ABSTRACT

The Understanding America Study (UAS) and analysis

Data for this analysis come from a nationally representative subsample of the Understanding America Study (UAS),^19,22 a probability-based internet panel of approximately 9,000 adults aged 18 years or older that is representative of the non-institutionalized US population. Panel members are recruited using address-based sampling, and tablets and broadband internet are provided if needed. Surveys are in English or Spanish. From April 1, 2020 to February 16, 2021, the UAS invited a subsample of the online panel to participate in biweekly surveys about the COVID-19 pandemic.¹⁹ Respondents were randomly assigned to an invitation day, and invitations were distributed over the course of 14 days. Since February 17, 2021, this invitation period has been doubled – invitations are now distributed over the course of 28 days. Once respondents are invited into each survey wave, they are given 14 days to respond. Following best survey practices,²³ incentives are provided to encourage respondents to complete each survey at a rate of 20 USD per 30 minutes of survey time (and proportionately less for shorter surveys). To encourage respondents to complete the interview on their assigned day, they receive an additional 1 USD incentive if they complete the survey on the day they receive their invitation. For survey waves fielded prior to February 17, 2021, over 90% of responses were received within the first 2 weeks of fielding; for surveys fielded since February 17, 2021, over 90% of responses are received within the first 4 weeks of fielding. Thus, a two-week sample of surveys collected prior to February 17, 2021 is comparable in size to a four-week sample of surveys collected since February 17, 2021.

We use a graphical format to describe trends in vaccination likelihood since April 2020, presenting 2-week averages that span April 1, 2020, to March 16, 2021. We present our statistical analyses in a table format. To evaluate post-election trends, our statistical analyses focus on two time points: the two-week period of October 28^th-November 10^th, 2020, which occurred around the US Presidential election (6,396 respondents, response rate 78%), and the four-week period of February 17-March 16, 2021, which is comparable in sample size to prior two-week periods and was the most recent survey period available (5,891 respondents, response rate 71%).

In each survey wave, respondents were asked the question: “How likely are you to get vaccinated for coronavirus once a vaccine is available to the public? (very likely, somewhat likely, somewhat unlikely, very unlikely, or unsure). We classified “very likely” or “somewhat likely” responses as likely to vaccinate; we included those who already had a COVID-19 vaccine as likely to vaccinate. We interpreted “very unlikely,” “somewhat unlikely,” and “unsure” responses as indicative of COVID-19 vaccine hesitancy.

We examined how trends in vaccination likelihood varied according to demographics (age group, sex, race and ethnicity, and educational attainment) and political affiliation (Democratic party, Republican party, or other).

To assess reasons for hesitancy about COVID-19 vaccinations, we analyzed five questions on the most recent survey (February 17-March 16, 2021). First, we asked about level of agreement with statements regarding benefits or side effects of the vaccines. To explore deeper concerns about vaccine effectiveness or harms, we asked respondents to estimate the likelihood that a vaccinated individual might still catch coronavirus or the likelihood that the vaccine will cause serious side effects or long-term harm. Finally, we asked two questions about level of trust in the vaccine development process in general and about the governmental approval process for COVID-19 vaccines.

We graphed trends in the likelihood of getting a COVID-19 vaccination for the entire population and for racial/ethnic subgroups (Figure 1). We calculated percentages and margins of error (based on a 95% confidence level) with adjustment for survey sampling weights. We performed a cross-sectional analysis (Table 1) of reported intentions in the most recent February 17-March 16, 2021 survey using a multivariable Poisson regression model with robust standard errors, showing differences between groups as risk ratios with 95% confidence intervals (CIs). To assess trends since the US election, we also analyzed the change in intentions between the October 28-November 10, 2020 survey and the most recent February 17-March 16, 2021 survey. We used linear regression models of changes in intentions with cluster-robust standard errors, with clustering at the respondent level and summarized changes over time using least squares means and 95% CIs. We used significance levels of 0.05 and Holm’s correction to account for multiple hypothesis testing. Finally, we used descriptive statistics to summarize vaccine perceptions among individuals who were hesitant about COVID-19 vaccines (Table 2).

Table 1.

