Chris Hanretty

The fall of the Prodi government was disappointing, but entirely predictable. The aim of lasting for the entire legislative term (5 years) was unrealistic, with less than a 2% chance of success. Conversely, it was 95% certain that the government would fall sometime between September 2007 and May 2009; the most likely date was sometime in June 2008.

Spurious certainty, probably, but hear me out. King et al. (1990) give a reasonably simple exponential model of cabinet duration as a function of seven variables:

1. formal investiture requirement
2. party system fractionalization
3. party system polarization
4. majority versus minority government
5. number of formation attempts
6. whether the government was formed after an election or not
7. whether the government is a caretaker government or not

What values can we plug in for the Prodi government, and what result do we get?

1. Formal investiture requirement: true, so a value of 1
2. Party system fractionalization: the Rae index of fractionalization taking the party groups in the Senate immediately following the election is 850.8 (ignoring senatori a vita)
3. Party system polarization: the percentage of votes received by ‘extremist parties’, following the definition used by Powell (1981) is 18.63, most of which is taken up by Rif.Comm, PdCI, Verdi, Lega Nord + various separatist parties, FT, and others
4. Majority government status, so a value of 1
5. Number of formation attempts was 1
6. The government was formed after an election, so a value of 1
7. The government wasn’t a caretaker government, so a value of 1

Plugging these values into the King et al. model, one gets a predicted duration in months of 25.9 - that is, the Prodi government was predicted to last until June 2008. Thus, the actual fall of the Prodi government was not that far from what could have been predicted the day after the election. I’m sure no-one would begrudge Prodi for not being able to hold on to the coalition for a few more months, especially given that the fractionalization of the party system increased due to a remarkable rate of party fission in the Senate.
The data used in the King model is available here; the R code used to generate these predictions was:

coal<-read.table("coalcold.asc",

header=F,

col.names=c("DURAT","BELGIUM","CANADA","DENMARK","FINLAND","FRANCE","ICELAND",

"IRELAND","ISRAEL","ITALY","NETHER","NORWAY","PORTUG","SPAIN","SWEDEN","UK",

"T","POPINFL","OPPINFL","ROPINFL","POX","IDENT","VOLAT","RESPONSE","INVEST",

"POLAR","FRACT","NUMST2","PARLBAS","CRISIS","FORMAT","OPPCONC","ELTIME2",

"ELTIMEB","ELTIMEN","CARETK2","NONPART","ELECTC","ELECTM","SINGPR","EXTSUP",

"SALIEN","OPPCON2","OPPCOND","NP","DECPROX","CIEP12","CIEP24","CIEPTW"),skip=1)

z.out<-zelig(Surv(DURAT,CIEPTW)~INVEST+FRACT+POLAR+NUMST2+FORMAT+ELTIME2+CARETK2,model="exp",data=coal)
x.out<-setx(z.out,INVEST=1,NUMST2=1,ELTIME2=1,CARETK2=0,FRACT=fract2,FORMAT=1,POLAR=18.63)
s.out<-sim(z.out,x.out)
summary(s.out)