE14, while WVS6 was completed less than a year before GE13. The BN returned to power in the 2013 election, while it lost in the latter. Therefore, the comparative analysis aims to determine if the mediating effect of trust, if any, is unique to the GE14 context. The intuition is that if the levels of LS and TRUST had remained the same over these two time points, and the mediation analysis for these two time points returned the same result, then any evidence of the mediating effect of trust will not explain the GE14 election outcome as it is not specific to the GE14 context. Thus, we test the following hypothesis:

H4

Life satisfaction and trust levels are lower in GE14 (WVS7) than in GE13 (WVS6), and the mediating effect of trust in government on the relationship between life satisfaction and incumbent voting is significant in GE14, but not in GE13.

Last, to address possible endogeneity concerns about the relationship between political trust and voting due to projection and persuasion biases³ (Miller & Shanks, 1996), Bélanger (2017) offered counterarguments to mitigate such concerns, justifying the use of political trust as an independent variable in empirical analyses. For example, Bélanger (2017) argued that conceptualising trust as an endogenous variable is incongruent with the idea of a “significant reservoir of dissatisfaction towards political institutions among citizens” (p. 249). Particularly, the rise of any challenger parties, such as PH in the Malaysian GE14 context, requires as a condition the prior existence of a significant reservoir of political dissatisfaction among the electorate. Furthermore, the author argues that studies have shown that trust is only weakly related to voters’ pre-established partisan identification (Bélanger & Nadeau, 2005; Clarke & Kornberg, 1993).

3 Data and Methodology

3.1 Data

To test the hypotheses H1 to H3, we utilise WVS7 which surveyed a nationally representative sample of respondents in Malaysia. To capture the electorate’s preference for the incumbent ruling coalition (i.e., BN), we used the following question as the dependent variable, VOTE: ‘If there were a national election tomorrow, for which party on this list would you vote for?’ The variable VOTE takes a value of 1 if the respondent chooses to vote for the BN, and 0 otherwise. Similar to Ward (2020), we exclude responses from respondents who indicated that they have never voted or are not allowed to vote. This leaves a total of 893 ‘likely voters’ who either usually or always vote at national elections. It should be noted that the WVS for Malaysia was concluded on 27 April 2018, shortly before GE14, which was held on 9 May 2018. Thus, the responses to this question can be considered reliable and representative of respondents’ actual voting behaviour. To further demonstrate the representativeness of the data, Table 1 compares the actual BN vote share in GE14 for each of the Malaysian states with the proportion of WVS respondents who indicated that they would vote for the BN. Table 1 shows that the actual BN vote share in GE14 is similar to the reported vote share in the WVS, except for a few states. The WVS significantly overestimates the BN vote share for Kelantan and Terengganu. The WVS indicated that more than 50% of respondents would vote for the BN for these two states, while the actual GE14 result revealed otherwise. Although the WVS also overestimated the BN vote share for WP Kuala Lumpur, both the WVS and GE14 vote shares are below the 40% threshold, indicating the state voters’ preference for the opposition party.

Table 1 Comparison of the BN vote share in GE14 and in the WVS

The main independent variables in this study are life satisfaction (LS) and trust in government (TRUST). To measure life satisfaction, we follow convention by using the responses to the following question: ‘All things considered, how satisfied are you with your life as a whole these days?’ Respondents rate their life satisfaction on a scale from 1 (completely dissatisfied) to 10 (completely satisfied). To enable comparisons, we standardised this variable into a z-score, such that it has a mean of 0 and a standard deviation of 1. The electorate’s trust in government is measured through responses to the question ‘Could you tell me how much confidence you have in government: 1 = none at all; 2 = not very much confidence; 3 = quite a lot of confidence; 4 = great deal of confidence?’ We also standardised this variable into a z-score. It should be noted that researchers often treat political trust as citizen’s confidence in political institutions (Catterberg & Moreno, 2006; Levi & Stoker, 2000; Mariën & Hooghe, 2011). Moreover, there is evidence that respondents cannot draw a meaningful separation between the concepts of trust and confidence as the distinctions between them are too narrow (ONS, 2016).⁴ Therefore, we use the ‘confidence’ question to measure trust in government.

