How Might Undecided Italians Vote Tomorrow?
Ahead of Sunday’s Italian general election, a significant proportion of voters, up to 35%, are currently undecided or refuse to tell pollsters for whom they will vote. To a large extent, their choice will determine the outcome of the election. In this article, I analyse Ipsos Italia* data to analyse the views, past voting behaviour and socio-demographics of this group in order to predict how they might vote.
Ahead of Italy’s election this Sunday, 4 March, the country’s polls have been pointing in a fairly consistent direction. The anti-establishment Movimento 5 Stelle (M5S) are likely to receive the most votes, around 27% of the total according to the latest polls. Following them should be the incumbent, centre-left Partito Democratico (PD), who have been in continuous decline since the 2014 European elections and Matteo Renzi’s ill-fated 2016 constitutional referendum, at around 22%. With around 16% and 14% respectively, are the ‘centre-right’ Forza Italia and the populist radical right Lega, the latter of which has been running on a particularly anti-immigration platform.
As has been noted elsewhere, these headline polling figures overlook the proportion of voters who remain undecided or refuse to disclose for whom they intend to vote. This figure is usually between 20% and 30%. In a recent poll by Ipsos Italia, a full 35% of respondents either said they did not know how they would vote or refused to say (see Figure 2).
How might this group vote? First, let’s consider their ideological outlook. In Figure 3, we can see how voters of the main parties, undecided voters and the Italian electorate as a whole responded when asked to place themselves on a 1–7 left-right ideological scale. The ideological distribution of the four major parties is predictable enough, with only M5S voters showing a high degree of internal variation. The Italian electorate as a whole is multi-peaked, with a slight left-ward skew. Conversely, undecided voters (labelled as “Don’t Know”) are strongly overrepresented in the centre-right — 34% describe themselves as such. Moreover, undecideds are more moderate than any other group — only 22% class themselves as far left, left-wing, right-wing or far right.
However, in spite of their centre-right skew, when we consider what undecideds (“Don’t Knows”) think of the current centre-left government led by Prime Minister Paolo Gentiloni of the centre-left Partito Democratico, we see a high degree of support. As shown in Figure 4, when asked to rate the current government on a 1–10 scale, the average undecided voter responds with a 5. This makes them more positive to the Gentiloni government than voters of any of the other parties, particularly M5S, as well as those who plan to abstain.
Third, I consider the voting history of these undecided voters. In Figure 5, we see the 2013 general election and 2014 European election voting behaviour of the entire electorate in the two outer rings and those who are currently undecided in the inner rings. We can see that in 2013 currently undecided voters were, compared to the electorate, less likely to vote for the PD, more to vote for the M5S and more likely to vote for the parties that inherited the Christian Democrat vote, the Union of the Centre (UdC) and Scelta Civica. At the 2014 European Parliament elections, currently undecided voters were slightly more likely to vote for Renzi’s PD and considerably more likely to vote for the M5S. In each election they were far less likely to vote for Lega than the average voter.
From all of this, it seems that some of Italy’s undecided voters have been faced with something of a dilemma. On the one hand, they are relatively sympathetic to the current government. They also have far less history of supporting extreme parties, including Lega Nord, than other Italians do, rather turning towards smaller centrist parties and the M5S. On the other hand, they are considerably more likely to be centre-right in orientation than PD or M5S voters. Furthermore, they have a history at two consecutive national elections of being more likely than the average voter to support M5S.
Under such circumstances, it might be expected that the behavior of one’s peers, the electoral campaign and habit to have especially strong effects on these groups. We can use variables that might pick up such effects and the ones already discussed to build a model to predict the vote choice of each of the currently undecided voters, based on their socio-demographics, political attitudes and voting history. Using the results of a logistic multinominal model with relative risk ratios, we can then compare the vote choice of the undecideds with the rest of the electorate. I use the following socio-demographic predictors: geographical area, gender, age, university degree, employment status and religiosity. Also included is a variable that measures whether individuals have close friends from a non-Italian ethnic background, as placed in the Ipsos Italia polling by the Observatory of Public Attitudes to Migration. The four political variables already discussed above are also included: left-right self-placement, approval of the current government, vote choice in 2013 and vote choice in 2014.[1]
Naturally, such a model based on just one poll comes with important caveats, mostly related to the representativeness of the sample, whether the predictors of party choice for decided voters are the same as those for undecided voters and the correct estimation of turnout (discussed below). Each of these will only be at best partially met.
The resultant model predicts the breakdown of undecided voters on March 4 as shown in Table 1, below. As we can see, aside from the quarter of undecided voters who will not vote, using socio-demographic and political variables predicts that undecided voters will be more likely to vote for M5S, for Forza Italia and for the pro-European centrists of Emma Bonino’s Più Europa than other voters are, as shown in our data, and less likely to vote for Lega Nord. However, the latter may be underestimated due to the party’s transformation since the 2013 and 2014 elections.
Overall, despite concerns about extremism and instability in Italy, recent data suggests that the large proportion of undecided voters in the electorate are actually fairly centrist and reasonably positive towards the current government. However, when predicting future voting behaviour (albeit from just a single poll) by taking this into account, alongside past voting behaviour and socio-demographic indicators, undecided voters seem more likely to vote for M5S and the centre-right than other parties and so these two groups may do better than polls have expected. The latter have been predicted to do particularly well in the constituency vote. Lega have been seen by many as the big winners of the election, but there seem to be limits to its appeal not least because of emerging tensions during the campaign with supporters in its traditional strongholds in northern Italy as it tries to recast itself as a national party. Finally, it should be noted that the new hybrid electoral system is likely to translate vote share to seats in still unpredictable ways.
[1] In order to correct for differential turnout, the following three groups are reclassified as ‘will not vote’: those who say they won’t vote; those who say they will vote for a party but when asked of their probability of turning out say it is less than five out of ten; those who say they are undecided but whom the model predicts won’t vote. This excludes 18% of respondents — still very likely to be an underestimate.
*Thanks to Professor Andrew Geddes, Dr. Lenka Dražanová and Dr. Noah Carl for their help on the write up and the politics team at Ipsos Italia in Milan for providing the data for the above analyses.
