The Impact of the Pandemic on Opinion toward the Role of Government in Australia

Abstract

The 2020–23 COVID-19 pandemic resulted in a level of government-imposed restrictions on personal freedom unknown outside of wartime. How did these restrictions affect citizens’ views about the appropriate role of government? To answer this question, we use a unique longitudinal dataset that matches the public’s views about the role of government in Australia to local variations in restrictions on personal freedom aimed at reducing community infection. We identify the extent to which local level restrictions on personal freedom, which varied widely across Australia, affected the public’s views about government intervention. The results show that there was initial support for intervention but continuing restrictions on personal freedom resulted in less support for government intervention, particularly in areas of government policy that affected relatively small population groups. The findings suggest that restrictions on personal freedom significantly changed the public’s views about the role of government in their lives.

Keywords:

• Pandemic
• personal freedom
• public opinion

The views of the public on the role of government have shown relatively little change over the past four decades. This is despite increasing public support for economic deregulation across most of the advanced democracies over the same period (Hadler et al. 2019) and weak economic growth in the wake of the 2007–08 Global Financial Crisis (van de Walle and Jilke 2014). To the extent that there are significant variations in the public’s views of what governments should achieve on their behalf, it is influenced more by national variations in the types of welfare regimes that exist around the world or by changes in social structure than by macroeconomic conditions (Eder 2017). In short, variations in political cultures and path dependencies matter more than major events in shaping mass attitudes to the role of government (Page and Shapiro 1992).

This view of long-term opinion stability concerning the role of government is challenged by the 2020–22 COVID-19 pandemic. The first responsibility of a government is to protect its citizens from an existential threat, in this case, a pandemic. While rare—the COVID-19 pandemic was the first major global pandemic in a century—these events risk large numbers of deaths, widespread illness, and major economic dislocation if infection across the population is not reduced. However, to achieve this aim the government must introduce restrictions on personal freedom with no parallels outside of wartime (Ebrahim et al. 2020). In 2020–22 the effects of these severe restrictions were major intrusions on the social and economic lives of billions of people, with only a small number of countries escaping relatively unscathed (Nguyen et al. 2021; Onyeaka et al. 2021).

This paper examines how far these government-imposed restrictions on personal freedom resulting from the pandemic changed the public’s views about the role of government. To answer this question, we use a unique dataset collected in Australia between 2018 and 2023 which matches views about the role of government with restrictions on personal freedom imposed by the country’s six states and two territories. The survey interviewed respondents nationwide using a question battery about the role of government at four separate time points—the first before the pandemic in 2018, then in 2021 and 2022 while the pandemic was underway, and finally in 2023 once almost all COVID-specific measures had ceased. Since the dataset is longitudinal, meaning we have information on the same individuals at multiple points in time, we are able to gain a unique insight into how the public revised their views about the role of government in response to changes in personal freedom as the pandemic progressed. Australia is a particularly appropriate case study because of its federal structure and the wide differences in how the states and territories approached the pandemic, coupled with a strong tradition of state intervention and a highly developed welfare state.

We extend the literature on the effects of COVID-related lockdowns on views of government in three ways. First, we analyze public opinion over a longer period than any previous study, stretching from August 2018, well before the onset of the pandemic, to January 2023, when the pandemic had effectively subsided. Second, we show that there were important subnational variations in how citizens responded to the lockdowns, based on differing local policies. In line with previous research, our results show that in the initial stages of the pandemic, the public gave strong support to the government in line with the “rally around the flag” effect (Bækgaard et al. 2020; Reeskens et al. 2021). However, public support for decisive government intervention declined in the local areas most affected by the restrictions. Third, we clarify what aspects of government intervention attracted the least support. We find that the decline in public support was particularly marked in areas of government intervention that affected relatively small proportions of the population, such as those without secure housing or who were unemployed.

The paper proceeds as follows. The next section examines how governments respond, and are expected to respond by their public, to existential crises. The second section traces the course of the pandemic in Australia, places it in a comparative context, and outlines the extent of the personal restrictions that existed across the eight jurisdictions. The third section describes the datasets and the analytic strategy, while the section that follows presents the main findings of the study. Finally, we discuss these findings and put them in the context of personal freedom and government intervention in democratic societies.

The role of government in a crisis

The ongoing responsibility of government in advanced societies is to ensure the welfare and prosperity of their citizens, through prudent economic management, support for an open labor market, and the provision of quality health services and educational opportunities, among other things (Gingrich 2014). These functions address a normative question about what the public believes to be the appropriate level of support for social welfare. Against this ongoing commitment to social welfare—at whatever level—governments must also ensure the safety of their citizens from existential threats, whether physical (such as a threat from an external actor) or health-related (such as an epidemic or pandemic).

A crisis inevitably upturns this settled role for government and the public attitudes that underpin it. The government is put under significant pressure to intervene socially and economically in all areas of citizens’ lives to mitigate the effects of the crisis (Hasenfeld and Rafferty 1989). Before the 2020–22 COVID-19 pandemic, the 2007–08 Global Financial Crisis (GFC), and the resulting austerity across many societies, caused relatively short-term changes in public opinion, with trust in politics recovering to its pre-GFC levels in most countries by the early 2010s (Bartels 2013; Hooghe and Okolikj 2020). However, confidence in the government in the countries that were most affected by the crisis—such as Spain and Greece—took much longer to recover, not least among the young, whose economic opportunities were most harmed (Cameron 2021; Sloam 2014).