We control for the confounding effects of other variables by including the same set of control variables used by Ng et al. (2017). These include respondents’ individual characteristics, income indicators, education indicators and labour force characteristics. The respondents’ ethnicity is also included in the model as an independent variable to account for its influence on voting preferences as established in the literature. Finally, dummy variables for the Malaysian states are also included to control for the differences between the states in terms of history, culture and economic conditions. The state dummies also account for how states differ in their voting preferences for the respective parties. This is necessary given the ‘resilience of regional identities’ (Hutchinson, 2018, p. 600), whereby the political dynamics in the peninsula’s west coast are starkly different from those in the northern and eastern states. In the same vein, Ostwald and Oliver (2020) also observed that Malaysia essentially comprises four electoral regions with distinct voting preferences. Except for the state dummy variables, the details for each control variable are listed in Table 2.⁵

Table 2 Definition of control variables

To test hypothesis H4, we include data from WVS6 and calculate the values of the same variables computed from WVS7 (i.e., VOTE, LS, TRUST and control variables listed in Table 2). The WVS6 was concluded on 15 June 2012, less than a year before GE13 was held. The sample contains 1121 respondents.

Our comparative analysis between the two elections is unlikely to be affected by the different electoral landscapes of GE13 and GE14. GE13 saw the contest between just two coalitions, BN and the then-opposition, People’s Alliance (Pakatan Rakyat, PR). In contrast, GE14 was marked by numerous three-way contests in the peninsula between BN, PH and a smaller coalition led by the Malaysian Islamic Party (Parti Islam Se-Malaysia, PAS). However, any empirical concern regarding splitting the incumbent vote does not arise as this would only occur if one were analysing actual vote share aggregated at the constituency level, as per Ng et al. (2020). Our analysis is based on the context of individual-level survey data whereby respondents indicate which party they would vote for if an election were held the next day. In other words, respondents freely indicating which party they would vote for in the WVS is unlike actual polling day, where splitting of the incumbent vote may arise in constituencies with three-way contests.

3.2 Methodology

The mediation analysis consists of testing hypotheses H1, H2 and H3, which are constituent components of the indirect effects in the life satisfaction–voting relationship. H4 compares the significance of the mediating effect of trust in the two GEs.

We conduct the empirical analysis using PROCESS, a macro implemented in the IBM SPSS version 20.0 software developed by Hayes (2012). PROCESS is a computational tool dedicated to path analyses with observed variables. It contains 92 pre-programmed models for estimating and testing various causal pathway relationships, including direct and indirect effects, conditional effects and conditional indirect effects, making it particularly suitable for mediation, moderation, and conditional process analysis. Sarstedt et al. (2020) highlighted the merit of PROCESS in estimating mediation and conditional processes for single-item observable variables.

The empirical analysis is conducted using PROCESS Model 4 (Hayes, 2018) to examine the mediating effect of trust on the relationship between life satisfaction and incumbent voting. Model 4 is used for testing H1, H2 and H3 as its specification aligns with our conceptual framework. Further, Model 4 is estimated twice by using data from WVS6 and WVS7, respectively, to assess H4.

Ordinary least squares (OLS) is the primary method used by PROCESS for parameter estimation. However, when the outcome variable (Y) is binary, such as VOTE in this study, PROCESS can model the binary response variable using logistic regression (Hayes, 2012). PROCESS also provides bootstrap confidence intervals as part of its estimation output for performing statistical inference on the indirect effects. Because bootstrapping generates an empirical estimate of the sampling distribution of the indirect effects, no distributional assumption is required. Subsequently, bootstrapping has been advocated as the preferred method for investigating indirect effects (Mackinnon et al., 2004; Preacher & Hayes, 2004; Preacher et al., 2007). It does so through a sampling-with-replacement strategy from the original dataset until many bootstrapped samples are generated. PROCESS constructs a 95% confidence interval around the indirect effect by default based on the bootstrap samples. The confidence interval must not contain zero to support the presence of a mediating relationship. According to Hayes (2013), inferences conducted using the bootstrapping procedure are expected to be more accurate. The probability of committing a Type II error is lower than alternative approaches such as those that rely on a normal distributional assumption.

