1 Introduction

In a context of scarcer budgetary funds, special attention is given to the more efficient use of public resources, with better government spending performance and efficiency being preferred by policymakers and taxpayers (see, Afonso et al., 2021a, 2021b). At the same time, a more efficient use of public resources and consequently better government performance, is also (positively) internalized by financial markets (see Afonso et al., 2022). We conjecture that such general efficiency-enhancing policy and approach to government´s assets (physical and human) can generate a higher degree of confidence and trust in the state.

Trust in the government has been identified as one of the most important foundations upon which the legitimacy and sustainability of political systems are built (Fukuyama, 1995). The trust citizens place in their government reflects their confidence in the government's actions. It is a function of the congruence between citizens’ preferences—their interpretation of what is right and fair and what is unfair—and the perceived actual functioning of government (Bouckaert & Walle, 2003). Public trust helps governments manage and administer a country on a daily basis in a way that reinforces the democratic institutions.Footnote 1 However, trust in the government has decreased not only in the US but also in several European countries (Intawan & Nicholson, 2018; Pérez-Morote, et al., 2020). Hence, the key question in this paper is whether we can empirically provide strong evidence on the relationship between government´s trust and public sector efficiency.

The relevance of public sector efficiency has been addressed by a growing literature. Several authors have identified substantial public spending efficiency differences between countries and scope for spending savings. Most public spending efficiency related studies report that there is room for improvement in terms of government spending efficiency, and this typically implies that more public services could be provided with the same public resources, or conversely, the same level of public resources might be provided with fewer public resources. For OECD and EU countries see, notably the evidence reported by Gupta and Verhoeven (2001), Afonso et al. (2005), Adam et al. (2011), Dutu and Sicari (2016), Afonso and Kazemi (2017), Antonelli and de Bonis (2019), and Afonso et al. (2023). Regarding Emerging Markets see, for instance, Afonso et al. (2010), Herrera and Ouedrago (2018), and for Latin American and Caribbean countries see Afonso et al. (2013). To explain these cross-country efficiency differences, studies have examined, in a two-step analysis, the so-called discretionary factors such as: population size, education, income level, quality of the institutions (property right security and corruption) and quality of the country’s governance level, size of the government, political orientation, voter participation, and civil service competence (Afonso et al., 2005; Hauner and Kyobe, 2008; Antonelli & de Bonis, 2019).

Regarding the literature on the level of trust that citizens place in their governments, we can infer that this will depend on the credibility of the government's commitment to the quality of public policies in relation to the amount of spending. For instance, Alesina and Warcziarg (2000) argue that a more pronounced polarisation of voter preferences in advanced economies and the low quality of government policy, which favour particular groups and less the median voter, both reduce trust. Moreover, unproductive government spending reduces public trust in the State, which might become more damaging for large and ineffective governments (Garen & Clark, 2015). Besley et al. (2010) mentioned that governments somehow associated with rent-seeking and lobbying activities contributed to a lower level of public trust. Hence, one can observe unproductive public spending and lower trust of voters in the government. This may be consistent with the decrease of citizens´ trust in government over the years (Intawan & Nicholson, 2018). On the other hand, Pérez-Morote et al. (2020) mentioned that economic events, corruption, or the disclosure of classified information tended to decrease the trust in government. On the same vein, Belabed and Hake (2018) reported that corruption and weak rule of law undermined trust in European governments. In addition, Foster and Frieden (2017) found via survey responses that economic factors at individual and national levels contributed to the trust in the State over the years. Finally, Rodrigues (2021), for a panel of developed and developing countries, reports adverse effects of inefficient public spending on public trust.

Moreover, one might consider the assumption that often politicians know what the ‘right’ policy is, but have private incentives to do something else. Still, it is less clear that politicians do indeed know about the “right” policies. Indeed, for instance, ideological views or lack of information can firmly support different convictions of what is “right” to do. For instance, there might be some interaction between knowledge held and gathered by policy makers and incentive problems that can skew some decisions.

