Final Recap of Our Forecast of the March 4th Italian Legislative Election

It is now time to recap what our model-based predictions have been saying. As explained at some more length in our first post, we have been trying to predict the results of March 4th based exclusively on the evolution throughout the term of a relatively small set of socioeconomic fundamentals. Using municipalities as our units of observation, we have first estimated the parameters of our model on the last legislative election (2013).

Then, we have exploited them to forecast the future performance of the five main parties that have "survived" the term.  Crucially, to predict the outcome of the upcoming race, we have introduced a penalty to the Democratic Party. Such penalty is aimed at capturing Italian voters' typical fatigue with the incumbent, and is computed based on insights from past elections. Tables 1 and 2 summarize the vote shares and seats resulting from this exercise.

Table 1 |  Predicted vote shares according to the evolution of socioeconomic fundamentals and incumbency disadvantage.

Table 2 | Predicted seats according to the evolution of socioeconomic fundamentals and incumbency disadvantage.

Even without resorting to any interview of individual voters, and thus leaving out any consideration related, for instance, to personal tastes, leaders' appeal, and the like, our model delivers reasonable predictions. With "reasonable", we mean fairly close to those of the most recent polls. An advantage of our framework is that, by predicting results at the most disaggregate level, it allows us to predict the allocation of seats in both the single- and the multi-member tier of the Italian Chamber of Deputies.

Tables 3 and 4 show how our forecast is affected when introducing a proxy for the intensity of a phenomenon that has been dominating the electoral campaign: the refugee crisis. Exploiting measures of the intensity of refugees' presence at the level of individual municipalities, we predict that the Center-right coalition will benefit from the phenomenon. Namely, Berlusconi's camp is expected to receive a bonus of around 10 seats, mainly acquired by snatching them to the 5-Star Movement in single-member districts of Southern Italy. As discussed in our last post, these seats will probably benefit Berlusconi's party Forward Italy, even if gained thanks to the better performance of the Northern League and Brothers of Italy, who have made opposition to immigration a central element of their platforms. 

Overall, then, our fundamentals-based forecast predicts that the center-right coalition will easily get a plurality of seats in Montecitorio. Nonetheless, not even the most favorable scenario in our framework would entitle it to gain a majority of the Chamber of Deputies, which requires 316 seats. As for the other competitors, the Democratic Party is expected to rank behind the Five-Star Movement. However, it could actually earn a higher number of seats than Di Maio's camp thanks to two factors: the contribution of its coalition allies and its relative strength in some single-member districts. Consistent with the polls, then, our predictions depict a stalemate, which will hardly be broken by any type of bargaining across parties. 

Predicted vote shares according to the evolution of socioeconomic fundamentals (including intensity of refugee crisis at municipal level) and incumbency disadvantage.

Predicted seats according to the evolution of socioeconomic fundamentals (including intensity of refugee crisis at municipal level) and incumbency disadvantage.

Technical Note: compared to our previous posts, the total number of seats accruing to the center-right and center-left coalitions according to each exercise is somewhat smaller in this recap. This is due to a change in the way we aggregate the vote shares of smaller allies (Noi con l'Italia, +Europa, Civica Popolare, Insieme). Rather than rescaling them to the size of the "Other parties" group implied by our predictions, we are now including directly the percentages from YouTrend's super-mean of polls of February 16th. We believe that this is a more accurate approach. In fact, our model is intrinsically not capable of producing estimates for such parties (that were not running in 2013), and can only get an approximation of their total performance as a residual vote share. 

About the authors

Massimo Pulejo is a Pre-Doctoral Fellow at Bocconi University, Department of Policy Analysis and Public Management.

Piero Stanig is Assistant Professor of Political Science at Bocconi University, Department of Policy Analysis and Public Management, and a fellow of the Carlo F. Dondena Research Center. 

Thanks to Giovanni Da Fre' for excellent research assistance.

Forecasting the Italian Elections: Winners and Losers of the Refugee Crisis

As mentioned in our first post, immigration has certainly been one of the most salient issues in the Italian political debate of recent years. This is why, in our baseline prediction model for the legislative election of March 4th, we have included the relative, within-term change of the number of foreign-born residents in each Italian municipality.  However, this might not be enough to capture how voters will respond to this issue. In fact, such parameter only measures the variation in the number of legal immigrants, having a regular permit to reside on the Italian territory. On the other hand, the aspect that has dominated the electoral campaign is rather the substantial inflow of undocumented immigrants or refugees, and its impact on native people. 

