To assess attitudinal and behavioral change as a result of events in 2020, we used a longitudinal panel approach, which provides us the ability to gather repeated observation on a set of variables for the same sample units over time. In this particular case, we re-surveyed respondents from our 2019 Hoosier Life Survey (HLS1). To gather our existing sample and data, HLS1 surveys were sent to 10,000 Indiana households across the state. Addresses were purchased from a private address-based sampling vendor. In mailing surveys to these households, we used a modified Dillman approach, with a total of five mailing waves.
For the second wave of the HLS survey, we re-surveyed those respondents from our original sample who agreed to participate in future studies. Of our approximately 2,700 respondents to the 2019 survey, 2,021 (76 percent) agreed to receive requests to participate in future studies requests. Our decision to draw from our existing sample reflects a non-probability design. However, such a panel-sampling design is widely considered one of the most effective means to determine the amount of change in key variables of interest—in our case, individual’s climate- change risk perceptions, views, and behaviors—and what drives changes in these areas.
That the 2019 HLS, our baseline measure, was completed just before the emergence of COVID-19 in the United States provided us with a unique opportunity to understand the impact of this pandemic and other events in 2020—particularly in its relation to changing views about climate change—on the Indiana public. To ensure our ability to assess this change, we included a subset of key questions from the 2019 HLS in the second wave of the Hoosier Life Survey (HLS2) exactly as they were worded in the original HLS1.
To reach this sample, we used a unique mixed-mode, adapted tailored design approach. This delivery approach capitalized on existing contact information. We split our sample into two categories: those who provided email addresses and those who did not. Of the more than 2,000 HLS1 respondents who agreed to participate in future studies, more than 1,200 provided their email addresses. To this group, we emailed an invitation, including a link to the survey. The online survey was designed and hosted on IU’s Qualtrics account, a widely used online survey-design software program. To the approximately 800 HLS1 respondents who agreed to participate but did not provide an email address, we mailed a survey packet and cover letter to their known home address. Non-respondents in both groups received follow-up notifications. This second wave of contacts also contained a cover letter and either a survey link or the survey packet (depending on delivery mode). Each respondent who completes the survey was entered to win a $50-dollar gift card (one of 20 available in total) as a post-incentive. Our survey ran between October 2020—March 2021
Of the 2,021 potential respondents, approximately 1,200 returned our survey for an unweighted response rate of approximately 59.4%.
To ensure accurate population estimates for this analysis, survey weights were used. Weighting incorporates: (1) a base weight adjustment for unequal probabilities of selection due to disproportionate stratified sampling by geography and due to the number of adults in the household, and (2) a calibration adjustment to the 2019 5-year American Community Survey estimates for age by sex, education, and race in the Indiana adult population. Weights have been trimmed and scaled to the unweighted number of respondents.
We appreciate the funding we received for this study from Indiana University’s Grand Challenge Initiative, Prepared for Environmental Change. We also would like to thank Indiana University’s Center for Survey Research and Jonathan Hines for their critical support in producing this study and report.
 To ensure maximum statistical accuracy for each analysis, we use pairwise deletion to address missing data. In consequence, the exact number of respondents will vary across questions. Unless otherwise noted, total sample size ranges between approximately 1,180-1,220 for this report. Our most conservative estimate for the confidence interval is +/-4.48% accounting for the design effects of sample weights (i.e., deff)
 For this report, “Republican” represents respondents who selected “Republican” or “Independent, Lean Republican. “Democrat” represents respondents who selected “Democrat” or “Independent, Lean Democrat.” Those who selected “other” political affiliation (n=71) were omitted from all political-focused analysis. This coding reflect grouping we used previously in the Hoosier Life Survey Politics Report.
 Age based definitions of the generational cohorts are as follows: Generation Z (age 18->24); Millennials (age 25->40); Generation X (age 41->56); Baby Boomers (age 57->73); and the Silent Generation (age 74 and above). These definitions are based on those used by the Pew Research Center. See here.