Support for bigger government: The principle‐implementation gap and COVID‐19

Abstract

We study the COVID‐19 pandemic's effect on government and market attitudes using within‐subject comparisons of survey responses elicited before and after the onset of the pandemic. We find that participants develop significantly less favorable opinions toward government and markets; and that participants increase support for bigger government significantly and for redistribution, in general, marginally significantly. There is no evidence this leads to an increase in support for specific redistributive policies, nor for government to play a larger role in specific functions. Our results echo the stubbornness of American preferences for redistribution and suggest the presence of a principle‐implementation gap.

Abbreviations

EU

European Union

FWER

Familywise Error Rate

OLS

Ordinary Least Squares

PPP

Paycheck Protection Program

U.S.

United States

WHO

World Health Organization

1. INTRODUCTION

The COVID‐19 pandemic has thrown into stark relief the actual and possible roles of government and markets. In the U.S., the majority of residents have been subject to stay‐at‐home orders, and many schools and businesses to government‐mandated shutdowns. To date, Congress has provided over $4 trillion in emergency aid, including grants to local and state governments, direct stimulus checks to individuals, and expanded unemployment‐insurance benefits; the Federal Reserve created emergency facilities broadening the scope of its role as lender‐of‐last‐resort to include non‐bank firms and financial markets; and a vast public‐private partnership produced an effective vaccine in record time.

Prior global crises have often ushered in substantial changes in the U.S. For example, after the great depression, the Social Security Act of 1935 and the Fair Labor Standards Act of 1938 were passed, establishing old‐age, maternal‐ and child‐welfare benefits, unemployment insurance, and workplace regulations including minimum wages, overtime pay policies, and limits on child labor. After World War Two, the U.S. ended a period of isolationism, helping to found the United Nations and the North Atlantic Treaty Organization, and led the international effort to rebuild Europe and Japan.

For many, the pandemic raised the question of what governments and businesses should do to prevent or ameliorate such crises, and what the polity's priorities will be henceforth. While some have focused on understanding the specific mix of policy responses that have proved most effective (Haug et al., 2020), a rapidly forming literature on individuals' attitudes and perceptions in relation to the COVID‐19 pandemic also seeks to address these questions. For example, using a large‐scale international survey, Fetzer et al. (2020) find that respondents think governments should be doing more in response to the COVID‐19 pandemic. They also find that decisive government response can improve mental health by reducing COVID‐related worry and depression. Many papers have focused on the partisan nature of attitudes and perceptions in relation to the COVID‐19 pandemic in the U.S.: Democrats were significantly more pessimistic than Republicans about their own chance of infection, the severity of the pandemic, and the concomitant economic downturn (Allcott et al., 2020; Barrios & Hochberg, 2020; Fan et al., 2020; Gadarian et al., 2021); Democratic governors were significantly more likely than Republican governors to issue stay‐at‐home orders (Baccini & Brodeur, 2021; Murray & Murray, 2020); political partisanship was a predictor of constituents' propensity to follow such orders (Clinton et al., 2021; Gollwitzer et al., 2020; Grossman et al., 2020); and the greater prevalence of COVID‐19 cases in a county negatively impacted Trump's vote share in the 2020 election (Baccini et al., 2021).

In this paper, we study the COVID‐19 pandemic's effect on attitudes toward government and markets. We contribute to the literature in a number of ways. First, because we survey participants both before and after the onset of the pandemic, we are able to measure how attitudes and support for specific policies changed. To our knowledge, this is the first paper to use a within‐subject design to identify changes between pre‐pandemic and pandemic survey responses in the U.S. Second, all communications with participants avoided mentioning the pandemic in any manner, allowing us to elicit general government and market attitudes without priming participants to think about the pandemic.

Third, the changes we observe among participants, who are college‐aged, are likely to persist over time. Exposure to recessions and pandemics during early adulthood (ages 18–25) has long‐lasting effects on political attitudes and support for government policies: exposure to a recession in early adulthood permanently increases support for redistribution (Giuliano & Spilimbergo, 2014), and exposure to an epidemic erodes confidence in political leaders, governments, and national elections more for those in early adulthood than in any other age group, an effect that lasts at least 2 decades (Aksoy et al., 2020).

Fourth, our results echo prior work revealing a gap between general attitudes and support for specific interventions. Kuziemko et al. (2015) show that while exposure to information about income inequality makes respondents significantly more likely to consider inequality a problem, it does not make them more supportive of specific redistributive policies. This principle‐implementation gap has also been identified in the context of racial equality. Dixon et al. (2017) assert that although individuals are supportive of racial equality in the abstract, support for concrete policies to ameliorate racial inequalities is less widespread. Because our survey includes items about both general attitudes and specific policy preferences, we are able to examine the principle‐implementation gap in relation to the COVID‐19 pandemic.

We find that participants develop significantly less favorable opinions toward government and markets. They also increase their support for bigger government significantly and for redistribution, in general, marginally significantly. In addition, we find evidence of a principle‐implementation gap: desire for a bigger government does not translate to corresponding increases in support for expansions of specific government functions or policies. Lastly, we do not find significant differences in the effect of the pandemic on government and market attitudes by political ideology.

2. LITERATURE REVIEW

A sizable COVID literature has emerged in the economics profession (see Brodeur et al., 2021 for a review). While there is relatively little on the pandemic's effect on attitudes toward the market, a number of studies consider the pandemic's effect on attitudes toward the government. Fetzer et al. (2020) conducted a survey early in the pandemic (late March/early April 2020) with over 100,000 respondents in 58 countries; citizens indicated preferences for strong policy measures from their governments and believed that their governments were not doing enough. This suggests that the decline in favorable attitudes toward government that we identify may be driven by the desire for the government to do more. Fetzer et al. (2020) also find that respondents think that their fellow citizens are not doing enough to contain the pandemic, and that they are not as supportive of strong government response as they are, which may redouble dissatisfaction with the government, as the government may be perceived as the institution most directly responsible for generating prosociality among the citizenry. Bellani et al. (2021) find that greater local exposure to COVID reduces trust in the government for not having been sufficiently responsive or prepared. Daniele et al. (2020) conducted an experiment in a May/June 2020 online survey of over 8000 Europeans, half of whom answered questions immediately after having the concept of the COVID‐19 pandemic primed. Compared to non‐primed respondents, primed respondents exhibited lower interpersonal trust, institutional trust, support for the EU, and support for tax‐financed social welfare spending.