To get a sense of this phenomenon, we exploit municipal-level data about SPRAR (Sistema di Protezione per Richiedenti Asilo e Rifugiati). SPRAR is a nationwide system, managed by the Ministry of the Interior, that provides accommodation and basic services to immigrants reaching Italy as refugees from at-risk countries. In a nutshell, it is based on the initiatives of local administrations, that get funds from the central government to finance the reception and care of asylum seekers. In particular, we have data about the number of SPRAR projects active in each municipality as of 2013 and January 2017. Moreover, for each active project, we know how many refugees it hosts at the time of observation. This allows us to track the development of SPRAR throughout the term, and use it as an additional predictor in our forecast. The inclusion of this information in the analysis induces the changes in votes, shares, and seats displayed in the table that opens this article. 

As one can easily spot, according to our estimates, the substantial increase in the number and size of SPRAR projects in the period under analysis is going to benefit those parties that have put the greatest emphasis on the migration issue during their campaigns. In particular, Matteo Salvini's Northern League and Giorgia Meloni's Brothers of Italy are set to earn a joint bonus of more than 150000 votes. However, it is interesting to note how a sizable portion of such bonus may consist of votes that would have otherwise accrued to the leading party in the coalition, Berlusconi's Forward Italy. As a result, the net gain for the center-right is pretty modest, and it is bound to be below one percentage point of the national vote. As to the other parties, a severe penalty is imposed by the electorate on the Five-Star Movement, whose position on this crucial issue has been thus far unclear, and that in 2013, the election we use to estimate the parameters of the model used for the forecast, did not emphasize the issue much. The table shows how such lack of clarity may persuade many of its supporters to either choose another party or to abstain. (We are at the same time aware that, being based on historical records rather than recent campaign position-taking, our model might miss any correction the Five-Star might have made to this disadvantage by acting tougher on immigration.) An increase of the abstention rate is in fact another important side effect of including the migration crisis into our framework. Namely, it may convince more than half a million additional citizens not to cast any ballot on March 4th. 

Mean of vote shares and seats allocated across 1000 elections simulated by our prediction model. Note that, as in our update of February 22nd, we are now summing up Democratic Party, +Europa, Civica Popolare and Insieme into the center-left coalition. For the latter three parties, whose performances cannot be estimated with our framework, we rely on YouTrend' s latest "super-mean" of polls, published on February 16th.

While the overall shift in terms of vote shares across parties is not dramatic, key insights come from the analysis about how the allocation of seats may change. In particular, even with a fairly limited increase in its vote share, the center-right would earn as many as 13 additional seats according to our exercise. Why is this the case? The core reason is that Berlusconi's coalition would gain from  the Five-Star Movement several single-member districts in close races in Southern Italy. This area of Italy, indeed, is at the same time the one with the toughest competition between the two camps and the one that has been most seriously hit by the inflow of refugees and undocumented migrants. 

With this additional bonus, the center-right would approach the threshold of 316 MPs needed to control the Chamber of Deputies. As discussed, this important step forward would entirely be the result of the better performance of the Northern League and Brothers of Italy. Ironically enough, however, the major beneficiary will probably be Berlusconi's Forward Italy, a perfect free rider in this situation. The reason is simple: most of the candidates in single-member districts of the center-right in Southern Italy are Forward Italy's members. As such, a shift of power towards the right part of the coalition in terms of consensus might instead result in a de facto shift in the opposite direction in terms of seats. 

About authors

Massimo Pulejo is a Pre-Doctoral Fellow at Bocconi University, Department of Policy Analysis and Public Management.
Piero Stanig is Assistant Professor of Political Science at Bocconi University, Department of Policy Analysis and Public Management, and a fellow of the Carlo F. Dondena Research Center. 

Thanks to Giovanni Da Fre' for excellent research assistance. 

Forecasting the Italian Election: The Impact of Small Parties

In our latest update, we have shown how our predictions are affected by the introduction of a "fatigue effect".  This effect would penalize the Democratic Party for having led the executive in the last five years. Specifically, in the context of our forecast, voters are expected to subtract about three percentage points to the Democratic Party compared to what socioeconomic fundamentals would suggest. Not a trivial loss for Matteo Renzi's camp, especially when translated into seats: 35 candidates would lose their place in Montecitorio to the advantage of their competitors. However, we have so far neglected the potential impact of small parties' performances on March 4th. In fact, we have gathered all these minor formations under a generic "Others" label, which is expected to gain as many as 76 seats in the upcoming race. It is now time to open this black box, and see how such seats are bound to alter the balance of strength in Italy's future Chamber of Deputies.