In the early stages of the pandemic, many governments around the world took decisive action to stop community infection. This involved restricting the movement and personal freedoms of their citizens, introducing major economic subsidies to avoid large-scale unemployment and economic dislocation, and expanding the health system to cope with the expected surge in cases (Altiparmakis et al. 2021). The speed of infection meant that there was little time for policy-makers to consult with the public about the appropriate policy responses, and for the most part, governments dictated policy without having much insight into how the public might respond to these unprecedented changes (Nguyen et al. 2021). Equally problematic was the fact that the mass public had no accumulated experience or information from similar past experiences with which to guide their response to government actions.

Public opinion initially reacted to these unprecedented restrictions with the familiar “rally around the flag” effect, with confidence and trust in the government increasing dramatically, including in Australia (Goldfinch et al. 2021). However, after an initial surge in support for governments and their actions to combat the pandemic, the “rally around the flag” effect was short-lived in many countries, with views on governments and other institutions returning to previous levels as the pandemic continued unabated (Johansson et al. 2021). This was particularly the case in many of the advanced democracies where governments were praised for taking decisive early action to mitigate the effects of the pandemic but whose measures rapidly appeared to have outlived their usefulness (Bol et al. 2021; Kritzinger et al. 2021; Schraff 2021).

Changes in public opinion toward major policy areas move relatively slowly (Page and Shapiro 1992). However, we do know that individual experiences, assessments of the performance of the political system in response to unexpected events, as well as the framing of these events by the mass media, will cause the public to reevaluate their views (Soroka and Wlezien 2010). In such a reassessment, we might expect that public policies directly impacting the individual—such as economic and social security—will attract the most public attention and support. By contrast, policy areas with more marginal implications on the majority of citizens might be seen to be less salient areas for government intervention in the midst of a crisis. In other words, a crisis might be expected to focus the public’s attention on what aspects of government intervention matter most to them.

The COVID-19 experience in Australia

Australia is a particularly appropriate case study to examine the impact of the pandemic on public opinion generally, and views of the role of the state specifically, for two reasons. First, by virtue of its geographical isolation, Australia was able to close its international borders quickly and effectively, meaning that control of the pandemic was almost exclusively in the hands of the government. In practice, the policy settings decided on by the government determined the course of the pandemic, providing a unique opportunity to gauge how policy shapes opinion. Second, as a federal system of government—federal, state, and territory—all levels of government intervened significantly in almost all aspects of society during the pandemic. There were wide variations between these subnational units in how they responded to the crisis (Murphy and Arban 2021), again making Australia an ideal case study to examine how citizens responded to different restrictions.

The Australian experience of COVID-19 and the associated public health response therefore differed markedly from that of almost every other advanced country. The early closure of the international border in March 2020—several months before anywhere in Europe or North America—was accompanied by closures of many interstate borders. This, coupled with widespread lockdowns of education, business, and public transport utilities, meant that by early August 2020, there had only been around 17,000 confirmed cases of COVID-19 in Australia (646 cases per million). This was far lower on a per capita basis than the US (13,471 cases per million), the UK (cases 4518 per million), and Canada (cases 3011 per million), but on a par with New Zealand (cases 234 cases per million). New Zealand had also introduced extensive border controls early in the pandemic.¹

The economic impact of the pandemic on individuals and the Australian economy was mitigated by a range of significant interventions (Lim et al. 2021). These included a “JobKeeper” scheme providing a wage subsidy and a “JobSeeker” scheme providing a supplement to those seeking work. While the unemployment rate rose to 7.4% by mid-2020, this was less than the estimate of 10% without these schemes (Borland and Charlton 2020). The Reserve Bank of Australia cut its target interest rate to a record low of 0.25% which reduced economic pressures on retail and wholesale borrowers. The net effect of these measures, coupled with a gradual opening of the economy, was a 3.3% rebound in GDP in the September quarter of 2020 (Lim et al. 2021).

The continuation of these public health policies into 2021 meant that infection rates remained low and by August 2021, Australia still had only 34,612 cumulative cases, one of the lowest per capita rates among the advanced democracies. Once first dose vaccination rates rose to 95%, restrictions were progressively eased, and infection rates correspondingly increased. COVID-19 infections reached a cumulative 462,955 cases by January 2022, 4.63 million by April 2022, and 9.47 million by August 2022.

In addition to internationally low infection rates, the level of mortality in Australia was initially much lower than in comparable countries. The low death rate, largely due to the border closures and lockdowns, remained until February 2022 when the opening of the borders and the widespread community infection that followed resulted in a substantial increase. Nevertheless, Australia had far fewer deaths from COVID-19 than countries of comparable size and economic development, such as Canada, and very substantially lower deaths still compared to the US and the UK. The public was constantly made aware of Australia’s favorable pandemic performance relative to its peers through intense media publicity. The government gained widespread credit among the public for having dealt with the initial stages of the pandemic effectively.

A distinctive feature of the Australian response to COVID-19, stemming from the federal system of government, was the use of geographically specific restrictions that differed across the six state and two territory borders. These restrictions were put in place in response to localized outbreaks of the disease (that is, variation in case numbers), political preferences within the state government, and confidence in their ability to eliminate or at least control outbreaks. This meant that from the onset of the pandemic, there were large geographic variations in the severity (stringency) of COVID-19 restrictions at any point in time across states and territories (Fenna et al. 2022).