In addition, people's trust in the government is probably also related to their trust in each other. Probably, one can consider that overall economic and societal prosperity is linked and depends on cooperation between individuals and large groups, which is only feasible if trust is indeed present, notably in institutions such as government. Hence, high-trust societies with high moral beliefs, particularly cultural beliefs, can result in better government performance than low-trust societies. This relation can then be perceived as a cultural question as well (see, notably Rose, 2011, 2018).

In this study, we first compute composite indicators of government public sector performance. Secondly, we calculate so-called input efficiency scores for the period 2006–2019. Third, we empirically assess the relevance of these efficiency scores on proxies of trust in the government in a panel setting of 36 OECD countries. It naturally follows that the idea of efficiency is also linked to some measure of fiscal prudence embedded in spending rationalization and optimization efforts.

We find that the more efficient countries in terms of government spending, in our baseline specification (Model 0), are Australia (2009–2011; 2013; 2019), Chile (2007–2016; 2019); Ireland (2015; 2019), New Zealand (2018), South Korea (2006–2018), and Switzerland (2006–2009; 2014–2016; 2019). Moreover, better spending efficiency is positively associated with citizens’ higher trust in the governments. This result holds using alternative proxies for public sector efficiency, alternative measures for trust, specifications with different control variables and different empirical approaches (instrumental variables). In general, political economy variables and the existence of fiscal rules do not seem to significantly affect our measure of trust.

The remainder of the paper is organized as follows. Section 2 discusses and constructs the indicators and scores of public sector efficiency. Section 3 conducts the empirical panel analysis of trust and efficiency. The last section concludes.

2 Public sector efficiency and data envelopment analysis

To compute the public sector efficiency scores, we use data envelopment analysis (DEA),Footnote 2 which compares each observation with an optimal outcome. For each country i, we consider the following function:

$${Y}_{i}=f\left({X}_{i}\right), i=1,\dots ,36$$
(1)

where \(Y\) is the composite output measure (Public Sector Performance, PSP) and \(X\) is the composite input measure (Public Expenditure, PE), namely government spending-to-GDP ratio. We compute the yearly efficiency scores for 36 OECD member countriesFootnote 3 between 2006 and 2019.

The output composite indicator for Public Sector Performance (PSP), as suggested by Afonso et al., (2005, 2022), includes two main components: opportunity and the traditional Musgravian indicators. The opportunity indicators evaluate the performance of the government in administration, education, health and infrastructure sectors. The Musgravian indicators includes three sub-indicators: distribution, stability and economic performance. Table 1 summarizes the variables used to construct the PSP indicators. PSP is the average between the opportunity and Musgravian indicators. Accordingly, the opportunity and Musgravian indicators result from the average of the measures included in each sub-indicator. To ensure a convenient benchmark, each sub-indicator measure is first normalized by dividing the value of a specific country by the average of that measure for all the countries in the sample.

Table 1 Total public sector performance (PSP) indicator
Full size table

Our input measure, Public Expenditure (PE), is lagged 1 year and expressed as a percentage of GDP in several sectors. More specifically, we consider government consumption, expenditure on education, expenditure on health, public investment, transfers and subsidies and total expenditure. Each area of government expenditure is equally weighted to compute the public expenditure input. Tables 7 and 8 in Appendix A provide additional information on the sources and variable construction. Further explanation on the variable’s construction is provided in Afonso et al. (2022).

We adopt an input orientated approach, to measure the proportional increase in inputs while holding output constant and assume variable-returns to scale (VRS), to account for the fact that countries might not operate at the optimal scale. The efficiency scores are computed through the following linear programming problemFootnote 4:

$$\begin{gathered} \mathop {\min }\limits_{\theta , \lambda } \theta \hfill \\ s.t. - y_{i} + Y\lambda \ge 0 \hfill \\ \theta x_{i} - X\lambda \ge 0 \hfill \\ I1^{{\prime }} \lambda = 1 \hfill \\ \lambda \ge 0 \hfill \\ \end{gathered}$$
(2)

