Graphs below. Click on the links to show trend-lines for different parties.

PDL | PD | M5S | Lega | SEL | Scelta Civica

Infrequently asked questions

Where do these polling figures come from? Here and here.

What are these trend lines? They’re estimates from a model which treats latent party support as something which evolves smoothly over time, and is made manifest through particular potentially biased polling snapshots.

Why are some trend lines way above the polls? Evidence from the last elections showed polling companies consistently under/over-estimated some parties. These biases are included in the model.

tl:dr version: a standard model of cabinet duration predicts the Letta government should last about 400 days.

Enrico Letta’s government was supposed to be a governo di scopo — a government that would take a number of targeted measures to reduce the cost of politics and increase the ease of governing.

Letta’s address to the Camera yesterday promised much more than that. This was true even for matters of institutional and electoral reform: the bicameral commission on constitutional reform — the fourth in thirty years — would take eighteen months (or rather, Letta would step in and possibly dissolve parliament if it hadn’t completed its work by then).

Given the ambitious nature of Letta’s announced programme, it’s fair to ask: how long might such a government last?

I’ve made a number of predictions about government duration in the past. All of these predictions have been based on statistical models, and some of them have performed reasonably well (Berlusconi). Here are some of the time-varying factors we can include in the model, which is taken from Strøm, Kaare; Müller, Wolfgang C. and Bergman, Torbjörn, eds. (2008). Cabinets and Coalition Bargaining: the Democratic Life Cycle in Western Europe:

Majority government:

the cabinet has a clear majority, so that works in its favour;

Minimal winning:

the cabinet is not minimal winning, because it includes a surplus party (Scelta Civica), which does make it…

Minimal connected:

the cabinet is ideologically connected (i.e., there are no gaps when the parties are lined up left to right)

Maximum Banzhaf score:

the cabinet includes the party with the largest Banzhaf score

Coalition:

it’s a coalition, rather than a single-party government, which counts against it

Eff. num. parl parties:

here, I’ve taken the effective number of parties in Camera, at 3.51 (but you could take the number in the Senate, or 3.92)

Conservative:

conservative parties don’t have a majority (conservative governments are longer-lived)

Add to this a number of factors which are constant (Opposition influence, positive parliamentarism, cabinet unanimity, PM dissolution powers (or lack thereof), bicameralism, lack of semipresidentialism), and what kind of prediction do we get?

The figure below plots the predicted probability of survival for days after formation. The probability dips below 0.5 after 396 days — which means that we should predict Letta’s government falling at the end of May next year. If we bear in mind that the 2014 European Parliament elections will, shortly after that point, provide a useful signal of parties’ electoral support, we might expect the government to fall after that point.

Note that the definition of cabinet termination used in the academic literature includes changes of PM and changes of party composition. A change of parties in the government — due to a PD split or Scelta Civica withdrawing — means a `new’ cabinet. Note also that the model doesn’t tell us anything about the causes of termination — the approach is to assume that causes come along (`events, dear boy, events’) and structural attributes modify governments’ ability to weather them.

R source code available as a GitHub Gist.

Update: This Oxford Analytica talk with David Hine gives slightly different estimates: 50% probability lasting until autumn, and only 20 to 25% chance of lasting until next spring. The reasoning is that Berlusconi will pull the plug as soon as he thinks that the gap between PDL and PD in the polls is sufficient to win a majority.

One of the more benign consequences of having being invaded by Napoleon lies in generally having very good administrative records in certain areas. This is certainly true of the Italian parliament, which has a wonderful collection of linked data available to download.

I’ve spent too great a part of my day getting to grips with querying the SPARQL endpoint in order to find out about the ages of Italian legislators over time. The effort spent on learning SPARQL somewhat exceeds the reward in terms of the graph below, but still, blah blah blah transferable skills blah blah blah semantic web blah.

Anyway, here’s a plot which shows the mean and median ages of legislators from the constituent assembly to the 16th Legislature (the one which recently ended rather than the one which was recently elected), upon the start of their mandate.
Continue Reading »

If the Letta government is confirmed next week, it will have to reform the electoral system, either by reforming the electoral law, or by amending the Constitution to end Italy’s perfect bicameralism, whereby the government must have the confidence of both the Chamber and the Senate.

The PD’s current proposal for electoral reform, presented at the end of last month, is a resubmission of the proposal they made during the last parliamentary term. It proposes the following for the Camera:

• 433 single-member districts, elected in two rounds if no candidate wins 50% in the first round, with all candidates polling over 10% progressing to the second round
• 173 seats allocated using party-list proportional representation using the existing 26 circoscrizionali.
• 12 seats allocated in a national constituency using the vote totals for party lists
• A strong link between the two tiers: voters will only have one vote, which will be valid for both single-member and proportional tiers
• Compensatory allocation of seats in the proportional tier: votes cast for a party in the single-member districts will be subtracted from their total before seats are allocated in the proportional tier.

For the Senate, the proposal is similar, except that 216 seats are elected in single member districts, and the remaining 93 seats are allocated using party-list proportional representation in the regions.

This is similar to the system used between 1994 and 2001, in that it is a mixed-member proportional system.

It is different, and more majoritarian in its likely effects, in that (a) it is harder for parties to avoid their votes being subtracted in the proportional element (there are no problems with the scorporo), and (b) the allocation of seats in 26 circoscrizioni creates a very high effective threshold.

It is different, and less majoritarian in its likely effects, in that (a) the two round system allows small parties to either try their luck in the first round, and possibly sneak through to the second round, or to negotiate stand-down pacts with other bigger parties and secure seats that way.

If, as Napolitano’s wise men have suggested, the size of both chambers must be reduced, then this system will become even more majoritarian in its likely effects.

Of course, just because the PD has proposed this reform does not mean that any eventual reform will bear any relationship to this proposal.

Suppose the Letta government receives a vote of confidence next Monday. That will mean that the coalition took 63 days to form, counting from the day of the election. Is that more than might reasonably have been expected?

If we use Sona Golder’s data, then we can say that it’s certainly more than the average for Western Europe (23 days), and more than the average for Italy (29 days). It’s certainly much more than the average for post-electoral formation exercises in Western Europe (12 days).

One response to that is to say that this time round was different. We can use Golder’s Cox survival model to make a prediction based on specific features of Italy in 2013. The main features are:

• the number of legislative parties: here, I’ve taken the effective number of parties in the Senate (4.29)
• the degree of polarization: here, lacking a left-right position for the M5S, I’ve taken the mean for Italy
• positive parliamentarism: governments must face an investiture vote
• continuation rules (governments can continue in office during electoral periods): Golder has Italy as lacking a continuation rule, strangely, but I’ve followed her usage
• majority party: no majority party here.

With this set of features (and ignoring interactions with time necessary to maintain the proportional hazards assumption), we’d predict that the median time to formation would be 33 days, with 95% confidence interval of [30,40].

So government formation this time round was almost 100% longer than we would have predicted. Why?

Two obvious suggestions: first, a President had to be elected. Parties had to negotiate on two levels. Second, there was a new anti-system party. Ascertaining that party’s intentions took time.

R code here.