Demographic and individual predictors of likelihood of getting a coronavirus vaccine

	February 17 – March 16, 2021			Change from October 28-November 10, 2020 to February 17 – March 16, 2021
	Percent Likely to Get Vaccine	Adjusted Risk Ratio (95% CI)	P Value	Unadjusted Change in Percent (95% CI)	P Value	Adjusted Change in Percent (95% CI)	P Value
Overall	64.2%			10.4% (8.4%, 12.3%)	<.001	10.4% (8.4%, 12.3%)	<.001
Sex
Female	60.2%	- REF -		11.8% (9.0%, 14.5%)	<.001	12.7% (8.8%, 16.6%)	<.001
Male	68.5%	1.10 (1.04, 1.16)	<.001	9.0% (6.3%, 11.7%)	<.001	9.8% (6.0%, 13.6%)	<.001
Age (y)
18–49	55.9%	- REF -		5.5% (2.7%, 8.3%)	<.001	2.9% (−0.8%, 6.6%)	.119
50–64	68.2%	1.29 (1.20, 1.38)	<.001	15.2% (11.6%, 18.8%)	<.001	12.6% (8.5%, 16.6%)	<.001
65–74	78.5%	1.44 (1.34, 1.54)	<.001	13.8% (9.2%, 18.4%)	<.001	12.2% (6.9%, 17.5%)	<.001
75+	86.4%	1.59 (1.46, 1.72)	<.001	18.7% (12.4%, 24.9%)	<.001	17.4% (10.5%, 24.3%)	<.001
Education
High School or Less	54.3%	- REF -		11.4% (7.6%, 15.1%)	<.001	11.5% (7.0%, 16.0%)	<.001
Some College	59.1%	1.11 (1.03, 1.21)	.008	11.1% (7.4%, 14.8%)	<.001	11.4% (6.9%, 15.9%)	<.001
Bachelor’s or More	78.9%	1.38 (1.29, 1.48)	<.001	8.8% (6.4%, 11.2%)	<.001	10.9% (7.0%, 14.7%)	<.001
Race/Ethnicity
Non-Hispanic White	64.9%	- REF -		9.2% (7.0%, 11.3%)	<.001	12.1% (9.8%, 14.5%)	<.001
Hispanic	61.9%	0.93 (0.85, 1.03)	.175	16.0% (9.4%, 22.6%)	<.001	19.4% (12.7%, 26.2%)	<.001
Non-Hispanic Black	56.5%	0.77 (0.69, 0.86)	<.001	16.6% (10.3%, 22.8%)	<.001	17.1% (10.7%, 23.5%)	<.001
Non-Hispanic Asian	83.8%	1.16 (1.06, 1.27)	.002	6.2% (−0.1%, 12.6%)	.055	9.9% (2.917%, 0%)	.006
Other	60.0%	0.93 (0.79, 1.09)	.371	−6.6% (−15.9%, 2.8%)	.169	−2.3% (−11.8%, 7.2%)	.634
Political Influence
Democratic Party	80.1%	- REF -		15.7% (12.7%, 18.8%)	<.001	15.9% (12.2%, 19.7%)	<.001
Republican Party	53.4%	0.64 (0.60, 0.68)	<.001	7.8% (4.5%, 11.1%)	<.001	9.2% (4.7%, 13.8%)	<.001
Other	54.7%	0.71 (0.66, 0.77)	<.001	5.6% (1.8%, 9.3%)	.003	8.6% (3.9%, 13.3%)	<.001

Open in a separate window

Bolded p-values indicate significance at the 0.05 level (Feb. 17 – March 16 adjusted risk ratios), or significance after Holm-Bonferroni correction (change from Oct. 28-Nov. 10 to Feb. 17 – March 16). Adjustments are for all variables in the table.

Individuals who have already received a COVID-19 vaccine were included as “Likely”.

Open in a separate window

Figure 1.

The figure shows likelihood of getting a coronavirus vaccine, or a self-report of having been vaccinated, for each two-week period from April 1, 2020, to March 16, 2021, except for June 10–23 and July 8–21, 2020 (surveys were not done during those two periods). The margin of error for 95% confidence level ranged from: Hispanic (5.1%-7.3%%), White (1.8%-2.9%), Black (5.2%-8.7%), Asian (4.8%-8.0%), Overall (1.7%-2.5%) per survey period

Percent of surveyed US adults who are likely to get a coronavirus vaccine if available: Overall and by race and ethnicity.