4 Results

4.1 Mediation Analysis Using WVS7

We generated 5000 bootstrapped samples to estimate bias-corrected 95% confidence intervals for the mediation effect. The control variables used in this study are a standard set of demographics and the state effect (see Table 2). While these control variables are independent variables, which are not part of the research study, their influence cannot be ignored. Results of the mediation analyses are presented in Table 3.

Table 3 Mediation regression analysis using Model 4 of PROCESS

The mediation analyses tested H1, H2 and H3. Results of the bootstrapping analysis show that after accounting for the effects of the control variables (see Table 3, adjusted coefficients), life satisfaction does exert an indirect effect on incumbent voting behaviour. There is a positive relationship between LS and TRUST in the incumbent ruling party (β = 0.1610, 95% CI = 0.1090, 0.2130, p < 0.0001). The results also show that there is a positive relationship between TRUST and VOTE (β = 0.8584, 95% CI = 0.7023, 1.0145, p < 0.0001). Hence, both H1 and H2 are supported. Further, the estimated indirect effect of LS on VOTE (β = 0.13825, 95% CI = 0.0907, 0.2019) supports the claim that trust is a significant mediator between life satisfaction and voting at the 5% significance level as the 95% bootstrapped confidence interval does not contain the value zero. Thus, H3 is supported.

4.2 Understanding the GE14 Election Outcome

Before presenting the mediation results using the WVS6 data, we compare the levels of life satisfaction and trust of respondents from WVS6 and WVS7. Figures 2 and 3 plot the distributions of life satisfaction and trust in 2013 and 2018, respectively. In Fig. 2, it is notable that the proportion of respondents evaluating themselves as completely satisfied (i.e., LS = 10) in 2018 was about half of that in 2013. However, the proportion of respondents rating their LS with a score of 9 was higher in 2018 than 2013. Nevertheless, there was a slight dip in the mean life satisfaction from 7.16 in 2013 to 6.99 in 2018.

Fig. 2

Distribution of life satisfaction

Fig. 3

Distribution of trust

Figure 3 paints a vastly different picture. The proportion of respondents trusting the government (i.e., TRUST = 3 and 4) in 2018 was almost half of that in 2013. On the other hand, the proportion of respondents having absolute distrust in the government (i.e., TRUST = 1) in 2018 was 16%, more than triple that in 2013 (4.8%). The government’s mean level of trust dropped from 2.9 in 2013 to 2.43 in 2018, representing a 16.2% decline.

Table 4 presents the mediation analysis using the WVS6 data, corresponding to Malaysia’s GE13. As before, we generated 5000 bootstrapped samples to estimate bias-corrected 95% confidence intervals for the mediation effects.

Table 4 Mediation regression analysis using Model 4 of PROCESS

There is a positive relationship between LS and TRUST in the incumbent ruling party (β = 0.0766, 95% CI = 0.0248, 0.1284, p = 0.0152). There is also a positive relationship between TRUST and VOTE (β = 0.5411, 95% CI = 0.3829, 0.6994, p < 0.0001). However, the estimated indirect effect of LS on VOTE (β = 0.0414, 95% CI = 0.0130, 0.0771) has a very weak effect size. Moreover, although the 95% bootstrapped confidence interval does not contain zero, the lower limit is almost zero. Thus, we can conclude that TRUST did not mediate the relationship between LS and VOTE in GE13. This result contrasts with GE14, where TRUST mediated the relationship between LS and VOTE, with a large effect size.

In summary, our comparative analysis supports H4. It shows that the ruling coalition’s electoral loss resulted from the mediating effect of trust in the unique context of GE14, manifested by a decline in both LS and TRUST over the preceding five-year period.