The local variations in COVID-19 policies are shown in Figure 1, using the Australian subnational version of the Oxford COVID-19 Government Response Tracker (OxCGRT) (Edwards et al. 2022; Hale et al. 2021). The index is a daily measure of the government closure and containment policies intended to reduce the spread of COVID-19.² The index ranges from a minimum of zero (fewest restrictions) to a maximum of 100 (most restrictions). Across the COVID-19 period in Australia, stringency values of around 80 or above indicate significant lockdown conditions, values of between 40 and 60 occurred when moderate restrictions were in place, and the indices converged toward 20 when most restrictions had been eased. The results show that all jurisdictions were impacted by the initial lockdown restrictions from March to May 2020. One state—Victoria—also experienced another spike in restrictions, between July and September 2020. All the jurisdictions, then, had a relatively low stringency value for the second half of 2020 and the first half of 2021, though there were some isolated periods with restrictions, in response to localized surges in cases.

Figure 1. Stringency index by state/territory, 2020–2022. Note. See text for details of construction.

Source: https://csrm.cass.anu.edu.au/research/projects/oxcgrt-australian-subnational-dataset

The first two lockdown periods in Australia occurred before the widespread availability of COVID-19 vaccines. Despite a slow start due to supply problems, by the end of July 2021, almost one-third of Australians had received at least one dose, with vaccination rates much higher among the elderly and those with preexisting conditions. In response to the so-called Delta-wave of infections three jurisdictions—the Australian Capital Territory (ACT), New South Wales (NSW), and Victoria—substantially increased their stringency measures from August to October 2021, with significant and lengthy lockdowns in all three jurisdictions. Thereafter, stringency measures declined and generally stayed low for all jurisdictions.

Over the 2020 to 2022 period, these data allow us to distinguish three patterns in the restrictions on personal freedom that were introduced to reduce community infection. First, Victoria experienced three major lockdowns, two in 2020 and one in 2021. Indeed, Victoria’s state capital, Melbourne, was at the time one of the most locked down cities in the world. Second, NSW and the ACT had two major lockdowns, in 2020 and 2021, respectively. Third, the rest of Australia (Queensland, South Australia, Tasmania, Western Australia, and the Northern Territory) had one major lockdown in 2020 and none thereafter. This was in part a consequence of their geographical isolation from the major south-eastern states. These three very different patterns provide an ideal opportunity to test the impact of restrictions on personal freedom on beliefs about government.

The public’s initial response to the pandemic—through the “rally around the flag” effect—suggests that this may lead to support for a more interventionist role for government. We know that the public’s trust in government around the world surged in this early period (Bækgaard et al. 2020; Johansson et al. 2021; Kritzinger et al. 2021) so it would be reasonable to expect that the public would also come to support a more interventionist role for government in line with greater levels of trust. This would be in response to the perceived need for government to introduce measures to restrict personal freedom to stop community infection, as well as to intervene economically to reduce hardship. This leads to the first hypothesis:

H1: There will be increased support for an interventionist role for government in the initial stages of the pandemic.

Democratic societies are committed to ensuring as much personal freedom as possible for their citizens. Many citizens were clearly prepared to accept the need for physical lockdowns and restrictions on what they could do and where they could go to counter the pandemic. This appeared to be the case even in the absence of an informed debate about the utility of such measures. However, we would expect that because such measures are antithetical to the values of a democratic society, as well as the lack of experience that almost all citizens would have of such measures, more restrictions would lead to less support for an interventionist role for government (Johansson et al. 2021). This leads to the second hypothesis:

H2: The greater the restrictions on personal freedom the less public support there will be for an interventionist role for government.

Government intervention has the potential to affect large areas of society, from the regulation of the economy to the maintenance of social welfare and the physical protection of citizens. This raises the question of whether restrictions on personal freedom affect the public’s views about all areas of government intervention equally or whether the public differentiates between them supporting some and not others. We might expect that areas of policy that are likely to affect the well-being and prosperity of the largest proportion of citizens would attract the most support, while areas affecting small groups would be regarded as less relevant. This leads to the third hypothesis:

H3: Government intervention in policy areas less important to the majority of the population will see the largest declines in public support compared to other policy areas.

We test these three hypotheses in the paper using a unique longitudinal dataset that matches public opinion about beliefs in government with the level of restrictions on personal freedoms within a geographical area. The wide variations in the public’s experience of restrictions across a relatively homogeneous population provide an ideal opportunity to test these hypotheses and to evaluate the impact of the pandemic on the public’s views about the role of government.

Data, measurement, and method

Data

The survey data are derived from the COVID-19 Impact Monitoring Series,³ a 14-wave longitudinal survey nationally representative of the adult population conducted between April 2020 and January 2023, as part of the ANUpoll series of surveys.⁴ Three of the survey waves fielded during the pandemic included the Role of Government module derived from the International Social Survey Program (ISSP). The first survey was conducted in January 2021 (wave 6 of the series; n = 3459 respondents), the second in January 2022 (wave 10; n = 3472 respondents), and the third in January 2023 (wave 10; n = 3472 respondents). Also included in the analyses is an August 2018 ANUpoll which asked the Role of Government module.⁵ There is significant longitudinal overlap across the samples and across the four waves of data used in this paper—in 2018, just before the onset of the pandemic, and in 2021, 2022, and 2023—there is a total of 12,521 observations, collected from 5853 unique individuals.⁶

Measurement

In the 2018 to 2023 surveys, the respondents were asked their opinions about 11 specific roles of government using the ISSP Role of Government module.⁷ The question was: “On the whole, do you think it should or should not be the government’s responsibility to …?” The 11 roles ranged from “providing a job for everyone who wants one” to promoting equality between men and women’.⁸ The order in which the roles were presented to the respondents was randomized. The responses were coded “definitely should be,” “probably should be,” “probably should not be,” and “definitely should not be.” The responses to the 11 items were combined additively to form a single scale and rescored from zero (least support for government having a role) to 10 (most support for government having a role).⁹ The validity of a single index was confirmed by a factor analysis of the 11 measures undertaken separately for each year.

