No truth stands alone.

Pandemic Politics: Minor Party Voting Intention

In our previous pieces in the Pandemic Politics series, we examined whether elections held during the COVID-19 pandemic saw significant swings in the vote for incumbent governments and minor parties from the last election held in that electorate (parts 1, 2, 4). However, that approach has a significant limitation, namely that it can only assess whether COVID-19 had an impact on voting compared to the last election held in each country, instead of assessing if COVID-19 had an impact on voting intention going into 2020. For example, if there was a rally-around-the-flag effect in a particular country, a government on track for a landslide loss might instead see little vote change from the last election, producing the appearance of no change.

Hence, in order to assess whether minor party voting intention changed as a result of the COVID-19 pandemic, we have to use public opinion polling before and during the pandemic. This brings up several issues of its own (click/tap each for a full explanation):

  1. This is, broadly speaking, the many variations of polling error which many people have become familar with in recent years.

    Polling error can have many sources, ranging from sampling error (randomly finding more/less voters for party X in a sample than in the population), unrepresentative samples stemming from improper or failing to weight data by relevant demographics, to more prosaic issues such as pollster herding (where there are abnormally few outlier polls) and non-response bias (where people who don’t respond to polls voting significantly differently from those who do despite sharing common demographic characteristics).

    Unlike the elections we looked at in our previous pieces, polling is not as comprehensive a tool in assessing public opinion. It still remains a very useful tool for discerning trends in public opinion (and in this case, for assessing the impact of a pandemic on electoral politics).

    However, when dealing with instruments with higher margins of error like polling, less confidence is required when analysing the data collected.
  2. Generally speaking, countries with regular polling of voting intention tend to be disproportionately Western, educated, and more populous than democracies as a whole.

    One particular difference that stands out in such democracies is a tendency toward ideological parties instead of “personal vehicle” parties, i.e. parties in such countries are more likely to run on a single, unifying ideology (e.g. progressive, conservative, nationalist, libertarian) rather than being formed to support the political ambitions of a popular leader.

    This may reduce swings due to public opinion of a party leader (as ideological parties tend to have a core base of partisans who will vote for the party even if they dislike the leader) and increase the size of any effects of the pandemic.

    Furthermore, as we are looking at minor parties, the electoral systems used in democracies here will tend to be disproportionately non-proportional (i.e. plurality, majoritarian), as many democracies with proportional representation tend to see the development of a multi-party system in which it is unclear which, if any, parties are major/minor. The use of non-proportional systems may dampen the effect of any swing to/from minor parties, as it discourages voting for candidates who are perceived to be unlikely to win, ensuring that the only voters who do tend to be dedicated to that party.

    In plurality systems, this also links to points 1 and 4; for example, it is generally accepted that third-party voting intention declines over time in polling of United States of America elections. Hence, in non-proportional systems, polling too far out from an election may overestimate the minor party vote.
  3. While polling of minor parties was fairly accurate this year (polling error ranged from -6.2 to +5.0, which isn’t too bad as far as polling errors go), the pandemic may produce systematic non-response biases by making it more likely that some voters respond to polling (e.g. voters whose jobs allow them to work from home may be overrepresented in landline polling) than others.

    In particular, early hypotheses about the cause of polling error in the 2020 US elections suggest that polling error and coronavirus cases may have been correlated, which would suggest that the pandemic may have made it harder for pollsters to contact a representative sample.
  4. In some ways, this is a subset of point 1, but it’s worth pointing out that polling too far out from an election may not be accurate.

    For example, it’s generally accepted that third-party voting intention declines in US presidential elections as the election campaign wears on; this may also apply to other democracies with plurality electoral systems such as the United Kingdom.

    This is likely due to polled respondents providing an affective response (a snap judgement of the parties based on how they feel about them at the moment) rather than a considered opinion on how they intend to vote. For example, a leftist disenchanted with their centre-left major party’s centrism may tell pollsters that they will be voting for a left-wing minor, only to switch back to the centre-left major once negative partisanship (dislike of “the other side”) kicks in.

    Furthermore, although I intend to go into this in more depth in the incumbent government polling piece, a cursory look at some of the polling of voting intention for incumbent governments suggests unsustainable levels of support, likely from voters who would usually never vote for that party but who support their government’s COVID-19 response.

    Even in a country whose government is widely recognised as having successfully contained COVID-19 and whose 2020 re-election is often credited to said containment (New Zealand), some reversion to the mean in their polling from the massive highs seen in the early months of the pandemic is fairly evident.