The effect of the pandemic on attitudes toward the government is of course dependent on the country analyzed. Sibley et al. (2020) compare attitudes of a matched sample of New Zealanders before and after the first 18 days of the lockdown and find that the lockdown increased respondents' sense of community, patriotism, satisfaction with the government, and trust in science, politicians and the police, despite an increase in mental distress. Within‐person comparisons to survey responses a year before revealed the same changes as the matched‐sample analysis, in addition to a decrease in satisfaction with business. In Sweden, a within‐person comparison of over 11,000 individuals revealed an increase in institutional and interpersonal trust from the initial (late February to early March 2020) to acute phase (late March to mid‐April 2020) of the pandemic, and also from before the pandemic (December 2018) to the acute phase (Esaiasson et al., 2021). More generally, data from a variety of countries find that the rating of government's response is strongly correlated with trust in government (Lazarus et al., 2020). Important subgroup differences have also been identified: Balasundharam and Dabla‐Norris (2021) analyze data from 11 countries and find that women, those without college education, the unemployed, and those who cannot telework favor their government's pandemic responses the least.

The effect of the pandemic on attitudes toward the government also depends on attitudes toward particular policies. For example, Fazio et al. (2021) find that approval of the Italian government's COVID management was negatively correlated with individuals' desired penalties for lockdown violations. In early April 2020, a survey of over 7000 Europeans found that the policies that respondents most approved of were the fining of quarantine violations, the banning of public gatherings, and border closures; the policies respondents most opposed were the suspension of public transportation, the banning of medical exports, the use of mobile phone data for tracking, and curfews (Sabat et al., 2020). In the U.S., a survey of 3000 Americans in late March 2020 found that controlling for state, county, and urban versus rural location, political orientation was a strong determinant of policy preferences (Gadarian et al., 2021). Compared to Republicans, Democrats were more supportive of social distancing, paid sick leave, and socializing diagnosis‐ and treatment‐costs, and less supportive of import taxes, expansionary macroeconomic policy, international‐travel restrictions, and delaying elections; Republicans and Democrats were statistically indistinguishable in their support for firms, domestic‐travel restrictions, and state governments. In June 2020, Lazarus et al. (2020) surveyed hundreds of people in each of 19 countries about their trust in their government's handling of the pandemic. The item for which the U.S. average exceeded the international average the most was: “The government helped me and my family meet our daily needs during the COVID‐19 epidemic in terms of income, food, and shelter.” The item for which the U.S. average fell short of the international average the most was: “The government cooperated with other countries and international partners such as the World Health Organization (WHO) to fight the COVID‐19 pandemic.”

In the current study, participants' negative attitudes toward government both before and after the onset of the pandemic may be attributable to their cohort. College students in 2020 grew up during a recession, which has been shown to subsequently increase support for left‐wing candidates, which would be consistent with a negative attitude toward a right‐wing administration (Giuliano & Spilimbergo, 2014). Further, in both survey waves, the U.S. president (Donald Trump) was one who was particularly disliked among college students, even conservatives; for example, Republican college students' approval of Trump, while starting relatively high, declined to 25% points below that of the overall Republican population (Harvard Kennedy School Institute of Politics, 2017; Steinmetz, 2018). Another factor may have been the COVID caseload in the areas respondents lived. Using three waves of panel data from 3000 respondents in Germany between May 2020 and May 2021, Bellani et al. (2021) document a negative association between a county's COVID caseload and support for redistribution, which is explained by declining trust in the government. In a European sample, Asaria et al. (2021) found that individuals who experienced COVID‐related health or financial shocks became less inequality averse. In a priming experiment in the U.S., Cappelen et al. (2021) found that while respondents primed to think of the pandemic were more willing to prioritize societal problems over their own, they also became more tolerant of luck‐based inequalities in income and health. In contrast, Rees‐Jones et al. (2022) found that support for safety‐net expansion increased with the impact of COVID at the county level in the U.S., and that while willingness for tax funding did not increase, willingness to increase deficit spending did.

It is important to note that the pandemic experiences of the current study's participants were likely shaped by COVID exposure and government policies—and perceptions of them, often correlated to political orientation—at multiple levels, including federal (both in the US and in the country of origin for foreign students), local (both where students' colleges are located and where they are originally from), and college/university. For example, since all participants were students in either California or New York (the first and fourth states to adopt stay‐at‐home orders) and since these states are among the most progressive, students may be self‐selected on left‐leaning political attitudes relative to others from their home states; such students may then be particularly upset at a less coordinated federal response or by their home states' lack of stay‐at‐home order (McCannon & Hall, 2021).

Turning to market attitudes, the moderate increase in disagreement with pro‐market positions may have arisen due to business's response to the pandemic. For example, Chetty et al. (2020) find that loans to small businesses from the Paycheck Protection Program (PPP) did little to restore employment, with firms that were eligible for PPP no more likely to retain employees than those that were not. PPP also often failed to reach the businesses and areas that most needed it, which could be attributed by the public to mismanagement by the financial institutions that determined which businesses ultimately received aid. The increase in disagreement with pro‐market positions may have also arisen due to the business vulnerabilities exposed by the pandemic (Susskind & Vines, 2020). For example, Mazzucato and Kattel (2020) note that the pandemic exposed the limited reserves that businesses hold, forcing them to cut costs in response to the pandemic; the public may attribute this to myopia or lack of preparedness on the part of businesses.

3. DATA AND METHODOLOGY

We use within‐subject analyses—comparing survey responses before and after the onset of the COVID‐19 pandemic—to examine the impact of the pandemic on government and market attitudes. The “pre‐pandemic” survey was administered from December 2–11, 2019 and the “pandemic” survey from April 29‐May 28, 2020. The survey sample includes students enrolled in introductory economics courses and introductory natural/physical sciences courses in Fall 2019 at Barnard College, Santa Clara University, and Skidmore College. ¹ In total, 480 participants started the pre‐pandemic survey; of these, 329 started the pandemic survey. Our analytical sample includes 269 participants who finished both surveys ² ; 60 participants were excluded as they did not finish one or both of the surveys. ³ A summary of participants' characteristics is found in Table 1.

TABLE 1.

Participant characteristics

Variable	Mean/percent	Std. dev.
Age	19.17	1.07
Gender
Female	67.3
Male	32.7
Race/ethnicity
White	66.7
Black	7.7
Latino	12.6
Asian	27.6
American Indian/Alaska Native	1.1
Pacific Islander	0.4
Other	2.3
Field of study
Economics/business	37.3
Political science	6.0
Intro economics course	61.7
Religious	48.1
Political ideology
Progressive‐conservative scale	4.37	2.39
Progressive (less than 6)	65.1
Moderate (equal to 6)	15.7
Conservative (greater than 6)	19.2
Libertarian‐authoritarian scale	5.29	1.86
Libertarian (less than 6)	45.0
Moderate (equal to 6)	32.9
Authoritarian (greater than 6)	22.1
Employed	50.8
Part‐time	50.0
Full‐time	0.8
High school household annual income
Low income (less than $25,000)	5.8
Middle income (25,000 to less than $200,000)	52.7
High income (greater than or equal to $200,000)	41.6

Note: Participants were directed to choose as many racial/ethnic categories as applied. Political ideology was reported on a scale from 1 (very progressive) to 11 (very conservative) and 1 (libertarian) to 11 (authoritarian).