The factor loadings for the four surveys are given in Appendix Table A1, and the eigenvalues for the first factor are all above 3.5, and for the second factor they are all below 0.52. Furthermore, while there is some variation for specific variables through time, the eigenvalues for all the variables are above 0.3, and the ordering of the variables by their eigenvalues is very similar. The empirical support for a single index also accords with other research using the same set of questions which confirms the existence of a single underlying attitude about the role of government (Breznau 2019).

The main analyses are based on the scale using these 11 items. The weighted scale mean for all 11 items is 7.74.¹⁰ A secondary analysis is also undertaken based on each of the 11 items considered separately. We assume that the null hypotheses for the significance tests for the individual items will be less likely to be rejected than for the index, as there is less variation in a four-point scale compared to an additive index. While we do not have strong a priori assumptions about which of the treatments (described below) will be significant for which individual variables (if any) we do not expect that there will be any variables for which the sign of the treatment is in the opposite direction to the summary index and is significant.

Method

Our analytical approach estimates the extent to which the experience of more severe lockdowns had a differential impact on belief in the role of government compared to experiencing less severe lockdowns using difference-in-difference regression models. The impact of three “treatment” variables capturing the impact of the severity of lockdowns on the belief in government scale, as well as views on the individual roles of government, are estimated.

Specifically, the first treatment is individuals living in Victoria after the second Victorian lockdown of mid-2020 (i.e., the January 2021, 2022, and 2023 surveys). Specifically, we test whether experiencing the extended lockdown in mid-2020 had an effect on views on the role of government throughout the remainder of the pandemic and after the pandemic had finished. The effect of this treatment is estimated by comparing changes in responses of people living in Victoria to the questions on the role of governments when asked before the second Victorian lockdown and then afterward, with the differences over the same time periods for people living in other areas of Australia who experienced less severe lockdowns. The second and third treatments capture the impact of the third lockdown, which occurred from August through to October 2021 in three jurisdictions—NSW, Victoria, and the ACT. For NSW and ACT residents, this was their second major lockdown, whereas for Victoria it was their third lockdown.

The timing of the surveys does not allow for as neat a treatment effect, as in January 2022 there were still several restrictions in place in these three jurisdictions, as well as in the rest of Australia (albeit at a lower level than during the lockdown period). For this reason, we use two different treatment measures. The jurisdictions are the same for both (NSW, Victoria, ACT), whereas the surveys as part of treatment two are January 2022 and 2023, compared to January 2023 only for the third treatment. That is, for treatment 2 we compare changes in NSW/Victoria/ACT between the August 2018/January 2021 surveys and the January 2022/January 2023 surveys with changes over the same time period in the rest of the country. For treatment 3, we compare changes between the August 2018/January 2021/January 2022 surveys and the January 2023 survey.

Our base model does not control for any other observable individual characteristics and simply compares the difference in the outcome variable pretreatment with the difference in the outcome variable post-treatment. We measure this assuming repeated cross-sections, not considering the longitudinal nature of the survey. In our final specification, we use panel-data techniques and estimate a model with demographic and socioeconomic controls, as well as an individual-level random effect. The equation takes the form: $$Y_{i,t}=f\left({{\beta }_0+{\beta }_{1}X}_{i,t}+{{\beta }_{2}R}_{i,t}+{\beta }_3{T}_{i,t}+{{\beta }_{4}R}_{i,t}.T_{i,t}+{\mu }_{i,t}\right)$$

In the equation, Y_i,t is the outcome variable for individual i at time t. The first outcome variable modeled is the belief in government scale, which is a continuous variable. A second set of outcome variables is the individual roles of government. There are ordered variables with four categories ranging from “definitely should not be” = 0, “probably should not be” = 1, “probably should be” = 2, and “definitely should be” = 4. T_i,t is the severity of the lockdown treatment variable which takes the value if the individual had experienced more severe lockdown and zero otherwise.

The functional form f() differs depending on the outcome variable in that particular equation. It is either a simple linear transformation if we are analyzing the index, or an ordered-probit transformation if we are estimating the effect of lockdowns on the individual-level roles of government. β₀ is a constant term that captures the outcome for those living outside the treatment regions, before the treatment surveys. µ_I,t is an individual-level random effect, which is assumed to vary across time and across individuals and is assumed to be uncorrelated with the other explanatory variables. β₁ captures the impact of the other observed variables in the model (X_i,t), with the impact assumed to be constant across the treatment and control regions/times. β₂ captures the average differences between the individuals in the treatment regions (R_i = 1) and the control regions (R_i = 0). β₃ captures the average differences for observations during the treatment period (T_i = 1) and the control period (T_i = 0).