However, as compared to swings in elections, polling has the advantage of enabling more of an apples-to-apples comparison by virtue of allowing comparisons from similar time periods. This also helps to isolate the effects of the pandemic on voting intention by allowing comparisons to be made between voting intention before the pandemic and during the pandemic itself, instead of looking at swings between elections held at different times. To decide which countries’ polling to include in our analysis, we used the following criteria:

  • Either minor parties + independents combined won >= 5% of the vote, or a minor/independent won a seat at the last election, or at the 2020 election (for countries which held an election in 2020).
  • Elections are generally agreed to be fair and impartial, and held in nations with a Democracy Index >= 6 as of 2019
  • Opposition or other significant parties (won >= 5% of vote at last fair election) did not boycott election
  • Need to be able to determine voting intention from given polling data
  • Polling data had to be available for the relevant periods under investigation
    • At minimum, there had to be polling data from Nov/Dec 2019, and either polling from Apr 2020 or Dec 2020 [if no 2020 election] / final pre-election polling [if the country held an election in 2020]
  • Must be polling for a general election and not a by-election
  • Where possible, legislative elections were used instead of executive elections.
  • Where there were separate constituency and proportional polls conducted simultaneously, the results of the proportional election were used if possible.
  • If major party status is unclear in a system with separate constituency and proportional elections held simultaneously, the vote share of the parties in the constituency elections will be used to determine which party holds the apparent status of major party.
  • Must be reasonably able to determine major parties.
  • Countries with elections held during or before Apr 2020 were excluded.

First up, let’s take a look at how minor party voting intention changed 1 month into the pandemic:

(if you’re on a mobile device, scroll right for full data or turn your device landscape)

The pandemic seems to have initially suppressed minor party voting intention
Nation/subdivisionSwingEnd-20191 month into pandemicRegionHeld an election in 2020?
South Korea-417.213.2AsiaYes
Basque-0.336.736.4EuropeYes
Vienna-84739EuropeYes
Dominican Republic7.821.429.2North AmericaYes
New Brunswick-23129North AmericaYes
British Columbia-83325North AmericaYes
Queensland-13130OceaniaYes
New Zealand-0.51514.5OceaniaYes
Romania11.933.845.7EuropeYes
Croatia-0.625.224.6EuropeYes
Galicia-5.336.230.9EuropeYes
Saskatchewan-0.813.913.1North AmericaYes
Sweden-1.233.332.1EuropeNo
Denmark-4.350.646.3EuropeNo
Canada-43733North AmericaNo
Australia-12524OceaniaNo
United Kingdom-2.72320.3EuropeNo
France24446EuropeNo
Portugal-5.140.435.3EuropeNo
Malta033EuropeNo
Greece-4.529.525EuropeNo
Bulgaria1.845.247EuropeNo
Brazil10.62737.6South AmericaNo
Poland-43632EuropeNo
Average-130.629.7

The pattern becomes even clearer when you look at a histogram of minor party voting intention swing 1 month into the pandemic:

Histogram of minor party voting intention swing 1 month into the pandemic, with a fitted skew-normal distribution.
Histogram of minor party voting intention swings (in all countries) 1 month into the pandemic (April 2020), along with a skew-normal distribution fitted using maximum-likelihood. Data was combined from both nations which held an election in 2020 and those which did not as there isn’t much difference between the two.

(An interesting data-quality check here can be seen in the shape of the data collected. Given that we’re analysing minor parties’ voting intention, the skew seen above [where the distribution seems “tilted” to one side instead of being symmetrical] is expected from a theoretical standpoint Since we’re analysing minor party voting intention here, we expect that minor parties have a relatively small share of the vote (by definition). As negative vote shares are not possible (in commonly-used voting systems), if the pandemic’s effect is to reduce minor party vote share, then we should expect a degree of “bunching up”, where we get a lot of negative swings of roughly the same magnitude (in the histogram it seems to be 0 to -5), with some of larger magnitude in countries where minor parties have a large share of the vote, and relatively few positive swings. In other words, the distribution of minor party voting intention swings theoretically should not be symmetrical if the pandemic reduced minor party voting intention. , which helps us be more confident in the data we’ve collated)

This swing is statistically significant p-value = 0.0442 using a Wilcoxon signed-rank test. I opted to use this instead of the more standard t-test due to the skew evident in the data. , suggesting that it is very likely that there was a shift against minor parties in the early months of the pandemic. During this time period, there was a very weak correlation between COVID-19 caseload (measured by COVID-19 cases per 1m population) and the swing in minor party voting intention:

Scatterplot of Apr 20 COVID-19 caseload vs minor party swing
Scatterplot of COVID-19 caseload versus change in minor party voting intention as of April 2020, along with a linear regression (in red) fitted using least-squares.

