February 27, 2021

The 2021 Norwegian Parliamentary Election

This page estimates party support over time based on opinion polls. The page is updated daily and reports estimates for each week from the last election until the next.

Who will get a parliamentary majority?

The plot shows the probability that a party constellation would get a majority if an election had been held during a given week.

The blue block currently has a 10.8% probability of winning a parliamentary majority at the next election. If an election had been held today, the blue block would have had a <1.0% probability of winning a majority.

Who will get above the 4% threshold?

The plot shows the probability that a party would get above the 4% threshold if an election had been held during a given week.

The Green Party (MDG) currently has a 19.5% probability of getting more than 4% of the national vote at the next election. The equivalent probability for the Liberal Party (V) is 12.9%. For the Christian Democratic Party (KrF) it is 11.7%, and for the Red Party (R) 7.2%.

How many seats will each party get?

The plot shows weekly levels of support in the form of predicted parliamentary seats.

If an election had been held today, the Conservative Party (H) would be predicted to be the largest party in parliament, with 48 seats, ahead of the Labor Party (Ap), at 42 seats. The Centre Party (Sp) would be predicted to gain the most seats, adding 18 compared to the last election. The Labor Party (Ap) would be predicted to lose the most, obtaining 7 fewer seats.

How is party support geographically distributed?

The map shows the total predicted vote share for the blue block at the next election. Move the marker over a district for details (on touch screens: tap). Click and drag to zoom. Double-click to return to the original position (on touch screens: tap with two fingers).

Rogaland is the most Right-leaning district in Norway, with a 53.8% predicted vote share for the blue block at the next election. Nord-Trøndelag is estimated to be the most Left-leaning, with a 27% predicted vote share for the blue block.

How do the polls look?

The reported date for each poll is the midpoint between the start and end of the data collection. The plot only shows national polls, although the model also uses polls at the district level.

In addition to national polls, the plot shows the estimated trends of support in the population. Because the model corrects for differences between polls and actual results in previous elections, some trends are adjusted slightly up or down relative to the moving average of the polls.

How does the model work?

The results reported above are produced by a Hamiltonian Monte Carlo simulation of a Bayesian model. To get a satisfactory grip on the uncertainty of the results, all parameters are estimated simultaneously via simulation: No point estimates are made until the joint probability distribution of all the parameters has been estimated.

The model accounts for both random variation and polling bias, although the latter is fairly limited. If we consider the polls conducted during the last two weeks before the last two elections, their averages have for the most part predicted the election results reasonably well, but there are some minor deviations that the model corrects for. These deviations have two likely explanations: (1) The voters of some parties are more prone to answer surveys (and the polling companies are not fully able to correct these differences through weighting), and (2) the voters of some parties have a lower turnout rate.

At the district level, the data are more limited, and national polls therefore play an important role in modeling support also at this level. The model estimates how election results at the district level normally relates to national results, which in turn allows the national polls to influence district level estimates. However, the predictions at the district level are still more uncertain than those at the national level.

The predictions for the next election are based on five elements: (1) Vote shares at the last election, (2) the degree of change in each party’s vote share from one election to the next, (3) current party support, (4) the degree of change in each party’s support from one week to the next, and (5) the degree of covariation in how support for each possible pair of parties evolves on a weekly basis.

To keep track of new polls, this project relies on Poll of polls and the project might not have been feasible without their effort to systematize the polls. The idea behind this project is to complement their contribution by going further in terms of modeling party support and predicting election results.

Downloading and using the data

To facilitate further use of the content on this page, the .csv files listed below are updated daily. These files can be used freely, provided appropriate attribution is given: “Jørgen Bølstad / Estimite.com”. The files contain results for the current week, as well as predictions for the election:

About this website

Further information about this site can be found here.