As previously specified, however, our forecast can only predict the vote shares of the 5 parties that have already run in the last legislative election, dating back to February 2013. Therefore, we have used the latest YouTrend's super-mean of polls to get an idea of how the remaining votes are likely to be shared across the different minor parties that form our "Others" group. Among them, the most competitive one is certainly Free and Equal, the splinter faction of the Democratic Party led by the former Speaker of the Senate, Pietro Grasso.  As of now, this is the only formation that is expected to overcome the threshold of 3% of vote shares that is necessary to take part to the allocation of seats assigned in multi-member districts.

Our prediction of vote shares and seats from February 21st, gathering all minor parties together under the "Others" label.

How seats change when the votes of allies are redistributed to the Center-Right coalition and to the Democratic Party.

All the other parties, some of which are additional allies of the Democratic Party (+Europa, Insieme, Civica Popolare) or the Center-Right (Noi con l'Italia), will probably see their votes and seats redistributed to their stronger allies. This will have a major impact on the overall size of Parliamentary delegations, shrinking the one of the Five-Star Movement to the benefit of the Center Right and, most important, the Democratic Party. A quick comparison of the tables above, in fact, gives a sense of how the Democratic Party would fully recover from the negative incumbency shock, and get as many as 162 seats.

A substantial part of this bonus would come from single-member districts, where the Democratic Party is set to regain18 seats thanks to the contribution of its allies. Interestingly, almost all such seats would be snatched from the Five-Star Movement in some highly contested constituencies of Central and Northern Italy, where Matteo Renzi's party used to have its strongholds. As for the Center-Right, an overall gain of 30 seats would project it closer to the absolute majority in our forecast, even though 37 seats still look like a difficult gap to close.

Finally, one important caveat: Italy's current electoral system makes it extremely hard to figure out who will actually be getting the seats in Montecitorio. In particular, a major source of uncertainty is the result of Emma Bonino's party, +Europa. Probably the strongest ally of the Democratic Party, it has been performing better and better in polls as the weeks passed, and it has a fairly high chance of actually overcoming the 3% threshold and gain some seats from multi-member districts. If this will turn out to be the case, 10 to 15 seats that are currently assigned to the Democratic Party in our prediction table would instead accrue to this formation, thus redrawing the balance of power both within the center-left coalition and in the Parliament as a whole.

About the authors

Massimo Pulejo is a Pre-Doctoral Fellow at Bocconi University, Department of Policy Analysis and Public Management.

Piero Stanig is Assistant Professor of Political Science at Bocconi University, Department of Policy Analysis and Public Management, and a fellow of the Carlo F. Dondena Research Center. 

Thanks to Giovanni Da Fre' for excellent research assistance.

Updated Forecasts of the Italian Legislative Election

As an update of our previous post, we ran some simulations including an additional "political" effect (voters' fatigue with the incumbent) and additional sources of uncertainty. While the simulations we reported last week were basically a rerun of the 2013 election, only with current economic conditions and immigration levels, and should therefore be considered a very rough baseline, these simulations are closer to a forecast of this election, as they take into account that the Prime Minister in office belongs to the Democratic Party.

Specifically, these plots display the results of new simulations that take into account the fact that usually the ruling party (i.e., the party of the Prime Minister) does worse in the election at the end of its tenure. We base the simulations on a plausible value of the magnitude of voters' fatigue with the incumbent, consistent with recent Italian legislative elections.

The red line is the mean prediction from our simulations, and the dashed lines identify the 95% credible interval.

Interestingly, this seems to be, to an extent, a case of the "why are campaign polls so variable if elections are so predictable" effect, documented in the U.S. case.

In particular, the Democratic Party polls seem to be converging to the prediction we make based on the economy and immigration, once the "fatigue" of voters with the incumbent is taken into account. If anything, maybe the Democratic Party, as of January, was polling a bit higher than the fundamentals and incumbency predict.

To an extent, the same is true also for the Five Star Movement, that was polling exceptionally well in the first half of 2017, but seems to be converging to the level we forecast based on the economy and immigration patterns.

On the other hand, the Center-Right coalition seems to be polling, as of January, quite higher than what one would otherwise expect based on fundamentals. This effect could be attributed to the return of Berlusconi, who, in 2013 (the election on which we base our forecast) was arguably quite less popular than he is now (and than he was prior to the crisis). Voters seem to have forgiven (or forgotten) Berlusconi's role in the debt crisis.

The mean of our predictions for the national vote share of the Center-Right coalition, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3, and standard deviation 6.7.

The mean of our predictions for the national vote share of the Democratic Party, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3 and standar deviation 6.7.

