Russia Polling institutes compared: accuracy and forecasting quality

The PolitPro Score evaluates the reliability of polling institutes by objectively comparing their data with actual election results and the calculated Election Trend. A high score indicates a precise representation of the political mood without systematic distortions. If an institute's values for individual parties repeatedly deviate significantly from the overall trend, this leads to point deductions. Such differences can indicate so-called house effects – methodological peculiarities in weighting that systematically over- or underrepresent certain parties. The score thus offers you transparent guidance on which polls most accurately reflect reality. The score has a maximum value of 100.

Accuracy shows how close pre-election polls were to the actual election result. We compare the pre-election polls of the institutes with the actual result of an election. We compare the average deviation of the parties' values and form an average from them. In the calculation, we consider all parties with an election result of 3% and higher. Election polls have a margin of error of 2-3 percentage points. We rate average deviations of up to one percentage point as very good (green), and up to 2 percentage points as good (yellow). Values up to 3 percentage points deviation are rated as ok (orange), but already show significant outliers for some parties. Average deviations of more than three percentage points indicate deficiencies in accuracy and significance and are rated as bad (red).

The deviation calculates how often an institute's values significantly deviate from the Election Trend. Frequent deviations for individual parties can indicate favoritism or disadvantage. We calculate an institute's deviation from the values of individual parties in election polls. For this, we compare each party's values with the PolitPro Election Trend, which calculates a weighted average from current voting intention polls. If a party's value is within +/-1 percentage point of the Election Trend, we rate it as equal. If it is lower, we rate it as lower, and if it is higher, we rate it as higher. From all polls, we then calculate in how many cases an institute rates a party higher or lower (see overview on the institute subpages). We consider deviations of up to 20% as normal. If deviations are higher/lower in more than 20% of cases, we indicate the party with the corresponding arrow. Note: We recommend always considering the deviation in combination with accuracy in elections.