The final coefficient in the model (β₄) is our main coefficient of interest; this captures the difference between the treatment and control regions, during the treatment periods. If the coefficient is statistically significantly >0, then we conclude that the lockdown increased people’s support for a stronger role in government. If the coefficient is significantly <0, we conclude that the lockdown reduced support for a stronger role for government. Finally, if the coefficient is not statistically significant, then there is no evidence of a lockdown impact.

Limitations

The data used here was not constructed to answer the research questions posed at the start of the paper. When the baseline data was collected in August 2018, it could not have been anticipated that a global pandemic was around 18 months away and that governments would respond with major public health measures that impacted economic and social interaction. This leads to three limitations in the data. First, the role of government module, while validated during the pre-pandemic period, was not validated during the pandemic. The potential roles of government that were asked about therefore reflect the pre-pandemic world, rather than the role of government during the pandemic or afterwards. Second, there was no data collection between the lockdown that occurred across all of Australia in April/May 2020 and the lockdown in Victoria between July and September 2020. The first treatment effect presented in this paper (Victoria compared to the rest of Australia) is therefore difficult to separate from the effect of the pandemic on all Australians. Third, the sample is not designed for jurisdictional comparisons, but rather for national-level estimates. This means that comparisons between the two largest jurisdictions (New South Wales and Victoria) and other jurisdictions across the country are limited by sample size.

Results

What do Australians think about the appropriate role of government judged over an extended period? Overall, there was relatively little change in opinions over the 1985 to 2016 period during which the role of government module has been fielded in Australia, except for slightly stronger support for the provision of social welfare, albeit from a low base (Hadler et al. 2019:179). The two pivotal events of this period—an economic recession in the early 1990s, and the 2007–08 Global Financial Crisis—had relatively little effect on opinion. Placed in comparative perspective, public opinion in Australia is closer to the United States and the United Kingdom when compared to Germany, Norway, or Israel which have stronger welfare regimes.

Figure 2 shows that there has been consistency in the relative distribution of opinions over the 2018–23 period of the surveys, but absolute declines in support for most roles of government. The roles of government that have the greatest level of support are providing health care for the sick, controlling who enters Australia’s borders, and providing a decent standard of living for the old. The roles of government with the lowest levels of support are providing a decent standard of living for the unemployed, providing industry assistance, and providing a job for everyone who wants one.

Figure 2. Public support for areas of government intervention, 2018–2023. Notes. Estimates are the percent who “definitely” consider the area a government responsibility. The “whiskers” on the bars indicate the 95% confidence intervals for the estimate. See text for full question wording.

Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

Comparing the results from the pre-pandemic survey (August 2018) with the post-pandemic surveys (January 2021, January 2022, and January 2023) shows that the roles of government for which there were the greatest declines in the proportion of the population thinking it should “definitely” be a role are in promoting equality between men and women (from 55 to 34% of respondents thinking it “definitely” should be a role), providing the industry with the help it needs (44 to 28%), and providing a decent standard of living for the old (67 to 55%). These declines in the proportion of the population thinking that these should definitely be a role of government are reflected in the scale mean, which was 7.8 in August 2018 and 7.4 in January 2023. This difference is statistically significant at the 1% level of significance and is equivalent to a drop by 29% of a standard deviation.

There was, then, a significant decline in beliefs about the role of government over the COVID-19 period. Respondents clearly thought that the government should have fewer responsibilities by the end of the pandemic than they did at the start. To what extent did the restrictions on personal freedom play a role in any decline? To provide a preliminary answer to these questions, Figure 3 shows the mean score on the full 11-item belief in government scale for the four time points, between 2018 and 2023.

Figure 3. Beliefs in government scale by jurisdiction grouping, 2018–2023. Note. The estimates are the mean on a zero (least intervention) to 10 (most intervention) scale. The “whiskers” on the bars indicate the 95% confidence intervals for the estimate. Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

The results in Figure 3 show a decline in support for the role of government but with a very different trajectory depending on the jurisdiction in question. Those living in Victoria in the pre-COVID period had a higher value on the scale than those in NSW and the ACT (p-value = 0.027) meaning they were more supportive of a strong role for government. Those outside of those three jurisdictions, on the other hand, had a significantly lower index value than NSW, Victoria, and the ACT (p-value = 0.005), meaning they were less supportive of a strong role of government.

By January 2021, the difference between Victoria and NSW/ACT had largely disappeared, and all three jurisdictions had a similar belief in the role of government. However, there was still a large and significant difference between these three jurisdictions and the rest of Australia (p-value = 0.001). This pattern was very similar in January 2022, with a large difference between the COVID and non-COVID jurisdictions (p-value = 0.002), but no difference between NSW/ACT and Victoria. In January 2023, the final wave of data, there was still no difference between NSW/ACT and Victoria. The differences between these three jurisdictions and the rest of Australia had, however, substantially reduced and were no longer statistically significant (p-value = 0.230).

In sum, those jurisdictions with greater levels of restrictions converged in their view on the role of government with the rest of the country, and Victoria with the most severe restrictions had the greatest convergence. The two end points of the distribution in Figure 3 show the scale of the lockdown effects. The difference in the belief in government for those outside of NSW, ACT, and Victoria, who experienced the least restrictive lockdowns, was −0.321. For those in NSW and the ACT, who experienced more severe lockdowns, the difference was equal to −0.445. The “difference-in difference” is therefore equal to 0.142. For Victoria, which experienced the most severe lockdowns of any jurisdiction, the decline in the scale was equal to −0.634, or a “difference-in-difference” relative to NSW and the ACT of 0.189. The findings indicate that the three types of lockdowns had a significant effect on the public’s views of the role of government, reducing their support for government intervention.