All three schools are in metropolitan areas (New York City, the Silicon Valley, and the New York Capital Region) that experienced early outbreaks, well in advance of the pandemic survey (Appendix Figure A1 situates the timing of the surveys relative to the total number of U.S. COVID‐19 cases.). Further, only three of the pandemic survey responses were completed after the killing of George Floyd on May 25, 2020, eliminating the possibility that the ensuing protests against police brutality—and government response to those protests—confound the results.

This research was approved by the Institutional Review Board from each school and was registered with the European Economic Association COVID‐19 registry. Participants were offered course credit for completing the pre‐course and pre‐pandemic surveys; it was at the instructors' discretion to determine the course credit offered. Participants had the opportunity to opt out and still receive course credit by emailing the instructor. To incentivize completion of the pandemic survey, participants were entered into a lottery with a five‐percent chance of winning a $30 Amazon gift card.

3.1. The survey

The survey consists of 31 items assessing market attitudes, 14 assessing government attitudes, five assessing support for redistributive policies, and two assessing beliefs regarding 13 specific government functions. ⁴ Appendix Table A2 presents the text of each item and corresponding response scales, as well as a link to the survey.

For each market item, participants are asked to report their agreement with a statement about markets on a Likert scale from “strongly disagree” (=1.0) to “strongly agree” (=5.0) (with increments of 0.1). For example, “In my opinion, market systems lead to quality‐improvements and technical advances in products and services.” All items with a Likert scale include a “don't know” response option.

For seven of the government items, participants are asked to report their agreement with a statement using the same Likert scale as above. For example, “Government should do more to solve problems.” The other seven government items vary in format and response scale. For example, “If you had to choose, would you rather have a smaller government providing fewer services, or a bigger government providing more services?”

For four of the redistributive items, participants are asked to report their support for specific redistributive policies on a ternary scale. For example, “The federal minimum wage is currently $7.25 per hour. Do you think it should be decreased, stay the same or increased?” The remaining redistributive item asks participants to report their agreement with the following statement, using the same Likert scale as above: “It is appropriate for the government to redistribute resources from high‐income individuals to low‐income individuals using policies such as progressive income taxes and social safety net programs.”

For the first of the two items assessing beliefs regarding 13 specific government functions, participants are asked to report whether government does a “very bad,” “somewhat bad,” “somewhat good,” or “very good” job in addressing functions ranging from “strengthening the economy” to “advancing space exploration.” The second item asks participants to report how much of a role, if any, the federal government should play for the same functions: “no role at all,” “a minor role,” or “a major role.”

The order of the items is randomized to control for potential order effects. Specifically, items are organized into two blocks: (I) market‐attitude items and (II) all other items. Block (II) is further divided into three sub‐blocks: (i) all items with a Likert scale, (ii) the two items assessing beliefs regarding specific government functions, and (iii) all remaining items. Participants randomly receive block (I) or (II) first. Within block (II), participants receive sub‐blocks (i), (ii), and (iii) in random order. Further, individual items within each block are randomized.

After completing the government‐ and market‐attitudes survey, participants' sociodemographic characteristics are collected, including gender identity, race/ethnicity, age, college major, political ideology, religious affiliation, employment status, and family income.

3.2. Construction of dependent variables

To reduce dimensionality and, as appropriate, cluster items together that address the same latent construct, we apply exploratory factor analyses, as detailed in Appendix B.

Exploratory factor analysis identifies five distinct latent dimensions of market attitudes: efficient, harmless, fair, unregulated, and anodyne; this five‐factor model closely replicates the structure obtained in Goff and Noblet (2018). Efficient corresponds to the belief that markets efficiently allocate resources. Harmless corresponds to the belief that markets do not cause harm (e.g., that they do not give rise to greed, inequality, or environmental abuse). Unregulated corresponds to the belief that markets should be unregulated. Fair corresponds to the belief that markets provide equal opportunities and just outcomes. Anodyne corresponds to the belief that markets are an acceptable means for exchanging repugnant goods and services (e.g., a market for human organs would not be immoral). Appendix Table B1 lists the market‐attitude factors and their corresponding items.

Participants are assigned a value for each factor equal to the mean of their responses to the items included in the factor; response‐scales of the market‐attitude items are reverse coded as appropriate to make higher values indicative of more pro‐market attitudes. We also calculate a general market‐attitude index, overall, equal to the mean of the participant's five factor scores.

We use a similar approach to reduce dimensionality of the government‐attitude items. From the exploratory factor analysis, we extract two latent factors—beneficent and extent—and two factors that each comprises a single item—content and trust. Beneficent corresponds to the belief that government functions for the benefit of the common good. Extent corresponds to the belief that government should be bigger (e.g., should expand programs or do more than it currently does). Participants are assigned a value for each factor equal to the mean of their responses to the items included in the factor; response‐scales of the government‐attitude items are reverse coded as appropriate to make higher values indicative of more pro‐government attitudes. Lastly, content and trust correspond to contentment with and trust in government, respectively. Appendix Table B2 lists the government‐attitude factors and their corresponding items.

The five items measuring support for redistributive policies are analyzed separately. The first measures support for redistribution in general (redistribute), and the other four measure support for specific redistributive policies (minimum wages, food stamps, tax millionaires, and estate tax). Also analyzed separately are the two items assessing beliefs regarding 13 specific government functions. Lastly, we calculate two additional summary variables, does well and should do, equal to the mean of the participant's responses to the 13 government functions for each item.