5 Discussion

This study examined the role of trust as a mediating variable in the relationship between life satisfaction and incumbent voting based on the contextual premise that situational constraints can have a functional capacity to shape behaviour specific to citizens’ discontent of the incumbent. Furthermore, we compared the mediation results over two waves of data to determine if the mediating effect of trust explained the outcome of the 2018 election. The findings showed that trust mediates the relationship between life satisfaction and voting for the incumbent. We also showed how the mediating effect of trust was unique to the GE14 context, contributing to a six-decade-old government’s downfall.

Beyond the conventional ethnicity and urbanisation determinants of incumbent voting in Malaysia (Ng et al., 2015, 2020), our study sheds light on the nature of the relationship between life satisfaction and incumbent voting, thus contributing to the realm of political science. Using Malaysian data, the empirical results showed that greater life satisfaction leads to greater support for the ruling party. These findings add to a growing collection of studies that examine the effect of life satisfaction on voting outcomes (Flavin & Keane, 2012; Ward, 2019; Weitz-Shapiro & Winters, 2011; Zhong & Chen, 2002). Further, our research illuminates trust in government as a mediator in the relationship between life satisfaction and incumbent voting. Together, these findings suggest that individuals with greater life satisfaction are more likely to trust the government and, therefore, have a higher propensity to vote for the ruling party. By introducing a new mediator, we extend previous research to understand further how life satisfaction contributes to incumbent voting.

The finding that life satisfaction could positively predict trust in government corresponds with recent theory and research indicating that citizens are more likely to trust their governments if they are satisfied with their lives (Esaiasson et al., 2020). Further, correlational analyses revealed that social capital and trust are significantly and positively correlated with happiness (Bjørnskov, 2006; Dunn & Schweitzer, 2005; Helliwell et al., 2020). The finding that distrustful individuals were less likely to vote for the ruling party is consistent with Hirschman’s (1970) framework, which asserts that distrustful citizens would either abstain from voting or vote for the challenger party. Specifically, despite GE14 being on a weekday, Malaysians voted for the challenger party (i.e., PH) amid a high voter turn-out rate of 82.32%.⁶ Subsequently, our results indicate that political distrust was the mechanism through which life dissatisfaction among the voters translated into the ousting of the ruling party.

Based on our study’s findings, we discuss three important implications for Malaysia. First, we see our empirical results play out again when viewed in light of the collapse of the National Alliance (Perikatan Nasional, PN) government following the resignation of former Prime Minister Muhyiddin Yassin.⁷ Given the health and economic crises brought about by the COVID-19 pandemic, people’s life satisfaction has undoubtedly been affected. The White Flag and Black Flag movements are some examples of the growing hardship experienced by the nation.⁸ However, the twin crises were further exacerbated by PN’s poor handling of the challenges and its alleged attempt to “emasculate the legislature” (Said, 2021) through the suspension of parliament after the King of Malaysia assented to the government’s request to declare a national State of Emergency in January 2021. While Muhyiddin justified the need for the Emergency to rein in the pandemic, it was largely seen as a move for Muhyiddin to avoid any vote in parliament that challenged his razor-thin majority (John, 2021).

The PN government was criticised on numerous occasions for its reversals and conflicts in policies for handling the COVID-19 pandemic, causing confusion for the people (Hassan, 2020). It was also perceived to practice double standards, whereby ministers were easily let off the hook for violations of COVID-19 restrictions, while the commoner would be heavily fined for the same offence (Aziz, 2021). The government’s bungled efforts subsequently led to a “growing perception that it is hypocritical, unsympathetic and disconnected from the reality on the ground”, causing public trust in the PN government to be eroded (Loh & Khor, 2021). The large trust deficit incurred by the PN, coupled with its failure to lift the people’s wellbeing from the pandemic, led to the eventual resignation of Muhyiddin Yassin as Prime Minister. In this instance, although an election did not cause the change of government, the political crisis nevertheless revealed the importance of people’s wellbeing and trust in the rise and fall of governments.