To what extent are these findings replicated when other factors are considered? These estimates are shown in Table 1 in the form of a difference-in-difference regression model.¹¹ By using panel data techniques, we can show that the effects hold when we control for individual-level differences across the states and territories based on observable characteristics, as well as the clustering of standard errors from the longitudinal nature of the data.

Table 1. Treatment effects and beliefs in government, 2018–2023 (regression estimates).

	Treatment 1		Treatment 2		Treatment 3
Explanatory variables	Coeff.	Signif.	Coeff.	Signif.	Coeff.	Signif.
Victoria v rest of Australia	0.257	**
Post Lockdown 2	−0.151	**
Treatment 1	−0.089
NSW/ACT/Vic v rest of Australia			0.212	**	0.191	**
Post lockdown 3			−0.157	**
Treatment 2			−0.091	*
Post-COVID					−0.327	**
Treatment 3					−0.111	**
Female	0.476	**	0.471	**	0.470	**
Aged 18–24 years	0.334	**	0.376	**	0.380	**
Aged 25–34 years	0.172	**	0.197	**	0.192	**
Aged 45–54 years	0.005		0.011		0.012
Aged 55–64 years	−0.115		−0.102		−0.104
Aged 65–74 years	0.000		0.021		0.021
Aged 75 years plus	−0.112		−0.095		−0.095
Indigenous	0.275		0.345	*	0.342	*
Born overseas in a main English-speaking country	0.135	*	0.120	*	0.118	*
Born overseas in a non-English speaking country	0.109		0.116		0.120
Speaks a language other than English at home	0.032		0.028		0.025
Has not completed Year 12 or post-school qualification	−0.011		0.006		0.012
Has a post graduate degree	0.034		0.036		0.038
Has an undergraduate degree	0.037		0.032		0.033
Has a certificate III/IV, diploma or associate degree	0.047		0.030		0.033
Lives in the most disadvantaged areas (1st quintile)	0.128	*	0.115		0.114
Lives in next most disadvantaged areas (2nd quintile)	0.075		0.094		0.094
Lives in next most advantaged areas (4th quintile)	0.000		−0.003		0.000
Lives in the most advantaged areas (5th quintile)	−0.009		0.017		0.020
Constant	7.325	**	7.227	**	7.227	**
Sample size—Observations	11,839		11,436		11,436
Sample size—Individuals	5609		5346		5346

Notes. Random-effects, linear regression model. The base case individual is male; aged 35–44 years; non-Indigenous; born in Australia; does not speak a language other than English at home; has completed Year 12 but does not have a post-graduate degree; and lives in neither an advantaged or a disadvantaged suburb (third quintile).

Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

**p < .01, *p < .05.

For treatment 1, which captures the impact of the July to September 2020 lockdown, there was a difference between Victoria and the other states and territories pre-COVID (coefficient of 0.257, p-value < 0.001), a large and significant decline in belief in the role of government across Australia (−0.151, p-value < 0.001), and a slightly larger decrease for Victoria (interaction term −0.089, p-value = 0.127). This provides strong evidence that the lockdowns that occurred in 2020 made everyone in Australia less supportive of a greater role of government, and some support for the view that the specific lockdowns that occurred in Victoria from July to September 2020 made Victorians even less supportive of a greater role for government.

For treatment 2, which occurred between January 2021 and 2022 for NSW, the ACT, and Victoria, the pretrial differences between regions, national level differences across time, and region and time interaction are all in the expected direction and statistically significant. Specifically, before Australia’s third major lockdown, those who lived in NSW, the ACT, and Victoria had a higher index value and therefore were more supportive of a strong role for government (coefficient of 0.796, p-value < 0.001). However, after Australia’s third lockdown (treatment 3) all respondents decreased their support for a strong role for government (−0.157, p-value < 0.001), but those who lived in NSW, the ACT, and Victoria decreased their support by even more (difference-in-difference interaction effect of −0.091, p-value = 0.020). The treatment 3 tests are also significant, with a slightly larger difference-in-difference treatment effect than treatment 2 (−0.111 compared to −0.091 for treatment 2, and a p-value of 0.007 for treatment 3).

Taken together, the results provide strong evidence that lockdowns and the associated restrictions on personal freedom at the local level reduced support for governments playing a more interventionist role in public policy. We therefore confirm Hypothesis 2.

To what extent did the public’s views about specific roles of government change over the period? Given the government’s intervention across widespread areas, from physical movement to economic support, it is highly likely that the public’s views about some areas of government intervention experienced more change than others. Table 2 answers this question by showing the statistical significance of the various treatment effects on each of the 11 government roles. In all cases, the sign of the coefficient is negative indicating decreased support for the role of government in that area.

Table 2. Difference-in-difference treatment effects by role of government.

Role of government	Treatment 1	Treatment 2	Treatment 3
Provide a job for everyone who wants one
Keep prices under control			*
Provide health care for the sick		*
Provide a decent standard of living for the old		*
Provide industry with the help it needs to grow
Provide a decent standard of living for the unemployed			*
Reduce income differences between the rich and the poor
Give financial help to university students from low-income families
Provide decent housing for those who can’t afford it		**	**
Impose strict laws so industry reduces environmental harm/impact
Promote equality between men and women

Notes. Random-effects, linear regression model.

Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

**p < .01, *p < .05.

Treatment 1 shows the impact of the first Victoria-specific lockdown, and none of the effects are statistically significant. In treatment 2 (the third lockdown, which impacted NSW, Victoria, and the ACT) there are three significant effects, the most important of which is a decline in those seeing the government having a role in providing affordable housing. The two other roles that have significant declines relate to health and aged care, respectively. In treatment 3, for January 2023, providing affordable housing is again the most significant effect, followed by price controls and assisting the unemployed.

These results suggest two findings. First, in the initial stages of the pandemic, there were few changes in specific responsibilities that the public thought the government should undertake. This accords with the “rally around the flag” effect when, in the initial stages of a crisis, opinion surges in support of the government’s efforts to combat the threat, even when it involves significant restrictions on personal freedom. At this stage, the public does not differentiate between the specific government roles. Second, later in the pandemic, the public saw a reduced role for government generally, but particularly in providing affordable housing, a problem that affects a relatively small proportion of the population. This is also reflected in the final stages of the pandemic when looking after the unemployed was not viewed as a government priority, a policy that also affects a small proportion of the population. We therefore confirm Hypothesis 3.

Conclusion

Democratic societies rely on their legitimacy to provide as much personal freedom for their citizens as possible. When that freedom is curtailed, for whatever reason, it casts doubt on the public’s consent to have their lives regulated by the government. The COVID-19 pandemic tested the relationship of democratic consent between citizens and the state to a level unprecedented outside of wartime. While there were substantial increases in government expenditure to reduce economic hardship, which the public would generally support, there were often severe restrictions on daily life which the public might be expected to at least question (Anderson et al. 2021). Other research has suggested that the impact of the pandemic on opinions and values has been negligible (Reeskens et al. 2021).

How did these unprecedented restrictions on personal freedom shape the public’s view of the role of government in a democratic society? This paper has answered this question using unique Australian data that matches the public’s beliefs about government intervention with the wide variations in restrictions on personal freedom the public experienced across the eight subnational jurisdictions of the Australian Commonwealth. We reached three conclusions. First, the initial stages of the pandemic saw the familiar “rally around the flag” effect, with widespread public support for government intervention. Second, as the pandemic dragged on, restrictions on personal freedom caused citizens to seek a reduced role in government. This was notably the case in Victoria where the state capital, Melbourne, experienced “stay at home” orders by the government for an unprecedented 277 days.¹² Third, there were specific areas of government intervention that the public wanted to be reduced, notably in housing affordability and support for the unemployed.

These findings suggest two observations about the public’s longer-term response to these unprecedented restrictions on their personal freedom. First, individual circumstances are more important than macroeconomic performance in shaping the public’s views about the government’s role. This is evident in housing and unemployment being regarded as areas where there should be reduced government involvement since both affect relatively small groups of citizens. Generally, then, macroeconomic conditions matter less than micro factors in shaping the public’s views about the appropriate role of government. After the GFC, for example, the performance of the macroeconomy had relatively little influence on opinions about the role of government (Kenworthy and Lindsay 2011; van de Walle and Jilke 2014). This may be because factors, such as how people live their daily lives have more impact on opinions than how the national economy performs. In the case of the pandemic, due to major government economic intervention, economic hardship was minimized.

A second observation relates to how the public views the role of government in democratic societies. The main division in beliefs about government is usually interpreted as being between welfarist and liberal regimes, with the former supporting more intervention, and the latter less (Hadler et al. 2019). The results presented here suggest that personal freedom also represents an important secondary dimension, insofar as prolonged encroachments on freedom will reduce support for government intervention, particularly in areas that affect relatively small groups of citizens and are removed from the large majority. When subjected to an existential crisis, citizens may value their personal freedom from government interference more than their support for social welfare, particularly when that welfare appears to benefit more marginal groups within society, such as the unemployed.

Acknowledgments

Our thanks to two anonymous referees from this journal for their thoughtful and constructive comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The survey data is publicly available through the Australian Data Archive (doi:10.26193/BTCSY5).

Notes

Additional information

Funding

Data collection for this research was partially funded by the Australian Institute of Health and Welfare.

Notes

1 See https://coronavirus.jhu.edu/map.html.

2 The measure combines school, workplace and public transport closures, cancellation of public events and limitations on the size of gatherings, and restrictions on domestic and international travel. The estimates are coded from government policies in a way that makes them internationally comparable to other countries and within Australia (Hale et al. 2021).

3 See https://csrm.cass.anu.edu.au/research/publications/covid-19.

4 The ANUpoll series of surveys is collected on a probability-based, longitudinal panel. By using probability-based recruiting (predominantly telephone-based) the unknown and unquantifiable biases inherent in opt-in (non-probability) panels are minimised and it is also possible to quantify the uncertainty around the estimates due to sampling error using standard statistical techniques. This is not possible with non-probability surveys.

5 The methodology for the August 2018 ANUpoll is very similar to the more recent surveys, however the sample size is smaller with 2220. The COVID-19 Impact Monitoring series had a range of questions related to COVID-19, mental health, and wellbeing. The August 2018 ANUpoll, on the other hand, had far fewer questions, and no wellbeing questions.