The mean of our predictions for the national vote share of the Five-Star Movement, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3 and standard deviation 6.7.

More technical information: The new forecasts are based on  simulations that include a (stochastic) swing reflecting a plausible "incumbency disadvantage". In general, at least since the beginning of the century, the party of the Prime Minister lost several percentage points over a term in office. This loss ranges from a maximum of almost 17 percentage points (People of Freedom, 2013 compared to 2008) to a minimum of 2.3 percentage points (Berlusconi's coalition, 2006 compared to 2001).

These simulations include a swing at the municipal level, normally distributed with mean -2.3 (the negative swing received by Berlusconi's camp between 2001 and 2006) and standard deviation 6.7 (the standard deviation, across municipalities, of the negative swing received by the Center-Left coalition between 2006 and 2008).

For each of the 1000 simulations of the election, we draw one incumbency shock for each municipality from the distribution and add it to the vote share of the Democratic Party. We then allocate the gain to the other three camps (Center-Right coalition, Five Star Movement, and "others") based on weights drawn from uniform distributions with expected value 1/3 and summing to one. In practice, this reflects our ignorance regarding who gains more from the incumbency disadvantage of the Democratic Party. We assume that, on average, the votes the Democratic Party loses due to this phenomenon benefit equally the Center-Right, the Five Star Movement, and other smaller parties (e.g., Free and Equal, the splinter party set up by a faction of the Democratic Party). At the same time, we incorporate in the simulations the uncertainty about how the gains are split: in each simulation and in each municipality the gains are split based on different draws of the weights.

Below are more conservative simulations, where we include an incumbency effect like that of 2006 vs 2001, including its (very large) observed standard deviation across municipalities (22 percentage points, as observed in the actual data in 2006). In these simulations, the confidence intervals are quite wider, and the polling average for the Five Star Movement and the Democratic Party fall within our 95% credible interval. The Center-Right coalition, though, is still polling higher than our forecasts.

The mean of our predictions for the national vote share of the Center-Right coalition, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3 and standard deviation 22. 

The mean of our predictions for the national vote share of the Democratic Party, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3 and standard deviation 22. 

The mean of our predictions for the national vote share of the Five-Star Movement, plotted against the mean of surveys in each month going from January 2017 to January 2018. Dashed lines are 95% Bayesian credible intervals. The swing attributed to the incumbency disadvantage in each municipality is drawn from a normal distribution with mean -2.3 and standard deviation 22. 

About the authors

Massimo Pulejo is a Pre-Doctoral Fellow at Bocconi University, Department of Policy Analysis and Public Management.

Piero Stanig is Assistant Professor of Political Science at Bocconi University, Department of Policy Analysis and Public Management, and a fellow of the Carlo F. Dondena Research Center. 

Thanks to Giovanni Da Fre' for excellent research assistance.

A Fundamentals-Based Forecast of the 2018 Italian Elections

A comparison of our predictions with the mean of surveys conducted in Italy during the first week of February 2018.

The table above presents preliminary results from our project on 2018 Italian Parliamentary Elections. The aim of the project is to come up with a fundamentals-based forecast of the vote shares and resulting seats' allocation for the main parties and coalitions running on March 4th. By fundamentals-based, we mean that our prediction is grounded on what Italian voters are expected to do based on the recent dynamics of the two most debated issues: the migration crisis and the performance of the economy. To come up with our estimates, we build a bread-and-peace type model. In our framework, the vote shares of the main parties at the level of 7543 municipalities are the result of two fundamental elements: a baseline expected vote share plus a variation around the baseline, depending on a relatively small set of socioeconomic factors.

The mean of our predictions for the national vote share of the center-right  coalition, plotted against the mean of surveys in each month going from January 2017 to January 2018.