However, this relationship may be an artifact of the significantly positive outliers to the left, which saw large upticks in minor-party voting intention (Romania, Brazil and the Dominican Republic), or the very small number of high COVID-19 caseload countries in the dataset (this is a particular problem given least-squares regressions can be very sensitive to outliers).

Despite these issues, however, larger minor party voting intention declines in countries with high COVID-19 caseloads does lend support to the hypothesis that COVID-19 initially dampened minor party voting intention. If we found no correlation between COVID-19 caseload and the change in minor party voting intention, we would have to ask ourselves if the decline in minor party voting intention we observed previously was genuinely due to the pandemic or perhaps an artifact of the particular democracies which we had polling for.
Yes, the decline is statistically significant, but it’s not that hard to achieve statistical significance as statistical tests can only tell you if an outcome is unlikely under certain specific conditions (i.e. the null hypothesis).
If those assumptions are violated (e.g. in this case, we assumed the null hypothesis to be one of no change in minor party voting intention, but maybe we should expect some decline in minor party voting intention as electorates get closer to their elections), then it is entirely possible that we could have seen a statistically significant decline in minor party voting intention even if the pandemic had no impact.
However, this does not mean that COVID-19 produced a decline in the actual vote share of minor parties in elections; the rally-around-the-flag effect generated by disasters often fades rapidly especially as an election approaches. To assess this, we can examine how minor party voting intention shifted in final pre-election polling:

(if you’re on a mobile device, scroll right for full data or turn your device landscape)

Any effect of COVID-19 on minor party voting intention likely disappeared by election day
Nation/regionSwingEnd 2019Final pre-election pollingPolling errorElection dateRegion
South Korea-3.817.213.4-4.82020-04-15Asia
Dominican Republic8.421.429.8-2.22020-07-05North America
Croatia1125.236.21.32020-07-05Europe
Basque-3.136.733.60.12020-07-12Europe
Galicia-4.236.2320.62020-07-12Europe
Northern Territory-832245.22020-08-22Oceania
New Brunswick1.33132.3-62020-09-14North America
Vienna-84739-12020-10-11Europe
New Zealand8.11523.10.22020-10-17Oceania
British Columbia-12.53320.5-22020-10-24North America
Saskatchewan-6.913.970.12020-10-26North America
Queensland-43127-2.52020-10-31Oceania
Romania14.933.848.7-2.82020-12-06Europe
Average-0.528.728.2-1.1
One thing to note here is the polling error in the final pre-election polling, which ranges from -6 to +5.2. The fact that we're using polling means we need to be extra careful about drawing conclusions from this data due to the inherent inaccuracy in polling a subset of the electorate.

While there is still a small decline on average, this decline is non-significant p-value = 0.4721 from a Wilcoxon ranked-sign test, or p = 0.4141 from a t-test. , a fact which becomes rapidly apparent when graphing these swings:

Histogram of minor party voting intention change by final pre-election polling in countries which held an election in 2020.
Histogram of minor party voting intention change by final pre-election polling in countries which held an election in 2020, along with a fitted normal distribution.

Furthermore, there is very little correlation between COVID-19 caseload and minor party voting intention swing by election day:

Scatterplot of COVID-19 caseload versus change in minor party voting intention by election day.
Scatterplot of COVID-19 caseload versus change in minor party voting intention by election day, along with a linear regression (in red) fitted using least-squares.

Taken together, these seem to suggest that any effect of the COVID-19 pandemic on minor party voting intention disappeared by election day, not unlike a short-lived rally-around-the-flag effect which often follows natural disasters. To some extent, this is further backed up by polling data from countries which did not hold an election in 2020:

(if you’re on a mobile device, scroll right for full data or turn your device landscape)

A year into the pandemic, minor party voting intention has pretty much reverted to normal
NationSwingEnd 2019Dec 2020Region
Sweden-2.533.330.8Europe
Denmark0.350.650.9Europe
Canada-73730North America
Australia-42521Oceania
United Kingdom0.32323.3Europe
France64450Europe
Portugal-2.540.437.9Europe
Malta-231Europe
Greece0.629.530.1Europe
Bulgaria2.745.247.9Europe
Brazil52732South America
Poland8.73644.7Europe
Average+0.532.833.3

By December 2020, minor party voting intention recovered, with the average actually being slightly higher than during the same period in 2019 (although, again, it isn’t statistically significant). A graph of minor party voting intention swing makes the fact of there being no change abundantly clear:

Histogram of minor party voting intention change by Dec 2020 polling in countries which did not hold an election in 2020
Histogram of minor party voting intention change by Dec 2020 polling in countries which did not hold an election in 2020, along with a fitted normal distribution.

