A Join Point Analysis of COVID-19 Policy in Ontario

Join Point Analysis: COVID-19 and Public Opinion Arjumand Siddiqi, Carolyn Tuohy, & Blake Lee-Whiting How did public opinion regarding matters associated with the COVID-19 pandemic change in Canada over the course of the pandemic? Is the timing of change in public opinion associated with government policies regarding the pandemic? Or does timing of change in public opinion suggest that these changes responded to phenomena other than government policies? The innovation we present is to empirically examine when, how often, to what extent public opinion changed over time. This, in and of itself provides valuable information on the state of public opinion during a pandemic, which is a major indicator of how populations might react to societal circumstances and decisions. This information also allows us to cast a wider net on the possible drivers of public opinion. Traditionally, we tend to isolate a particular policy or other phenomenon, and then test whether public opinion demonstrates a response to this phenomenon. However, in the unprecedented and tumultuous times in which we find ourselves, it seems important to also approach these questions in a more open and agnostic way, to let the data indicate to us what might be driving public opinion. Our analysis addresses these questions using join point regression, a novel statistical technique that focuses on rigorous description of change in a variable, rather than hypothesis-driven inquiry of the association between variables. In other words, we will explore how public opinion changed, while holding ourselves blind to policy (or other societal) contexts. We propose analyzing public opinion during COVID-19 over time using join point analysis, and then, subsequently, conducting informed qualitative analysis of statistical-significant changes to determine whether changes in pandemic-era public opinion was associated with changes in the policy-sphere including the adoption of policies, the role of science advisers, the release of reports of commissions of inquiry, shifts in media narratives and so on. Join Point Analysis and Selection Bias Join point analysis is most frequently found in medical research, where studies have looked at cancer incidence rate over time. Tiwari et al (2005, 919-920) succinctly summarize the purpose of this type of analysis: "A question that is of particular interest when analysing cancer incidence and mortality rates is whether or not there has been a change in the trend over time and, if there has been a change, when it occurred. Questions of this type play an important role in measuring progress against cancer and in assessing the effect of population intervention on the outcome of disease. For example, a change in trend for lung cancer incidence may reflect the population effect of antitobacco programmes or changes in the trend for cancer mortality may be the result of new screening modalities… We have found that a log-linear model with random changepoints has been quite useful in modelling and interpreting cancer trends. Since the models define a changepoint as a change in slope, but do not allow a jump in the level at a change, we refer to these types of models as join point models." Join point analysis, therefore, allows researchers to detect statistically significantly changes in a trend over time. Join point models use algorithms which test where a multi-segmented line (time-series data, for instance) are segmented; these segment breaks, wherein a statistically significant change in the slope occurs, are called join points. The technique has been used to a very limited extent to date in examining changes in public opinion: a few examples of such applications have been undertaken by statisticians rather than political scientists (Tian and Porter 2022, Yeung et al. 2020). While we typically have reason not to select cases on the dependent variable in political science or public policy scholarship due to selection bias concerns (Lustick 1996), this concern is not shared equally across the discipline (Skocpol 1984), particularly in terms of historical analysis. Considering that the problem with selection bias is not a methods challenge, necessarily, but rather "occurs when the non-random selection of cases results in inferences, based on the resulting sample, that are not statistically representative of the population" (Collier 1995, 462 in Lustick 1996, 606), we can make a reasonable claim that a mixed-methods approach, in which the statistical join point analysis is complimented by thorough qualitative analysis, may mitigate these concerns. As scholars such as Rogers Smith (1988) and Andrew Stark (2020) have argued, qualitative and quantitative analysis can be complementary: qualitative thick description can provide a more comprehensive portrait and understanding of characteristics of political life that can then be treated as variables in further quantitative analysis. Our approach adds an initial iteration: our quantitative analysis will identify phenomena for further qualitative investigation. In other words, we can use join point analysis as a starting point by which to provide some descriptive analysis of a phenomenon, rather than testing for causation. Citations Collier, David. 1995. "Translating Quantitative Methods for Qualitative Researchers: The Case of Selection Bias." American Political Science Review 89(June):461-5. Lustick, Ian S. "History, historiography, and political science: Multiple historical records and the problem of selection bias." American Political Science Review 90.3 (1996): 605-618. Skocpol, Theda. 1984. "Emerging Agendas and Recurrent Strategies." In Vision and Method in Historical Sociology, ed. Theda Skocpol. Cambridge, UK: Cambridge University Press. Smith, Rogers. 1988. “Political Jurisprudence, the ‘New Institutionalism,’ and the Future of Public Law.” American Political Science Review 82, 1: 89–108. Stark, Andrew. 2020. “Bridges between Wedges and Frames: Outreach and Compromise in American Political Discourse.” American Political Science Review 114, 4: 1280-1296. Tian, Jiahao and Michael D. Porter, 2022. “Changing presidential approval: Detecting and understanding change points in interval censored polling data.” Stat 11:e463. Tiwari, Ram C., et al. "Bayesian model selection for join point regression with application to age‐adjusted cancer rates." Journal of the Royal Statistical Society: Series C (Applied Statistics) 54.5 (2005): 919-939. Yeung, Neil, Jonathan Lai and Jiebo Luo “Face Off: Polarized Public Opinions on Personal Face Mask Usage during the COVID-19 Pandemic.” 2020 IEEE International Conference on Big Data DOI: 10.1109/BigData50022.2020.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.