Table 2 illustrates the descriptive statistics for the dependent variables. The first and second columns present the mean scores for each dependent variable for the pre‐pandemic and pandemic responses, respectively; and the third and fourth columns present the percent of participants whose scores decrease and increase, respectively, between the pre‐pandemic and pandemic surveys for each dependent variable. Examining market‐attitude factors (see Panel A) one does not observe mean factor scores that lie near the ends of the response scale; they range from 2.34 (between “somewhat disagree” and “neither agree, nor disagree”) to 3.51 (between “neither agree, nor disagree” and “somewhat agree”). Further, we observe a substantial percent of participants reporting both lower (ranging from 40.1 to 44.6%) and higher (ranging from 32.0 to 48.0%) factor scores in the pandemic survey compared to the pre‐pandemic survey. A similar pattern is observed in Panel B for beneficent and extent, in Panel C for redistribute, and in Panel D for does well. All of the remaining dependent variables (trust, content, food stamps, minimum wages, tax millionaires, and estate tax) except for one (should do) are based on a single survey item and have a discrete response scale; thus, the percent of participants whose scores neither decrease nor increase (i.e., stay the same) is greater. Lastly, the means of four dependent variables are near the maximum of the response scale (food stamps, minimum wages, tax millionaires, and should do), which may limit the possibility for upward change between surveys.

TABLE 2.

Descriptive statistics

	Pre‐pandemic	Pandemic	Score decreased (% of sample)	Score increased (% of sample)
Panel A. Market attitudes
Overall	2.94	2.85	40.5	32.0
	(0.503)	(0.559)
Efficient	3.51	3.40	40.1	38.3
	(0.549)	(0.614)
Harmless	2.54	2.42	44.6	41.6
	(0.679)	(0.771)
Fair	2.89	2.80	42.0	47.6
	(0.778)	(0.883)
Unregulated	2.42	2.34	43.1	40.9
	(0.635)	(0.628)
Anodyne	3.26	3.24	40.1	48.0
	(0.661)	(0.711)
Notes: Market attitudes: 1 = strongly disagree with pro‐market position; 5 = strongly agree with pro‐market position.
Panel B. Government attitudes
Beneficent	2.68	2.61	46.1	40.1
	(0.616)	(0.671)
Extent	3.55	3.70	27.9	52.4
	(0.794)	(0.801)
Trust	2.21	2.09	21.6	14.5
	(0.508)	(0.562)
Content	2.14	2.02	19.3	17.1
	(0.620)	(0.643)
Notes: Beneficent: 1 = government does not work for common good; 5 = government does work for common good. Extent: 1 = does not support bigger government; 5 = does support bigger government. Content: 1 = angry with government; 3 = content with government. Trust: 1 = never trusts government to do the right thing; 4 = always trusts government to do the right thing.
Panel C. Redistribution
Redistribute	3.59	3.75	28.6	52.8
	(1.115)	(1.119)
Food stamps	2.70	2.77	7.4	16.4
	(0.541)	(0.467)
Minimum wages	2.72	2.77	5.2	13.4
	(0.510)	(0.438)
Tax millionaires	2.57	2.64	8.2	16.7
	(0.653)	(0.563)
Estate tax	1.78	1.77	17.8	22.3
	(0.703)	(0.672)
Notes: Redistribute: 1 = does not support redistribution; 5 = does support redistribution. Policies: 1 = decrease/do not support; 3 = increase/support.
Panel D. Government functions
Does well	2.38	2.28	48.7	29.0
	(0.484)	(0.498)
Should do	2.69	2.72	15.2	29.7
	(0.275)	(0.254)
Notes: Does well: 1 = does a very bad job; 4 = does a very good job. Should do: 1 = should play a smaller role; 3 = should play a larger role.

Table notes: Standard deviations are in parentheses. For continuous variables, decreased means diff <= −0.10 and increased means diff ≥ 0.10. For discrete variables, decreased means diff < 0 and increased means diff > 0.

4. Results

4.1. Regression analysis

To explore the impact of the COVID‐19 pandemic on government and market attitudes, we conduct within‐subject comparisons of pandemic and pre‐pandemic survey responses. Specifically, we regress the various dependent variables on a pandemic dummy, political ideology, school, college major, gender identity, race/ethnicity, family income, employment status, and a dummy for being religiously affiliated. Results are estimated with OLS and robust standard errors are clustered by participants.

Because we use multiple dependent variables to measure the impact of the pandemic, we use the Romano‐Wolf multiple hypothesis correction (rwolf command in Stata) to control for the familywise error rate (FWER) (Clarke et al., 2020; McKenzie, 2020). ⁵ The Romano‐Wolf correction is more powerful than earlier procedures, such as the Bonferroni correction, as it considers the dependence structure of the test statistics and is less prone to generating false negatives. We report Romano‐Wolf corrected p‐values for the coefficients of interest in all tables.

First, we find that after the onset of the COVID‐19 pandemic participants express less favorable attitudes toward markets. Specifically, the coefficient on overall is negative and statistically significant (b = −0.070, p = 0.017) (see Panel A of Table 3). Further, examining the factors individually, the coefficients on efficient and harmless are negative and statistically significant (b _efficient  = −0.097, p _efficient  = 0.026; b _harmless = −0.111, p _harmless  = 0.026). The estimated reduction in overall, efficient, and harmless represent 2.4% (= −0.070/2.94), 2.8% (= −0.097/3.51), and 4.8% (= −0.111/2.54) reductions in pre‐pandemic mean overall, efficient, and harmless scores, respectively.

TABLE 3.

Regression analyses to examine effects of pandemic on market and government attitudes (OLS)

Panel A. Market attitudes
	Overall	Efficient	Harmless	Fair	Unregulated	Anodyne
	(1)	(2)	(3)	(4)	(5)	(6)
Pandemic	−0.070**	−0.097**	−0.111**	−0.078	−0.067	−0.024
	(0.024)	(0.035)	(0.040)	(0.054)	(0.036)	(0.046)
	p = 0.017	p = 0.026	p = 0.026	p = 0.266	p = 0.179	p = 0.575
Political ideology	0.097***	0.075***	0.132***	0.144***	0.112***	0.028
	(0.016)	(0.019)	(0.020)	(0.024)	(0.020)	(0.021)
	p = 0.001	p = 0.003	p = 0.001	p = 0.001	p = 0.001	p = 0.175
Constant	2.172***	3.068***	1.476***	1.924***	1.383***	2.988***
	(0.175)	(0.209)	(0.229)	(0.269)	(0.233)	(0.248)
Observations	406	401	400	394	403	404
R‐squared	0.343	0.203	0.307	0.294	0.269	0.071