Second, the mediating effect of trust suggests that it may be difficult for new political parties to gain political ground in the short term. The Malaysian United Democratic Alliance (MUDA) is one of the latest political parties established in Malaysia. A multi-racial and youth-centric political party co-founded by the former Minister of Youth and Sports, Syed Saddiq, MUDA will contest its maiden elections in the upcoming Johor state polls. Campaigning on a platform of reforms that emphasises values and service regardless of race, local voters have expressed reservations about MUDA due to the unfamiliarity and inexperience of the party (Mohd, 2022). Therefore, although voters may want to vote for a challenger party in line with the Hirschman (1970) framework, their lack of trust in new political parties (or leaders) may impede them from doing so. Following this line of reasoning, any future political party should capitalise on the need to have trusted figures within its political ranks to overcome trust deficits that voters might initially have.

Third, the role of wellbeing in predicting incumbent voting may work as a double-edged sword in Malaysia. While similar studies as ours have concluded that there are political dividends to be gained by improving the wellbeing of the general electorate, the unique context of Malaysia may see such a move as a zero-sum game. This is due to Malaysia's entrenched clientelism climate. BN was able to stay in power for decades partly due to its clientelist practices, which PH continued after its GE14 victory, albeit with different characteristics (Weiss, 2020). This was despite its decry over such practices during its time in opposition. Therefore, patronage opportunities and clientelistic initiatives to improve voters' perceptions of their wellbeing will remain embedded in political strategies as long as voters are acculturated to such norms. However, Weiss (2020) highlighted that Malaysia's clientelism could promote public welfare, especially as the coalitions compete against each other in being providers of distributions and programmatic initiatives. At the same time, Weiss also cautioned that these practices could fortify clientelistic networks that distort incentives and discourage inclusive programmatic policies since people have recourse to patronage. Furthermore, given the recent political developments in Malaysia where no political parties or coalitions have a clear majority or a huge incumbency advantage, political parties may have inclinations to resort to pandering to specific ethnic groups at the expense of others to build their support base. In summary, although our results suggest that improving the wellbeing of the general population should result in favourable chances for the incumbent to retain power, the clientelistic norms in Malaysia may see the selective targeting and uplifting of specific voter groups’ wellbeing instead.

6 Conclusion

As Adserà et al., (2003, p. 445) state, “how well any government functions hinges on how good citizens are at making their politicians accountable for their actions.” Using self-reported measures of life satisfaction, the findings of this study suggest that Malaysian voters did use this piece of information during GE14 to hold the incumbent government accountable for their performance. Notably, the results indicate that the subjective well-being of voters contributed to the electoral fate of the ruling party, both directly and indirectly, through its influence on the government's trustworthiness. We subsequently discussed the implications of these results in Malaysia's context.

Our study concludes with several caveats. First, our empirical analysis was only considered in the context of the federal government's incumbency. We could not apply this at the state level in the context of state governments' incumbencies as these data were not available. Doing the latter would provide a more nuanced test of the hypotheses. Second, there have been concerns about using self-reported measures of life satisfaction as a broad and direct proxy for subjective well-being. An inherent limitation to this single measure is its multi-dimensionality, leading to the question of which domains of life satisfaction drive the link between life satisfaction and incumbent voting. Nevertheless, numerous studies have addressed general criticisms concerning the reliability and validity of subjective indicators. Overall, there seems to be sufficient evidence suggesting that the measure of life satisfaction is reliable and valid (Diener et al., 2013; Noll, 2013; OECD, 2013).