Findings: post-election likelihood of vaccination overall and by subgroups

For all US adults (Figure 1), the likelihood of getting a coronavirus vaccine increased from 54% in October 28-November 10, 2020 to 66% by March 3–16, 2021. As shown in Table 1, disparities still exist in March 2021. Males are more likely than females to state that they are likely to be vaccinated (69% vs. 60%, p < .001). Stated likelihood of vaccination increases by age group, with only 56% of young adults aged 18–49 years stating they are likely to be vaccinated compared to 79% of adults aged 65–74 years and 86% of adults aged 75 years or older. Individuals with some college (59%) or a Bachelor’s degree or higher (79%) are more likely to state they are likely to be vaccinated than those with a high school degree or less education (54%). Non-Hispanic Asian individuals (84%) are more likely than whites (65%) to state they intend to get vaccinated, while Black individuals (57%) are less likely and Hispanic individuals (62%) are similar to whites. Finally, individuals who affiliate with the Democratic party (80%) are far more likely to state they intend to get vaccinated than those who affiliate with the Republican party (53%), or have some other affiliation (55%), p < .0001 for each comparison with Democrats.

Table 1 shows the unadjusted and adjusted change in self-reported likelihood of vaccination between the October 28-November 10, 2020 survey period and the February 17 – March 16, 2021 survey period. The analyses reveal several important findings. First, virtually all subgroups have increased their stated likelihood of vaccination since the US election, with increases ranging from 3 percentage points to 17 percentage points. Second, disparities have widened by age group. While adults aged 50 years or older experienced a 12–17 percentage point increase in their stated likelihood of vaccination, young adults aged 18–49 years experienced only a 3-percentage point increase in their stated likelihood of vaccination. Third, disparities by educational attainment did not widen but persist, with an 11-percentage point increase in vaccination likelihood occurring across all three educational groups. Fourth, disparities by race/ethnicity have narrowed. While all racial/ethnic groups except “other race” have experienced substantial increases in their stated likelihood of vaccination, the increase has been greater among Non-Hispanic Black and Hispanic respondents than among Non-Hispanic White respondents. Finally, disparities by political affiliation have widened. Since the election, individuals affiliated with the Democratic party have experienced a 16-percentage point increase in stated likelihood of vaccination while individuals affiliated with the Republican party have experienced a smaller, 9-percentage point increase in stated likelihood of vaccination.

Table 2 shows perceptions among individuals who are hesitant about COVID-19 vaccines. A very high percentage of vaccine-hesitant individuals reportedly believe that the vaccines have harmful side effects and may even lead to illness or death. A considerable number of vaccine-hesitant individuals did not think that the vaccines are effective or beneficial to society. Most vaccine-hesitant individuals stated that a very high percentage of people who are vaccinated can still catch coronavirus or will have serious side effects or long-term health problems. Finally, a high proportion of vaccine-hesitant individuals did not trust the general vaccine development process to make safe vaccines, and did not trust the governmental approval process to ensure that COVID-19 vaccines are safe for the public.

Conclusions

A high proportion of the US population needs to receive a COVID-19 vaccine in order to end the pandemic, although the exact threshold for achieving herd immunity is unknown.⁷ Our analyses illustrate both encouraging and concerning findings. One encouraging finding is that for the overall US population, the likelihood of vaccination has risen by 10-percentage points, from a nadir in late October/November 2020 to 63% by March 2021. A second encouraging finding is that all subgroups have increased their stated likelihood of vaccination. A third extremely encouraging finding is that Non-Hispanic Black and Hispanic populations have substantially increased their stated likelihood of vaccination, greater than the increase among Whites, so that the racial/ethnic gap has narrowed since the election.