Panel B. Government attitudes, redistribution, and government functions
	Trust	Content	Beneficent	Extent	Redistribute	Does well	Should do
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Pandemic	−0.167***	−0.099*	−0.046	0.151***	0.131*	−0.111***	0.030
	(0.042)	(0.043)	(0.042)	(0.044)	(0.063)	(0.025)	(0.016)
	p = 0.001	p = 0.074	p = 0.258	p = 0.004	p = 0.099	p = 0.001	p = 0.105
Political ideology	0.021	0.098***	0.024	−0.122***	−0.166***	0.079***	−0.045***
	(0.016)	(0.017)	(0.018)	(0.022)	(0.029)	(0.014)	(0.010)
	p = 0.333	p = 0.001	p = 0.333	p = 0.001	p = 0.001	p = 0.001	p = 0.001
Overall market attitudes	0.214**	0.296***	0.261**	−0.667***	−0.746***	0.270***	−0.038
	(0.068)	(0.084)	(0.092)	(0.097)	(0.134)	(0.057)	(0.043)
	p = 0.016	p = 0.006	p = 0.024	p = 0.001	p = 0.001	p = 0.001	p = 0.399
Constant	1.585***	1.055***	2.234***	5.918***	7.240***	1.392***	2.914***
	(0.218)	(0.261)	(0.346)	(0.266)	(0.413)	(0.202)	(0.126)
Observations	406	406	406	406	390	404	404
R‐squared	0.166	0.308	0.146	0.511	0.481	0.366	0.222

Panel C. Support for specific redistributive policies
	Minimum wage	Food stamps	Tax millionaires	Estate tax
	(1)	(2)	(3)	(4)
Pandemic	0.044	0.059	0.064	−0.025
	(0.029)	(0.034)	(0.035)	(0.046)
	p = 0.222	p = 0.212	p = 0.212	p = 0.612
Political ideology	−0.056***	−0.062***	−0.109***	−0.047**
	(0.015)	(0.016)	(0.017)	(0.020)
	p = 0.004	p = 0.004	p = 0.001	p = 0.023
Overall market attitudes	−0.177**	−0.156*	−0.206**	−0.320**
	(0.067)	(0.080)	(0.074)	(0.096)
	p = 0.036	p = 0.076	p = 0.036	p = 0.014
Constant	3.266***	3.195***	3.881***	3.349***
	(0.187)	(0.220)	(0.228)	(0.293)
Observations	406	406	406	406
R‐squared	0.245	0.218	0.404	0.183

Note: Standard errors clustered at participant level are in parentheses. Romano‐Wolf step‐down adjusted p‐values are reported (1000 resamples): ***p < 0.01, **p < 0.05, *p < 0.1. Constant p‐values are not reported and starring uses unadjusted p‐values. Political ideology is measured on a scale from 1 (very progressive) to 11 (very conservative). Additional control variables: school, female gender, race/ethnicity, economics/business student, political science student, high school household income, employed, religiously affiliated.

Second, we find that after the onset of the COVID‐19 pandemic, participants are less trusting of and content with government, and they believe government is less effective at performing various functions (see Panel B of Table 3). Specifically, the coefficient on trust is negative and statistically significant (b = −0.167, p = 0.001) and the coefficient on content is negative and marginally significant (b = −0.099, p = 0.074). Further, the coefficient on does well, which summarizes participants' responses regarding government performance across 13 primary functions, is negative and statistically significant (b = −0.111, p = 0.001). ⁶ The estimated reduction in trust, content, and does well represent 7.6% (= −0.167/2.21), 4.6% (= −0.099/2.14), and 4.7% (= −0.111/2.38) reductions in pre‐pandemic mean trust, content, and does well scores, respectively.

Third, we find increases in support for government to do more, in general, than it currently does. Specifically, the coefficient on extent is positive and statistically significant (b = 0.151, p = 0.004), and the coefficient on redistribute is positive and marginally significant (b = 0.131, p = 0.099) (see Panel B of Table 3). The estimated increase in extent and redistribute represent 4.3% (= 0.151/3.55) and 3.6% (= 0.131/3.59) increases in pre‐pandemic mean extent and redistribute scores, respectively.

Fourth, we find no evidence that the COVID‐19 pandemic changes (i) support for government to play a larger role across 13 primary government functions (should do) (see Panel B of Table 3) ⁷ ; (ii) the belief that government functions for the benefit of the common good (beneficent) (see Panel B of Table 3); and (iii) support for specific redistributive policies (food stamps, minimum wages, tax millionaires, and estate tax) (see Panel C of Table 3).

Fifth, we find that the coefficients on political ideology are often as expected: more conservative participants are (i) significantly less supportive of bigger government (extent) (b = −0.122, p = 0.001) and redistributive policies, in general (redistribute) (b = −0.166, p = 0.001); (ii) significantly more likely to believe government should play a smaller role across 13 functions (should do) (b = −0.045, p = 0.001); and (iii) significantly more supportive of markets (overall, efficient, harmless, fair, and unregulated) (b _overall  = 0.097, p _overall  = 0.001; b _efficient  = 0.075, p _efficient  = 0.003; b _harmless  = 0.132, p _harmless  = 0.001; b _fair  = 0.144, p _fair  = 0.001; b _unregulated  = 0.112, p _unregulated  = 0.001) (see Panel B of Table 3). More conservative participants are significantly more content with government (content) (b = 0.098, p = 0.001) and more likely to believe government is doing a good job across 13 functions (does well) (b = 0.079, p = 0.001). This result may reflect that conservative participants thought of these factors in the context of the conservative federal administration that was in place at the time of the surveys.

Given the heterogeneous impact of the COVID‐19 pandemic on government attitudes as a function of market attitudes, we include overall as a covariate when the dependent variable relates to government attitudes (see Panel B of Table 3). We find that trust, content, beneficent, and does well significantly increase with overall (b _trust  = 0.214, p _trust  = 0.016; b _content  = 0.296, p _content  = 0.006; b _beneficent  = 0.261, p _beneficent  = 0.024; b _{does well}  = 0.270, p _{does well}  = 0.001), and that extent and redistribute significantly decrease with overall (b _extent  = −0.667, p _extent  = 0.001; b _redistribute  = −0.746, p _redistribute  = 0.001). This suggests that pro‐market attitudes are associated with more favorable opinions of government and, not surprisingly, less support for bigger government and redistributive policies, in general.

4.2. Heterogeneity

In sum, we observe the following in response to the COVID‐19 pandemic:

• Result 1. Participants develop less favorable opinions toward markets across all factors.

• Result 2. Participants develop broadly less favorable opinions toward government.

• Result 3. Participants increase their support for bigger government.

• Result 4. Participants do not increase their support for government to play a larger role in specific government functions, including redistributive policies.

These results are based on average effects and may mask important heterogeneities in the survey responses. For example, although attitudes toward markets become less favorable after the onset of the pandemic on average, for a large minority of participants this does not hold: 32.0%, 41.6%, and 38.3% of participants experience an increase in their overall, harmless, and efficient scores, respectively. This leads us to expand our analysis to determine whether heterogeneity in participants' reactions to the COVID‐19 pandemic may help further understand the impact of the COVID‐19 pandemic on government and market attitudes.

