Updated as of 8/30/2018
The Russian public opinion tracker includes public data from three leading Russian polling organizations: The Levada Center, The All-Russian Public Opinion Research Center (Russian acronym VTsIOM) and the Public Opinion Foundation (Russian acronym FOM).
The data are drawn specifically from similar questions asked by each organizations on a regular basis. Levada and FOM both ask respondents whether they approve of Vladimir Putin’s actions in his job. Unfortunately, VTsIOM’s only question that refers specifically to Putin asks about respondents’ trust of Putin, which obviously does not get at job approval.
As a compromise, the tracker substitutes VTsIOM’s question asking respondents about their approval of the actions of the president and assume that respondents understand that they are rating Putin’s actions. The approval ratings from all three polls track each other quite closely, so the difference in wording in the VTsIOM question does not appear to make a difference during the time analyzed in the tracker.
The questions, in English translation and in Russian, are as follows:
En: “On the whole, do you approve or disapprove of Vladimir Putin’s actions as president (prime-minister) of Russia?”
Ru: “Вы в целом одобряете или не одобряете деятельность Владимира Путина на посту президента (премьер-министра) России?”
En: “Do you think Vladimir Putin has performed well or poorly in his post?”
Ru: “Как вы считаете, президент В. Путин работает на своем посту хорошо или плохо?”
En: “On the whole, do you approve or disapprove of the actions of…the president of Russia?”
Ru: “Вы в целом одобряете или не одобряете деятельность…Президента России?”
Trend line calculation:
The trend line is calculated using local regression (loess), which plots a smooth curve over the data. In a nutshell, local regression weights the influence of each data point (in this case, each poll) on the trend line by its distance from the portion of the line being drawn. This means that closer points have a greater influence on the curve of the line than further away points. Sites like FiveThirtyEight and others use local regression in their trackers of American public opinion because it smooths out data from a wide range of polls and, hopefully, reduces some of the noise inherent in public opinion polls.
The degree, however, to which local regression techniques smooth the data is variable. Too much smoothing can produce a line that doesn’t actually capture changes in the data, making it not particularly useful for understanding shifts in political opinion. On the other hand, too little smoothing can produce a curve that moves with every little shift in the data and captures uninformative or misleading noise in the data. Since there are only three polls included in the tracker, I have opted for a model that is more conservative than US-focused approval trackers that use much more data. This means that the trend line calculation is less responsive to rapid changes in Putin’s approval, but is also less responsive to potential outlier poll results. The tracker does provides information on each individual poll result, so you can compare the trend line to the reported approval numbers at any point.
Finally, the trend line calculation also provides a measure of uncertainty in the form of 95% confidence bands. You’ll notice that a lot of the individual polls lay outside the bands. That is because they represent uncertainty around the estimation of the local regression curve rather than the region where we would predict 95% of the results. The latter region would be a prediction interval, which may replace the 95% confidence bands in a later version of the tracker.
The tracker is scheduled to be updated every Thursday. This allows for the incorporation of the weekly data published by FOM and VTsIOM. The actual date of the update is included in the tracker itself in the form of a ‘Last Updated on:’ note in the title.
Please feel free to use the tracker for any purpose – it just aggregates data that is publicly available from the three organizations described above. If you do use the tracker somewhere, please link to it or cite it however makes the most sense for your usage.
All of the R code and data used to create the tracker is available on GitHub. If you have any questions or comments, please don’t hesitate to reach out to me in the comments or send me an e-mail.