The baseline is jointly determined by the past vote share earned by party p in municipality i and a province fixed effect, aimed at capturing persistent geographical ideologies across the Italian territory. It is important to note that, upon observing the electoral dynamics in Italy during the last 25 years, we have decided to use electoral results in t-2, that is, two races before the one we are predicting, rather than in t-1. Therefore, in the case of 2018, the relevant shares included in the model are those obtained by parties in the elections of 2008. Once this starting point is set, our model predicts that a deviation will take place, based on how the state of the local economy and the migration phenomenon have shaped voters' preferences throughout the term. Specifically, in order to gauge the effect of possible occupational shocks  at the municipal level, we source information about the share of enterprises and jobs in the municipality that used to accrue to the manufatcuring sector as of 2011. Then, we interact these measures of the importance of manufacturing with the relative change in the unemployment rate at the provincial level during the last term preceding the election. This allows us to compute a measure of vulnerability to occupational shocks that varies across both terms and municipalities. With a similar spirit, and in order to account for the preferences of younger voters, we interact the share of people in the 18-29 age bracket at the municipal level with the provincial change in the inactivity rate for people of that age. As far as immigration is concerned, our proxy for the magnitude of this phenomenon is the relative variation of the share of foreign-born residents in the municipal population, again computed on a term basis. The equation is completed by controlling for the natural logarithm of the municipal population, and estimated via seemingly unrelated regression (Zellner 1962). This technique has indeed proved to be particularly suitable for multiparty data (Tomz et al. 2002), as it estimates all the equations for each party simultaneously. As a result, it takes into account that if party j is getting x% more votes, then they must be either sourced from other parties' electorates or from people who did not turn out in the previous election.

The mean of our predictions for the national vote share of the Democratic Party (center-left), plotted against the mean of surveys in each month going from January 2017 to January 2018.

Rather than coming up with a single prediction of the electoral performances of parties in each municipality, we follow King et al. (2000) and implement 1000 simulations of the predicted vote shares determined by our model. This maximizes the use of the information at our disposal, while exploiting the intrinsic uncertainty of our framework to produce estimates of Bayesian credibility intervals at the party-municipality level. The process leads to the results presented in the table above and in the plots. In a subsequent stage, by taking advantage of the extremely fine-grained nature of our data, we apply the electoral formula currently in force and aggregate votes at the level of electoral districts. This is done by weighting each municipality by the predicted number of people that will turn out, thus transforming vote shares into actual numbers of votes, and summing them up across all the municipalities in district j. The total number of votes provides us with a ranking of parties and coalitions in each of 231 single-member district. Following the electoral rules, we assign the seat to the first-ranked party or coalition of parties. The remaining seats are instead distributed by looking at the shares that parties are expected to gain at the national level. In fact, even if 387 multi-member districts have been designed, electoral prescriptions establish that they will matter only to determine where parties are going to earn their seats, but not how many of such seats they will be eventually entitled to get.

The mean of our predictions for the national vote share of the Five-Star Movement, plotted against the mean of surveys in each month going from January 2017 to January 2018.

In order to be used to predict the results of March 2018, the model had first been trained with an out-of-sample prediction of the results of the last Italian parliamentary election, which took place on February 2013. This means that we have adapted it to that election, and estimated the coefficients of each of our predictors on a pool of 5753 randomly drawn municipalities, which constituted our training set. After getting an estimate of how our variables had contributed to determine actual vote shares in those municipalities, we have used the coefficients to predict the results in the remaining 2005 units of observation, which formed our test set. The results, displayed in the table below, are quite encouraging.

The performance of our model in forecasting out-of-sample results for 2013 in 2005 Italian municipalities.

                            

Namely, the percentage-point differences between predicted and actual vote shares always lie in the 0.24-2.8 range. Moreover, Root Mean Squared Errors at the level of individual municipalities are remakably low, with the partial exception of those relative to the performance of the Five-Star Movement. However, this is probably due to the fact that we cannot include a baseline vote share for this party, as it was not on the political arena on 2006, which constitutes our reference election year for the prediction at stake. An identical consideration applies to the predictions for 2018, which use 2008 as the baseline election. In spite of these issues, however, the ranking of coalitions implied by our exercise mirrors the one depicted by surveys. Furthermore, the discrepancy between our forecast and the vote share computed by the most recent polls is quite limited. Notably, as one can easily see from the plots, most of the gap between our figures and those of surveys has opened up only in recent months. This is by no means surprising. In fact, as elections approach, survey interviews can obviously measure preferences for otherwise unobservable factors such as leaders' valence, candidates' qualities, or the effect of shocks that have hit one or more parties. For instance, it is extremely likely that the marked decrease in the vote share for the Democratic Party predicted by polls conducted in the last 6 months is reflecting the split that occurred within it, and eventually led to the formation of the leftist party Liberi e Uguali. Nonetheless, it is interesting to see how our framework, without resorting to any interview, is capable of producing credible estimates of a future electoral race based on a few, carefully chosen predictors.

About authors

Massimo Pulejo is a Pre-Doctoral Fellow at Bocconi University, Department of Policy Analysis and Public Management.
Piero Stanig is Assistant Professor of Political Science at Bocconi University, Department of Policy Analysis and Public Management, and a fellow of the Carlo F. Dondena Research Center. 

Thanks to Giovanni Da Fre' for excellent research assistance.