Appendix 1: Regression analysis using Model 4 of PROCESS for GE14 and GE13

Outcome variable: vote	GE14			GE13
	Coeff.	SE	95% CI	Coeff.	SE	95% CI
Intercept	0.6473	0.6801	[− 0.4714, 1.7659]	2.7645***	0.8239	[1.4094, 4.1197]
LS	0.1642	0.0839	[0.0262, 0.3022]	0.1516	0.089	[0.0051, 0.298]
Trust	0.8584***	0.0949	[0.7023, 1.0145]	0.5411***	0.0962	[0.3829, 0.6994]
Controls
Male	− 0.5108***	0.1626	[− 0.7782, − 0.2434]	− 0.3701**	0.1873	[− 0.6782, − 0.0621]
Age	0.0159**	0.0078	[0.003, 0.0287]	0.0031	0.0087	[− 0.0112, 0.0174]
Married	− 0.487**	0.2456	[− 0.891, − 0.0831]	− 0.8577**	0.3516	[− 1.4361, − 0.2794]
Children	0.256	0.251	[− 0.1569, 0.6689]	0.6699	0.3588	[0.0798, 1.26]
Religious	− 0.318	0.2088	[− 0.6615, 0.0256]	− 0.4534**	0.1937	[− 0.7721, − 0.1347]
Malay	0.0281	0.305	[− 0.4736, 0.5297]	− 0.6888	0.3636	[− 1.2869, − 0.0907]
Chinese	− 0.5348	0.3322	[− 1.0813, 0.0117]	− 1.5589***	0.3913	[− 2.2025, − 0.9152]
High edu	− 0.6478***	0.2196	[− 1.009, − 0.2866]	0.4518	0.3195	[− 0.0737, 0.9772]
Low edu	0.0727	0.2319	[− 0.3087, 0.4542]	0.3752	0.2113	[0.0277, 0.7227]
Employee	0.3976	0.2131	[0.0471, 0.7481]	0.4027	0.219	[0.0424, 0.763]
Self-employed	0.4061	0.3035	[− 0.0932, 0.9054]	0.227	0.2644	[− 0.2078, 0.6619]
Income	− 0.0349	0.0423	[− 0.1045, 0.0348]	0.0665	0.049	[− 0.014, 0.1471]
Johor	− 0.659	0.5086	[− 1.4955, 0.1775]	− 0.4817	0.64	[− 1.5345, 0.571]
Kedah	− 1.1576**	0.5591	[− 2.0773, − 0.2379]	− 1.0441	0.6792	[− 2.1613, 0.0731]
Kelantan	− 0.1009	0.5556	[− 1.0148, 0.8129]	− 1.5802**	0.6375	[− 2.6288, − 0.5316]
KL	− 0.3361	0.5555	[− 1.2498, 0.5776]	− 1.0626	0.6902	[− 2.1979, 0.0727]
Melaka	− 0.6484	0.6177	[− 1.6644, 0.3676]	− 0.6287	0.765	[− 1.887, 0.6296]
Pahang	− 0.7475	0.5499	[− 1.6521, 0.157]	− 1.8309***	0.6478	[− 2.8965, − 0.7654]
Penang	− 1.076	0.5602	[− 1.9974, − 0.1546]	− 2.4152***	0.6498	[− 3.484, − 1.3464]
Perak	− 1.053**	0.5129	[− 1.8967, − 0.2094]	− 1.2296	0.64	[− 2.2823, − 0.1769]
Perlis	− 1.3192	0.7684	[− 2.5832, − 0.0552]	− 1.4184	0.9703	[− 3.0144, 0.1775]
Sabah	− 1.0321***	0.5083	[− 1.8682, − 0.1961]	− 0.0879	0.6908	[− 1.2241, 1.0484]
Sarawak	− 0.4632	0.5163	[− 1.3124, 0.3859]	0.6944	0.8317	[− 0.6737, 2.0625]
Selangor	− 1.0917**	0.4865	[− 1.8919, − 0.2916]	− 1.4506**	0.5963	[− 2.4314, − 0.4698]
N. Sembilan	− 1.2497	0.6823	[− 2.372, − 0.1274]	0.1785	0.7886	[− 1.1186, 1.4756]
Indirect effects (Bootstrap)	0.1382	0.0344	[0.0907, 0.2019]	0.0414	0.0196	[0.0130, 0.0771]

1. ***p < 0.01; **p < 0.05
2. The results for the relationship between life satisfaction and incumbent voting, with trust in government as a mediator for GE14 and GE13, are presented in Tables 3 and 4, respectively