However, despite this, there is actually a correlation between minor party voting intention swing and COVID-19 caseload in countries which did not hold an election in 2020: The correlation is statistically significant, according to an analysis of variance (ANOVA), p-value = 0.042.

Scatterplot of COVID-19 caseload versus change in minor party voting intention by Dec 2020
Scatterplot of COVID-19 caseload versus change in minor party voting intention by Dec 2020 along with a linear regression (in red) fitted using least-squares.

Given the evidence for COVID-19 not having had an effect on minor party vote share, however, the reason for this is unclear. It may be a result of the particular democracies which both did not hold an election in 2020 and had readily available polling, many of which hold very high self-expression values In our dataset, the country with a lowest self-expression index is Bulgaria, at approximately -0.5, while the lowest country overall is Egypt, with an index of about -2.25.
Many of the countries in the dataset have very high self-expression values, with two (Denmark and Sweden) being part of the Protestant Europe cluster, which is so disproportionately self-expressive it distorts the graph (notice how the x-axis goes from -2.50 to +3.50).
(which is meant to correlate with a desire for autonomy from central authority). Hence, the increase in minor party voting intention seen here may be a turn to populist/libertarian minor parties in response to the lockdowns seen in countries with heavier COVID-19 caseloads. Alternatively, it may just be an artifact of the countries in our dataset – for example, in Poland, which has both a high COVID-19 caseload and the largest swing to minor parties in the dataset, the minor party voting intention uptick was due to the founding of the new Polska 2050 party, which had very little to do with the COVID-19 pandemic.

Although it is less than ideal to do so, if we combine data from the countries-with-2020-election dataset and the data from countries without an election in 2020, there seems to be very little effect of COVID-19 caseload on minor party voting intention overall:

Scatterplot of COVID-19 caseload versus change in minor party voting intention in all countries.
Scatterplot of COVID-19 caseload versus change in minor party voting intention in all countries.

Thus, given that we did not see any overall effect of the COVID-19 pandemic on minor party voting intention, it is more likely than not that the effect seen in countries with no election in 2020 is a spurious one. Furthermore, as polling from countries which did hold an election in 2020 shows, any effect of the COVID-19 pandemic appears to evaporate by election day , which suggests that even if COVID-19 caseload produces a shift towards minor parties, such an effect will likely fade by the time these countries hold their elections. Why should we trust the conclusions from analysing countries which held an election in 2020?
Well, firstly, the presence of an election helps to solidify voter choice in a way which simply asking a voter about in a survey may not. A good example of this is how voting intention for third-party candidates in the US Presidential election tends to decline as the election campaign continues; voters who may not be satisfied with the major party candidates may opt to tell pollsters they are voting for the third parties, then gradually switch to a major party candidate once faced with the possibility of the candidate they dislike more geting into office.
This is something known as an affective response, where respondents may answer pollsters based on a snap judgement rather than their considered opinion (something which we’ll explore further in the next piece).
In other words, if the data from countries which held an election in 2020 contradicts data from the ones which did not, the data from countries with a 2020 election should be considered.
In addition, research suggests that polling shifts due to crises generally fade fairly rapidly, especially by election day.
Hence, there is additional evidence for the notion that the effects of the COVID-19 pandemic “should” fade rapidly, making it more likely that any effect of COVID-19 caseload on minor party voting intention in countries without a 2020 election is either spurious or will fade by election day.

Hence, it is unlikely that the COVID-19 pandemic had any effect on minor party vote share or minor party voting intention, especially by the time an election rolls around. While it’s plausible that the issues raised in the midst of a pandemic can favour certain parties over others, the idea that minor parties running in elections held in 2020 suffered due to a lack of coverage or were otherwise crippled overall by pandemic politics is unlikely at least, especially when considering our previous piece found no change in the minor party vote in elections held during the pandemic.


Add Your Comment

* Indicates Required Field

Your email address will not be published.

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.

× Home Politics Projects
×

We detect you're using an outdated browser, for which this site is not optimised.

For your security, we strongly recommend you download a newer browser.

Although any of the latest browsers will do, we suggest the latest version of Firefox.

Download Firefox now

(Click on the button in the top-right to close this reminder)