One poll can be misleading and cherry picking a polling result, which fits your expectations, even more. We aggregate all polls we come across and show a weighted mean trendline in our interactive visualization. Additionally, every data point in the chart represents one result from one poll. This gives you a sense of the variance that comes with polling data.
Margin of Sampling Error: Variance between different polls is not the only source of uncertainty, also every individual poll comes with a margin of error of its results ranging from 1.5 to sometimes as high as 4.5 percentage points if the sample size is small. We are going to publish the margin of error (MOE) along with the sample size in our polling data tables when possible. Keep in mind that every polling number has to be interpreted as a range of plus-minus this error and that the "true" support for the respective party is going to lie with 95 percent confidence within this range.
The mean trendline for
a) the Kalman-Smooth trend: The default trend line in all our charts is based on the Kalman Filter values.
b) the Kalman-Filter prediction: Using a so-called Kalman filter we aggregate all polls by starting from the last election result as a fixed starting point and taking every new poll with its sample size and margin of error as a new piece of information for the trend line. Therefore, the sample size is automatically considered in this approach. As you can see the values and course for the weighted mean and the Kalman filter do not diverge heavily, but still, the two options give you an idea that the aggregation model matters as well, even though the difference might just be one percentage point up or down for a certain political party.