Our study also highlights several concerning findings. First, the likelihood of COVID-19 vaccination remains low among young adults aged 18–49, individuals with high school or less education, and individuals affiliated with the Republican party. Second, disparities remain in the likelihood of vaccination by gender, age group, education, race/ethnicity, and political affiliation. Finally, a high proportion of vaccine-hesitant individuals are skeptical about both the effectiveness and safety of COVID-19 vaccines and also do not trust the vaccine development process or the governmental approval process.

The marked rise in interest in COVID-19 vaccines among Black and Hispanic populations and the simultaneous narrowing of racial/ethnic disparities in the public’s interest in COVID-19 vaccines is heartening and likely reflects the success of outreach efforts at local, state, and federal levels, as well as the influence of the media. Given the barriers to accessing COVID-19 vaccines among many Black and Hispanic communities,²⁴ it is critical to pair this rising interest in vaccine with improved access to vaccines.²⁵ Communities across the US are focusing heavily on outreach to vulnerable populations and will need to continue and even intensify these efforts, particularly because the initial phases of the vaccination campaign have focused upon older adults who are more likely to get a vaccine and the later phases now include all adult ages. In fact, most US states have opened vaccine eligibility to individuals aged 16 and older; the present study findings underscore that vaccine hesitancy in younger age groups remains high and this needs to be addressed in future outreach efforts.

Understandably, many people remain hesitant about these new COVID-19 vaccines. Knowing someone who was vaccinated and is now protected might be a powerful motivator for behavior change. Studies suggest that personalized stories by physicians and others may be more powerful in countering hesitancy than recitation of numbers or frightening examples.^26,27 Thus, it is important for individuals and professionals to speak to others about their own COVID-19 vaccination experience and the feeling of protection that follows vaccination. Because disparities remain across all subgroups, trusted advisors and leaders across disciplines have an important role in educating their constituents about the benefit of vaccination. This is particularly important given our findings that vaccine-hesitant individuals lack trust in both the vaccines and in the vaccine-approval process.

A highly concerning finding is the large gap in likelihood of vaccination by gender, age and education. The gender gap may be partly related to concerns about impact on fertility or pregnancy, at least for young women. Although young adults are frequent spreaders of the virus, only about half of adults aged 18–49 years state they are likely to be vaccinated. It is possible that many young adults feel that high-risk individuals are getting vaccinated so they no longer need to receive the vaccine, or that vaccination is not necessary if they have already had COVID-19. Our findings regarding education level strongly suggest that messaging and outreach efforts must take literacy and health literacy into consideration.^28,29 Overall, intense effort is needed to uncover effective methods to address the risk/benefit concerns for COVID-19 vaccines, and focused efforts are needed to educate subgroups via both tailored messages and strategies to optimize access to COVID-19 vaccines.

The marked disparity in the stated likelihood of vaccination by political affiliation, even after adjustment for demographic factors, highlights the potential role of political leaders in assuaging concerns among their constituents and using their power to sway public opinion toward healthy behaviors. Many public officials have posted photos of themselves being vaccinated. To the extent that COVID-19 vaccination has been politicized, it is critical for public officials to lead by example to protect their constituents and our nation.

Our study has multiple strengths. The probability-based online panel is large and nationally representative, by virtue of the fact that respondents are recruited by address-based sampling and are provided with internet enabled tablets if needed. Several studies have demonstrated the representativeness of such panels in comparison to other modes of data collection.^30–33 In addition, the response rate is high per survey, and we incorporated information about receipt of COVID-19 vaccines. One limitation is insufficient survey sample size to assess the likelihood of vaccination among individuals with minority backgrounds other than Black, Hispanic, and Asian race/ethnicity. In addition, while we identified factors related to hesitancy for COVID-19, we cannot distinguish the relative importance of these factors.

We conclude that the likelihood of COVID-19 vaccination has risen overall and particularly among Black and Hispanic populations following the US election, with a narrowing of preexisting racial and ethnic disparities in the acceptance of these vaccines. At the same time, the likelihood of vaccination has not increased much among young adults, and disparities remain by gender, age, race/ethnicity, and education level. To reach herd immunity and achieve confidence in current and future COVID-19 vaccines, we will need to overcome vaccine hesitancy among the general population and all subgroups across the US by building trust in the vaccines and in the vaccine approval process.

Acknowledgments

References