where \({y}_{i}\) is a vector of outputs, \({x}_{i}\) is a vector of inputs, \(\lambda\) is a vector of constants, \(I1^{{\prime }}\) is a vector of ones, \(X\) is the input matrix and \(Y\) is the output matrix. The efficiency scores,\(\theta\), range from 0 to 1, such that countries performing in the frontier score 1. More specifically, if θ < 1, the country is inside the production frontier (i.e., it is inefficient), and if θ = 1, the country is at the frontier (i.e., it is efficient).We performed DEA for different models: baseline model (Model 0) includes only one input (PE as percentage of GDP) and one output (PSP); Model 1 uses two inputs, governments’ normalized spending on opportunity and on “Musgravian” indicators and one output, total PSP scores; and Model 2 assumes one input, governments´ normalized total spending (PE) and two outputs, the opportunity PSP and the “Musgravian” PSP scores. Detailed results are illustrated on Tables 9, 10 and 11 of Appendix B.

Table 2 provides a summary of the DEA results for the period 2009–2019 using input-oriented models. The purpose of an input-oriented assessment is to assess by how much input quantities can be proportionally reduced without changing the output quantities produced. Alternatively, and by computing output-oriented measures, one can assess how much output quantities can be proportionally increased without changing the input quantities used.

Table 2 Summary of DEA input efficiency scores
Full size table

Analyzing our results for the input efficiency scores, we find that the average scores of our baseline model ranged between 0.58 to 0.68, For Model 1, the average scores ranged between 0.63 to 0.71, which means that with the same level of outputs, inputs could decrease between 29 and 37%. Model 2’s input efficiency scores averaged between 0.61 and 0.69. Overall, the countries located in the production possibility frontier, hence the more efficient ones in terms of government spending for Model 0 are: Australia (2009–2011; 2013; 2019), Chile (2007–2016; 2019); Ireland (2015; 2019), New Zealand (2018), South Korea (2006–2018), and Switzerland (2006–2009; 2014–2016; 2019).

3 Trust and public sector efficiency

To estimate the impact of public sector efficiency (\({PSE}_{i,t})\) on trust \(({T}_{i,t}),\) we run the following reduced-form panel regression for the period between 2007 and 2020:

$${T}_{i,t}={\alpha }_{i}+{\delta }_{t}+\beta {PSE}_{i,t-1}+\gamma {{\varvec{X}}}_{i,t-1}^{\prime}+{\varepsilon }_{i,t}$$
(3)

where \({\alpha }_{i}\) are country-fixed effects included to capture unobserved heterogeneity across countries, and time-unvarying factors such as geographical variables which may affect the degree of trust; \({\delta }_{t}\) are time effects to control for global shocks (such as commodity prices or the world´s business cycle); \({\varepsilon }_{i,t}\) is an i.i.d. error term satisfying usual assumptions of zero mean and constant variance.

.Our dependent variable is trust in government (\({T}_{it})\) measured by the share of people who report having confidence in the national government. This indicator was retrieved from the OECD Stats (OECD, 2022) and it reflects the percentage of all survey respondents answering “yes” to the survey question: “In this country, do you have confidence in … national government?”.Footnote 5

The main independent variable is the 1 year-lag input efficiency scores\(({PSE}_{i,t-1})\), as computed in the previous section. We also include a vector of other determinants of trust in government, \(({{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}-1})\), lagged 1 year to reduce potential reverse causality concerns.Footnote 6 This vector includes the following variables: the logarithm of population and the age dependency ratio (as percentage of working-age population) included to control for the size of the social benefits, both variables retrieved from World Bank´s World Development Indicators; the debt-to-GDP ratio to control for the size of government retrieved from the IMF´s World Economic Outlook; a dummy variable equaling one for single-party majority government to control for political cohesion, and dummy variable for the right government to control for the political ideology, both retrieved from the Database of Political Institutions (Cruz et al., 2021) and Comparative Political Dataset, respectively.Footnote 7 According to related literature, left-wing governments prefer larger governments, which might be subjected to more elite capture, consequently less efficient (Blais et al., 1993; Cusack, 1997; Hick and Swank, 1992; Jensen, 2011).Footnote 8

4 Empirical results

We begin our empirical analysis by assessing the standalone (unconditional) link between the input level of government spending efficiency and trust. Columns (1) to (3) present our results for Model 0 (one input and one output), Model 1 (two inputs and one output) and Model 2 (one input and two outputs), respectively.