Previous research indicates that political ideology mediates responses to the COVID‐19 pandemic. To explore whether political ideology impacts changes in government and market attitudes in response to the pandemic, we re‐estimate the regression with an interaction term: pandemic x progressive. ⁸ We find that the only coefficient significantly different than zero is the coefficient on beneficent (b = −0.257, p = 0.044) (see Panel B of Table 4), indicating that after the pandemic progressives were less likely to believe that the government functions for the benefit of the common good. Otherwise, there is no evidence that the pandemic impacted the government and market attitudes of progressives and non‐progressives differently.

TABLE 4.

Regression analyses with progressive political ideology interaction (OLS)

Panel A. Market attitudes
	Overall	Efficient	Harmless	Fair	Unregulated	Anodyne
	(1)	(2)	(3)	(4)	(5)	(6)
Pandemic	−0.047	−0.015	−0.101	−0.102	−0.010	0.052
	(0.037)	(0.053)	(0.077)	(0.091)	(0.071)	(0.077)
	p = 0.634	p = 0.948	p = 0.634	p = 0.659	p = 0.948	p = 0.852
Progressive	−0.295***	−0.092	−0.486***	−0.442***	−0.386***	−0.033
	(0.069)	(0.086)	(0.095)	(0.112)	(0.098)	(0.104)
	p = 0.002	p = 0.455	p = 0.001	p = 0.002	p = 0.002	p = 0.770
Pandemic × progressive	−0.036	−0.123	−0.011	0.042	−0.084	−0.115
	(0.049)	(0.069)	(0.090)	(0.113)	(0.082)	(0.096)
	p = 0.800	p = 0.326	p = 0.917	p = 0.917	p = 0.687	p = 0.642
Constant	2.746***	3.415***	2.310***	2.779***	2.077***	3.119***
	(0.197)	(0.218)	(0.245)	(0.292)	(0.277)	(0.254)
Observations	406	401	400	394	403	404
R‐squared	0.268	0.153	0.254	0.217	0.229	0.069

Panel B. Government attitudes, redistribution, and government functions
	Trust	Content	Beneficent	Extent	Redistribute	Does well	Should do
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Pandemic	−0.130	−0.072	0.124	0.124	0.303*	−0.109*	0.064
	(0.078)	(0.080)	(0.075)	(0.087)	(0.125)	(0.042)	(0.039)
	p = 0.376	p = 0.376	p = 0.376	p = 0.376	p = 0.096	p = 0.078	p = 0.376
Progressive	−0.041	−0.381***	0.036	0.513***	0.765***	−0.318***	0.219***
	(0.082)	(0.092)	(0.085)	(0.098)	(0.140)	(0.070)	(0.045)
	p = 0.831	p = 0.001	p = 0.831	p = 0.001	p = 0.001	p = 0.001	p = 0.001
Pandemic × progressive	−0.056	−0.039	−0.257**	0.040	−0.260	−0.002	−0.053
	(0.092)	(0.094)	(0.089)	(0.100)	(0.143)	(0.053)	(0.041)
	p = 0.946	p = 0.947	p = 0.044	p = 0.947	p = 0.338	p = 0.968	p = 0.639
Overall market attitudes	0.234***	0.368***	0.280***	−0.747***	−0.876***	0.329***	−0.067
	(0.066)	(0.080)	(0.090)	(0.088)	(0.125)	(0.056)	(0.039)
	p = 0.005	p = 0.001	p = 0.008	p = 0.001	p = 0.001	p = 0.001	p = 0.100
Constant	1.642***	1.503***	2.255***	5.307***	6.412***	1.767***	2.666***
	(0.240)	(0.284)	(0.359)	(0.316)	(0.437)	(0.234)	(0.146)
Observations	406	406	406	406	390	404	404
R‐squared	0.164	0.295	0.154	0.507	0.465	0.350	0.217

Panel C. Support for specific redistributive policies
	Minimum wage	Food stamps	Tax millionaires	Estate tax
	(1)	(2)	(3)	(4)
Pandemic	0.087	0.087	0.145	0.014
	(0.058)	(0.078)	(0.085)	(0.071)
	p = 0.335	p = 0.443	p = 0.268	p = 0.872
Progressive	0.300***	0.341***	0.513***	0.189*
	(0.087)	(0.084)	(0.096)	(0.108)
	p = 0.001	p = 0.001	p = 0.001	p = 0.097
Pandemic × progressive	−0.065	−0.042	−0.123	−0.059
	(0.066)	(0.084)	(0.090)	(0.093)
	p = 0.686	p = 0.783	p = 0.505	p = 0.783
Overall market attitudes	−0.208***	−0.183**	−0.285***	−0.363***
	(0.062)	(0.076)	(0.072)	(0.094)
	p = 0.009	p = 0.043	p = 0.002	p = 0.003
Constant	2.926***	2.793***	3.320***	3.155***
	(0.220)	(0.241)	(0.273)	(0.332)
Observations	406	406	406	406
R‐squared	0.253	0.237	0.390	0.177

Note: Standard errors clustered at participant level are in parentheses. Romano‐Wolf step‐down adjusted p‐values are reported (1000 resamples): ***p < 0.01, **p < 0.05, *p < 0.1. Constant p‐values are not reported and starring uses unadjusted p‐values. Additional control variables: school, female gender, race/ethnicity, economics/business student, political science student, high school household income, employed, religiously affiliated.

We next examine if the same participants who developed less favorable opinions of markets (Result 1)—and specifically came to feel that markets are more harmful or less efficient—experienced a corresponding increase in support for government intervention in markets (Result 3). To examine this question, we first divide the sample into two subgroups based on an adjusted overall that omits unregulated—which is used as the dependent variable—to create overall _adj : those who became less favorable toward markets (difference in overall _adj  ≤ −0.10) and those who became more favorable toward markets (difference in overall _adj ≥ 0.10). We find increased support for government intervention in the market for those who became less supportive of markets overall (b = −0.128, p = 0.012) and no significant change for those who became more supportive of markets overall (see Panel A of Table 5). Performing an analogous analysis for harmless, we find increased support for government intervention in the market for those who came to rate markets as more harmful (b = −0.199, p = 0.001) and no significant change for those who came to rate markets as less harmful (see Panel A of Table 5). Lastly, performing an analogous analysis for efficient, we do not find any significant results.

TABLE 5.