6 On average, each respondent had information across 2.1 waves, with 2336 respondents (39.9%) collected in one wave, 1333 (22.8%) in two waves, 1217 (20.8%) in three waves, and 967 respondents (16.5%) with information across all four waves of data.

7 Full details can be found at https://issp.org/.

8 Two additional Australia-specific items were included, on border control and Aboriginal and Torres Strait Islander Australians, respectively. They are excluded here to ensure comparability with the international surveys.

9 The Cronbach’s alpha for the 11 variables in the scale was 0.83 in the 2018 survey, 0.85 in January 2021, 0.85 in January 2022, and 0.84 in January 2023. Because we are partially interested in changes through time in the belief in the role of government, we use an additive index with equal weight in our analysis, rather than the first estimated factor with weights based on factor loadings.

10 Those who did not give a response to any of the roles are excluded from the analysis.

11 We re-ran the model with a balanced panel (i.e. only those individuals that were observed in all four waves). The direction of the treatment effects were the same, and the magnitude was similar. However, because of the much smaller sample size (only 954 individuals) the p-values for the treatment effects are larger and not statistically significant. We also re-ran the model with the dependent variable being the number of roles that the person definitely thought should be a responsibility of government. We find exactly the same size and significance of the treatment effects—negative and not significant for treatment 1 (p-value = 0.181), negative and significant for treatments 2 and 3 (both with p-values of <0.01. It would not appear that the choice of sample or the choice of index construction is driving the result.

12 https://www.abc.net.au/news/2021-10-25/fact-check-is-melbourne-most-locked-down-city/100560172

Appendix A

Table A1. Loadings and eigenvalues for factor analysis of role of government variables.

Variable	Aug’ 2018	Jan’ 2021	Jan’ 2022	Jan’ 2023
Provide a job for everyone who wants one	0.5345	0.5695	0.5384	0.5279
Keep prices under control	0.5076	0.5161	0.4855	0.4514
Provide health care for the sick	0.5426	0.5691	0.5534	0.4921
Provide a decent standard of living for the old	0.5911	0.6299	0.6278	0.5989
Provide industry with the help it needs to grow	0.3178	0.3745	0.3564	0.3022
Provide a decent standard of living for the unemployed	0.6481	0.6943	0.7074	0.7069
Reduce income differences between the rich and the poor	0.6231	0.6621	0.6602	0.6562
Give financial help to university students from low-income families	0.6258	0.6508	0.6617	0.6491
Provide decent housing for those who can’t afford it	0.6861	0.7417	0.7453	0.7262
Impose strict laws to make industry reduce their environmental harm/impact	0.5094	0.5626	0.5720	0.5308
Promote equality between men and women	0.5633	0.6001	0.6061	0.5943
Eigenvalue 1	3.53566	4.02223	3.97710	3.68854
Eigenvalue 2	0.47825	0.43250	0.43434	0.51125
Eigenvalue 3	0.22041	0.22662	0.19139	0.22941

Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

Appendix B

Table B1. Treatment effects and beliefs in government using balanced panel, April 2018–January 2023 (regression estimates).

	Treatment 1		Treatment 2		Treatment 3
Explanatory variables	Coeff.	Signif.	Coeff.	Signif.	Coeff.	Signif.
Victoria v rest of Australia	0.001
Post Lockdown 2	−0.246	^–
Treatment 1	−0.003
NSW/ACT/Vic v rest of Australia			0.240	*	0.223	*
Post lockdown 3			−0.189	**
Treatment 2			−0.092
Post-COVID					−0.330	**
Treatment 3					−0.103
Female	0.490	**	0.478	**	0.476	**
Aged 18–24 years	0.463	*	0.504	*	0.526	*
Aged 25–34 years	0.248	*	0.270	*	0.277	*
Aged 45–54 years	−0.093		−0.088		−0.091
Aged 55–64 years	−0.205		−0.220		−0.230	*
Aged 65–74 years	−0.023		−0.052		−0.065
Aged 75 years plus	0.032		−0.031		−0.053
Indigenous	−0.115		−0.148		−0.153
Born overseas in a main English-speaking country	0.222		0.234		0.232
Born overseas in a non-English speaking country	0.109		0.128		0.130
Speaks a language other than English at home	−0.172		−0.179		−0.181
Has not completed Year 12 or post-school qualification	−0.102		−0.028		−0.002
Has a post graduate degree	0.162		0.176		0.182	*
Has an undergraduate degree	0.006		0.014		0.021
Has a certificate III/IV, diploma or associate degree	0.007		0.009		0.014
Lives in the most disadvantaged areas (1st quintile)	0.008		0.029		0.028
Lives in next most disadvantaged areas (2nd quintile)	0.146		0.152		0.150
Lives in next most advantaged areas (4th quintile)	0.058		0.074		0.074
Lives in the most advantaged areas (5th quintile)	−0.018		−0.003		−0.003
Constant	7.372	**	7.157	**	7.143	**
Sample size—Observations	3703		3703		3703
Sample size—Individuals	954		954		954

Notes. Random-effects, linear regression model. The base case individual is male; aged 35–44 years; non-Indigenous; born in Australia; does not speak a language other than English at home; has completed Year 12 but does not have a post-graduate degree; and lives in neither an advantaged or a disadvantaged suburb (third quintile).

Sources: ANUpolls (August 2018, January 2021, January 2022, and January 2023).

**p < .01, *p < .05.