Results reported in Table 3 show that better spending efficiency is positively associated with citizens’ higher trust in governments. These results hold for alternative output efficiency scores (for Models 0, 1 and 2 in Appendix C, Table 13).Footnote 9 As a next step, we estimate the initial baseline specification augmented with a set of control variables, notably: population, age dependency ratio, the debt-to-GDP ratio, right-wing ideology, and majority. Table 4 reports this new set of results again for alternative input efficient scores (for Models 0, 1 and 2).

Table 3 Unconditional regression on input efficiency scores
Full size table
Table 4 Conditional regression on input efficiency scores
Full size table

We continue to find that better public spending efficiency contributes to strengthening the trust in governments, notably for the input and output efficiency scores variables, except for output efficiency scores in Model 2. Results for the output efficiency scores (for Models 0, 1 and 2) are reported in Appendix C, Table 14. Regarding the control variables, we find that countries with larger population and higher level of government indebtedness are associated with lower government trust across both the input and output efficiency scores. Countries with higher levels of age dependency ratio tend to exhibit higher levels of government trust. Finally, no statistically significant result is found for the political economy variables, namely majority and right ideology.

At this point, it is important to address a relevant concern, the possible endogeneity of the efficiency score variables. We estimated specification (2) using panel fixed effect model, however, there might be a potential bi-directional relationship between the efficiency scores and trust in the government. Public sector efficiency may influence trust scores, but trust scores may also have an impact on public sector performance. For example, the citizens trust scores will affect the way they may opt out of the public services such as in the health sector and get treatment in the private sector or simply purchase private health insurances. This could then transform into a vicious cycle for the lack of need for public investment or increased efficiency in public hospitals and other primary care providers (gatekeepers) as less people are using them. A similar reasoning could apply to the provision of public education services. To account for this issue, we used the lagged efficiency score to explain the current trust score. Furthermore, we employ an instrumental variable (IV) or Two-Stage Least Squares approach. To instrument for the efficiency score variables, we select the government effectiveness index from the World Bank´s Governance Indicators. This measure is likely to be correlated with our measure of public sector efficiency, but presumably not directly related to trust. Table 5 reports the IV estimation results using alternative input efficiency score variables.

Table 5 Endogeneity unconditional and conditional regression on input efficiency scores
Full size table

Input efficiency scores are again positively related to the trust, except for Column (2). These main results are also captured for output efficiency scores (see Table 15 in Appendix C). Note that for an instrument to be valid the following conditions have to be satisfied. First, the instrument needs to be correlated with the endogenous variable. In Appendix C, Table 16, we see that this condition is met, except for input efficiency scores for Model 0. Second, the lagged values of the instrument should not be strongly correlated with the trust score (our dependent variable), otherwise the estimated coefficient would still be biased. To test the relevancy of the instrument, we report the Kleibergen-Paap (2016) Wald F statistics. The results are reported at the bottom of Table 5. The rejection of the Kleibergen-Paap rk LM statistics indicates that the instruments are not redundant and hence, they are valid ones.

To assess the heterogeneity of our results, we split countries considering their share of public administration employment and their type of government (coalition versus non-collation governments). From OECD, we retrieved data on employment from on all activities (in services) as well as employment in public administration and defense. Additionally, we collected coalition data from the Comparative Political data and we created a dummy variable that takes the value of one for governments classified as either “Minimal winning coalition” or “Surplus coalition”, and zero otherwise (for all other types). The results of both analyses are presented in Columns (1)–(4) of Table 17 in Appendix C. We find that the positive relationship between public sector spending efficiency and citizens’ trust in governments only holds in countries with lower share of public employment and with non-coalition governments.