Regression analyses with heterogeneity by changes in attitudes (OLS)

Panel A. Support for unregulated markets, by changes in other market attitudes
	Support for unregulated markets		Support for unregulated markets		Support for unregulated markets
	Less supportive of markets overall	More supportive of markets overall	Rates markets as more harmful	Rates markets as less harmful	Rates markets as less efficient	Rates markets as more efficient
	(1)	(2)	(3)	(4)	(5)	(6)
Pandemic	−0.128**	0.046	−0.199***	0.082	−0.014	−0.128
	(0.057)	(0.061)	(0.052)	(0.056)	(0.059)	(0.068)
	p = 0.012	p = 0.452	p = 0.001	p = 0.213	p = 0.789	p = 0.106
Political ideology	0.105***	0.121***	0.115***	0.101**	0.087***	0.088**
	(0.026)	(0.032)	(0.023)	(0.037)	(0.029)	(0.032)
	p = 0.001	p = 0.001	p = 0.001	p = 0.019	p = 0.002	p = 0.019
Constant	1.181***	1.035**	1.649***	1.250***	1.389***	1.676***
	(0.230)	(0.427)	(0.352)	(0.406)	(0.286)	(0.464)
Observations	185	125	186	163	175	142
R‐squared	0.394	0.331	0.366	0.265	0.307	0.258

Panel B. Support for bigger government, by changes in government attitudes
	Support for bigger gov't		Support for bigger gov't		Support for bigger gov't
	Less trusting of government	More trusting of government	Less content with government	More content with government	Rates government as less beneficent	Rates government as more beneficent
	(1)	(2)	(3)	(4)	(5)	(6)
Pandemic	0.267**	0.259	0.192*	0.129	0.192**	0.174
	(0.108)	(0.225)	(0.130)	(0.126)	(0.069)	(0.066)
	p = 0.020	p = 0.750	p = 0.062	p = 0.750	p = 0.014	p = 0.350
Political ideology	−0.114**	−0.082	−0.110**	−0.006	−0.090**	−0.132
	(0.037)	(0.031)	(0.041)	(0.032)	(0.036)	(0.031)
	p = 0.011	p = 0.350	p = 0.020	p = 0.858	p = 0.020	p = 0.157
Overall market attitudes	−0.942***	−0.799	−0.901***	−0.278	−0.919***	−0.502
	(0.211)	(0.681)	(0.163)	(0.237)	(0.157)	(0.139)
	p = 0.001	p = 0.750	p = 0.001	p = 0.750	p = 0.001	p = 0.212
Constant	7.362***	4.747*	6.293***	6.609***	6.642***	4.697***
	(0.762)	(2.439)	(0.772)	(0.329)	(0.424)	(0.725)
Observations	98	32	78	46	198	154
R‐squared	0.575	0.754	0.637	0.828	0.555	0.528

Panel C. Support for government actions, by changes in support for bigger government
	Support for unregulated markets		Support for redistribution		Support for Gov't taking on specific roles
	Decreased support for bigger gov't	Increased support for bigger gov't	Decreased support for bigger gov't	Increased support for bigger gov't	Decreased support for bigger gov't	Increased support for bigger gov't
	(1)	(2)	(3)	(4)	(5)	(6)
Pandemic	0.074	−0.165***	−0.100	0.244**	0.026	0.029
	(0.055)	(0.055)	(0.124)	(0.097)	(0.027)	(0.026)
	p = 0.392	p = 0.005	p = 0.531	p = 0.014	p = 0.531	p = 0.252
Political ideology	0.039	0.047**	−0.175	−0.152**	−0.040	−0.051***
	(0.043)	(0.020)	(0.074)	(0.048)	(0.020)	(0.014)
	p = 0.450	p = 0.038	p = 0.202	p = 0.016	p = 0.202	p = 0.005
Overall market attitudes	0.694**	0.678***	−0.773**	−0.893**	−0.005	−0.047
	(0.201)	(0.112)	(0.215)	(0.276)	(0.100)	(0.072)
	p = 0.037	p = 0.001	p = 0.037	p = 0.023	p = 0.958	p = 0.565
Constant	0.020	0.022	7.535***	7.876***	2.742***	2.743***
	(0.581)	(0.396)	(0.785)	(0.912)	(0.294)	(0.322)
Observations	121	203	115	197	122	203
R‐squared	0.555	0.453	0.585	0.479	0.190	0.288

Panel D. Support for specific redistributive policies, by changes in general support for redistribution
	Support for food stamps		Support for minimum wage		Support for estate tax		Support for taxing millionaires
	Decreased support for redistribution	Increased support for redistribution	Decreased support for redistribution	Increased support for redistribution	Decreased support for redistribution	Increased support for redistribution	Decreased support for redistribution	Increased support for redistribution
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Pandemic	0.016	0.091	0.031	0.081	−0.234***	0.020	−0.016	0.111
	(0.061)	(0.058)	(0.047)	(0.051)	(0.085)	(0.064)	(0.061)	(0.056)
	p = 0.953	p = 0.263	p = 0.865	p = 0.263	p = 0.009	p = 0.730	p = 0.953	p = 0.136
Political ideology	−0.054*	−0.059	−0.087*	−0.043	0.003	−0.026	−0.073*	−0.138***
	(0.022)	(0.027)	(0.029)	(0.021)	(0.048)	(0.028)	(0.023)	(0.028)
	p = 0.084	p = 0.131	p = 0.065	p = 0.131	p = 0.954	p = 0.354	p = 0.065	p = 0.001
Overall market attitudes	−0.214	−0.227	−0.111	−0.329*	−0.304	−0.275	−0.299*	−0.166
	(0.141)	(0.147)	(0.107)	(0.123)	(0.190)	(0.146)	(0.088)	(0.142)
	p = 0.489	p = 0.267	p = 0.489	p = 0.069	p = 0.489	p = 0.225	p = 0.059	p = 0.290
Constant	3.320***	3.455***	3.553***	3.312***	3.325***	3.151***	3.844***	3.966***
	(0.441)	(0.533)	(0.357)	(0.477)	(0.678)	(0.524)	(0.415)	(0.449)
Observations	128	198	128	198	128	198	128	198
R‐squared	0.275	0.220	0.280	0.292	0.270	0.128	0.527	0.391

Note: Standard errors clustered at participant level are in parentheses. Romano‐Wolf step‐down adjusted p‐values are reported (1000 resamples): ***p < 0.01, **p < 0.05, *p < 0.1. Constant p‐values are not reported and starring uses unadjusted p‐values. Political ideology is measured on a scale from 1 (very progressive) to 11 (very conservative). Additional control variables: school, female gender, race/ethnicity, economics/business student, political science student, high school household income, employed, religiously affiliated.