We performed several robustness analyses. Our first robustness exercise considers alternative measures for trust in the government. For that purpose, we retrieved data from Gallup confidence, EuroBarometer trust and World Values Survey confidence. Unfortunately, these datasets have more missing observations compared to our measure of trust in the baseline results. The results for alternative measures for trust are presented in Columns (5)–(6) of Table 17 in Appendix C. We continue to find a positive relationship between public sector spending efficiency and citizens’ trust in governments. Note that the magnitude of the coefficients associated with trust are similar to the baseline results presented in Table 4, nonetheless the standard errors are larger.

Our results are still kept when we restrict our sample to a sub-sample of 22 European countriesFootnote 10 and control if they have complied with or deviated from the rules set out in the Stability and Growth Pact (SGP). This is an important issue as the interaction between rules, fiscal space, counter-cyclical policies and credibility has been subjected to more and more scrutiny in recent times (see. e.g. Kopits, 2001; Nerlich and Reuter, 2016). To avert cross-border impact of a country budgetary decisions or jeopardize the functioning of the Economic and Monetary Union, the SGP encompasses four distinct numerical rules: the deficit rule, the structural budget balance rule, the expenditure rule and the debt rule.Footnote 11 Data on the rules of the SGP was retrieved from Larch and Santacroce (2020), Table 6 presents the results for the restricted sample using fixed effects and instrumental variable approach. We continue to find a positive effect of the input efficiency scores on trust for the unconditional regression (results not reported) using fixed effects and instrumental variable approach. When we include the control variables, the positive effect of input efficiency scores on trust is statistically significant in the fixed effect model specification and for input efficiency under Model 1 for instrumental variables.

Table 6 Fixed effects and endogeneity conditional regression on input efficiency scores
Full size table

Finally, we considered an alternative instrument variable. Instead of using government effectiveness, we considered the regulatory quality variable, retrieved from World Bank’s Governance Indicators. Using this as instrument for our key trust variable yields a positive and significant result but only for unconditional regression.

5 Conclusion

The 2007–08 Global Financial Crisis led to a significant loss of trust in governments. In contrast, the response by governments amidst the COVID-19 pandemic inverted that situation. A context of high-inflation and a situation of war in Europe are eroding trust in the State again. As governments search for a path to economic resilience to avoid a recession, the challenge they face is not only knowing what policies to choose, but also how to implement them. Yet, capacity to implement depends crucially on citizens’ trust and this on the other hand, depends on the ability of governments to efficiently use and allocate public moneys.

This paper empirically assessed the role of public sector efficiency scores in shaping the degree of trust in governments. By means of DEA, we first constructed several proxies of public spending efficiency and then related these, in a reduced form panel setting for a sample of 36 OECD countries over the 2007–2019 period, to a measure of trust. We find that the more efficient countries in terms of government spending are Australia, Chile, Ireland, New Zealand, South Korea, Switzerland.

Moreover, we found that indeed the more efficient a government is in managing its expenditure, the higher the level of trust it will gather from voters and citizens. This has important policy implications as the fiscal space available to conduct counter-cyclical fiscal policy is more and more limited.Footnote 12 Being able to convince the median-voter that the appropriate policies are being designed and implemented at times when tax burdens in OECD countries are at historic heights is the counterpart of benefitting from more trust which has positive externalities across other segments of the economy. In general, political economy variables and the existence of fiscal rules do not seem to significantly affect our measure of trust. Our results hold using alternative proxies for public sector efficiency, alternative measures for trust, specifications with different control variables and different empirical approaches (instrumental variables).

Future work could consider exploring more closely the way fiscal policy discretion versus rules matters in shaping government trust. On the one hand, too much discretion can erode trust if governments mismanage freely; on the other, too many rules can limit the necessary actions from the government to cope with crises and hence reap the needed trust so that policies are effective.