We also examine whether the same participants who became broadly more critical of government (Result 2) wanted a corresponding decrease in government (Result 4). To examine this relationship, we divide the sample into two subgroups based on each of trust, content, and beneficent: those who became less trusting of (content with) government (difference in trust (content) ≤ −1) and those who became more (difference in trust (content) ≥ 1), and those who come to believe less strongly that government functions for the benefit of the common good (difference in beneficent ≤ −0.10) and those who come to believe it more strongly (difference in beneficent ≥ 0.10). We find that support for bigger government (extent) increased significantly for those who became less trusting of government (b = 0.267, p = 0.020) or came to believe less strongly that government functions for the benefit of the common good (b = 0.192, p = 0.014), and marginally significantly for those who became less content with government (b = 0.192, p = 0.062) (see Panel B of Table 5).

Lastly, we examine whether the same participants who want a bigger government (Result 3) do not want government to play a larger role in carrying out specific functions (Result 4). To examine this we first divide the sample into two subgroups based on extent: those who decrease their support for bigger government in general (difference in extent ≤ −0.10) and those who increase it (difference in extent ≥ 0.10). We find that those whose support for bigger government in general increased experienced a significant decrease in unregulated (b = −0.165, p = 0.005) and a significant increase in redistribute (b = 0.244, p = 0.014), indicating that participants who became more supportive of big government after than before the pandemic were also more supportive of government intervention in the market and redistributive policies after than before the pandemic (see Panel C of Table 5).

To further explore this relationship, we also divide the sample into two subgroups based on redistribute: those who became less in favor of redistribution (difference in redistribute ≤ −0.10) and those who become more in favor (difference in redistribute ≥ 0.10). We find that those in the former group became significantly less supportive of the estate tax (b = −0.234, p = 0.009) (See Panel D of Table 5). We find no other significant changes for those in the former or latter group.

5. DISCUSSION

Comparing pandemic and pre‐pandemic survey responses, we demonstrate that participants, on average, develop significantly less favorable opinions toward government and markets. We also find that participants increase their support for bigger government significantly and for redistribution, in general, marginally significantly. We do not find evidence of increased support for specific redistributive policies, nor for government to play a larger role in specific functions. These results are obtained controlling for covariates and adjusting for multiple hypothesis testing.

Heterogeneity analyses reveal that those whose support for bigger government increased—and those who developed less favorable opinions of markets—came to more strongly support government intervention in markets. Increased support for bigger government did not translate into increased support for specific functions or policies, nor did increased support for redistribution, in general, translate into increased support for specific redistributive policies. Rather, our results echo the stubbornness of American preferences for redistribution and the existence of principle‐implementation gaps (Dixon et al., 2017; Kuziemko et al., 2015). We find evidence that a principle‐implementation gap exists between general support for bigger government and support for government to play a larger role across a range of specific functions.

One potential explanation for the principle‐implementation gap is suggested by the importance of trust on support for redistribution, which has been identified both in general and in the context of the COVID‐19 pandemic. Kuziemko et al. (2015) find that trust mediates support for redistribution in the U.S.—with decreased trust associated with decreased support for redistribution—and Bellani et al. (2021) find that declining trust in government explains a negative correlation between COVID caseloads and support for redistribution at the county level in Germany. Despite large decreases in trust, we do not witness a decrease in general support for redistribution. Rather, we see a statistically insignificant increase in support, coupled with little to no change (and in one context, a significant decrease) in support for specific redistributive policies. We also see that those who become less trusting of government desire a bigger government, but this does not translate into support for government playing a larger role in carrying out specific functions. These results indicate that a decrease in trust is an unlikely explanation for the gap we observe. One potential explanation for the gap is that support for bigger government assumes an idealized government, whereas support for specific functions and policies are grounded in their existing forms. Another is that respondents' ostensible support for redistribution is really support for increased deficit spending. Rees‐Jones et al. (2022) document a COVID‐related increase in support for universal healthcare and unemployment insurance in the U.S.; while there is no corresponding increased support for raising tax revenues, there is increased support for deficit spending.

We recognize scaling issues may contribute to the observed principle‐implementation gap. For example, support for three of the four specific redistributive policies is quite high in the pre‐pandemic survey (roughly 2.7 on scales with a maximum of 3.0), so there may be less scope for an increase than there is for support for bigger government (with a pre‐pandemic mean of roughly 3.5 on a scale with a maximum of 5.0). The same applies to the set of items regarding whether government should play a larger role in 13 specific functions. However, given that support for the estate tax was low in both the pre‐pandemic and pandemic surveys, scaling seems unlikely to be the sole explanation. In addition, we do see some movement in support for food stamps—though not statistically significant—despite mean support of 2.7 (out of 3.0) in the pre‐pandemic survey. Lastly, these potential top‐coding issues cannot explain the general lack of significant decreases in support for specific government functions or policies among those who became less supportive of bigger government. In all, we do not find convincing evidence that scaling issues explain away the principle‐implementation gap.

Strengthening our confidence that the COVID‐19 pandemic is responsible for the observed attitudinal changes, we note the following. First, we find no significant differences between pre‐pandemic and pre‐course survey responses. (Recall that the pre‐course survey was implemented in early fall 2019 as part of a separate study.) This demonstrates the test‐retest reliability of our survey instrument under non‐pandemic conditions and suggests that our results are likely driven by the onset of the pandemic. Second, in assessing the 13 government functions, we find no differences for non‐pandemic related functions. For example, participants do not change their attitudes about whether government does a good job at advancing space exploration, but, as one would expect, they do experience changes in their attitudes about whether government does a good job addressing access to health care.

Although we acknowledge that our sample is not representative, questions about the appropriate roles of government and markets in addressing crises loom particularly large for young people, who are often hardest hit in the long‐run by economic downturns (Hoynes et al., 2012; Verick, 2011), whose confidence in political institutions is most diminished by exposure to epidemics (Aksoy et al., 2020), and who are most concerned about the health risks associated with COVID‐19 (Bordalo et al., 2020).

For many Americans in early adulthood, November 2020 marked the first presidential election in which they could vote, making the immediate and long‐run political consequences of how their attitudes were affected by the pandemic particularly important. This prompts questions about whether the attitudinal changes we observe might be of a scale and scope substantial enough to be decisive in future real‐world contexts. Prior crises appear to have been substantial enough to spur sweeping changes to policies and programs that broke with the past and formed a new consensus. Whether the changes wrought by the COVID‐19 pandemic are substantial enough to have a similar impact remains to be seen.

Supporting information

Supplementary Material

Supplementary Material

Goff, S.H. , Ifcher, J. , Zarghamee, H. , Reents, A. & Wade, P. (2022) Support for bigger government: the principle‐implementation gap and COVID‐19. Contemporary Economic Policy, 1–19. Available from: 10.1111/coep.12593