Comparing final polls to election results

I found the last polling results from the 2002, 2003, and 2006 elections, and couldn’t resist studying them a bit better and comparing them to the eventual election outcome. Two trends emerged: one that I more-or-less expected and one that was a complete surprise.

The expected trend was that the polls did an excellent job of capturing the primary theme of an election cycle, but that a secondary, less important but still quite unexpected, theme surfaced in the actual elections.

The surprising trend was that relative to the last polls, the right block as a whole won three seats; two from the left and one from the christians. This happened both in 2003 and 2006, while in 2002 this last-minute shift was exactly twice as large.

Methodology

I took the results of the individual pollsters and entered them in my current poll app, which yielded the weighted average I use. The tables below compare this weighted average to the election results.

2002

2002 was the year of the Fortuyn revolt and his tragic murder. The last polls (which I think were taken after his murder) correctly predicted the size of the LPF and the deep, deep fall of the Purple parties. These were the main themes of the 2002 cycle.

1998 2002 final polls 2002 elections
CDA 29 +5 34 +9 43
LPF - +26 26 - 26
VVD 38 -12 26 -2 24
PvdA 45 -21 24 -1 23
GL 11 +2 13 -3 10
SP 5 +3 8 +1 9
D66 14 -6 8 -1 7
CU 5 +1 6 -2 4
SGP 3 -1 2 - 2
LN - +3 3 -1 2
Blocks
Right 67 +22 89 +6 95
Left 75 -22 53 -4 49
Christian 8 - 8 -2 6

What the polls did not predict was the size of the CDA victory. Initially the CDA gains in the polls were moderate, but they surged up to an unexpected size of 43 seats in the elections.

This can be explained by the fear that was caused by Fortuyn’s murder and the subsequent flight to the safe centre. That was what brought the CDA back to power after the dark and dismal Purple years.

Note the dramatic loss of GL: this was caused by the fact that Fortuyn was murdered by an animal rights activist, a theme GL owned at that time. GL party leader Rosenmüler was even threatened just before the elections, and GL’s slight gain melted away to a slight loss.

2003

After the failure of Balkenende I (CDA+VVD+LPF) the major theme of the 2003 elections was the flight back from LPF to PvdA, and it pretty soon became a prime-minister race between Bos (PvdA) and Balkenende (CDA).

2002 2003 final polls 2003 elections
CDA 43 -1 42 +2 44
PvdA 23 +20 43 -1 42
VVD 24 +5 29 -1 28
SP 9 +2 11 -2 9
GL 10 -3 7 +1 8
LPF 26 -20 6 +2 8
D66 7 -1 6 - 6
CU 4 - 4 -1 3
SGP 2 - 2 - 2
LN 2 -2 0 - 0
PvdD - - 0 - 0
Blocks
Right 95 -18 77 +3 80
Left 49 +18 67 -2 65
Christian 6 - 6 -1 5

However, all polls from the last week or so pointed to a slight but clear PvdA advantage. When the actual results came in (and I remember my surprise) the CDA turned out to have won slightly, and the PvdA lost slightly.

The reason for this last-minute change has general validity; also for the current cycle.

A few voters who’d planned to vote strategically for the PvdA in order to prevent the CDA from becoming the largest party decided to switch to their true preference when it seemed the PvdA remained in front of the CDA.

In other words, strategic voters may switch back to their true choice when it seems that their strategic vote is not necessary. That could conceivably happen to the VVD in the current elections, although it’s not easy to see where these votes would go to. CDA? PVV? Or maybe even D66?

Besides, there’s no neck-and-neck race between VVD and PvdA. So on balance it’s not very likely that the VVD will be crippled by the same effect.

2006

The 2006 elections are hard to characterise; they didn’t have one outstanding theme like the previous two. However, the most important political story was the dramatic rise of the SP, which combined protest voters with left-wing voters disappointed by the PvdA.

2003 2006 final polls 2006 elections
CDA 44 -3 41 - 41
PvdA 42 -6 36 -3 33
SP 9 +16 25 - 25
VVD 28 -6 22 - 22
PVV - +5 5 +4 9
GL 8 -1 7 - 7
CU 3 +4 7 -1 6
D66 6 -4 2 +1 3
SGP 2 - 2 - 2
PvdD 0 +2 2 - 2
LPF 8 -8 0 - 0
EenNL - +1 1 -1 0
Blocks
Right 80 -11 69 +3 72
Left 65 +5 70 -2 68
Christian 5 +4 9 -1 8

Although it was clear that the extreme right would not disappear from parliament, it was widely held that the populist vote would be divided under three, maybe even four parties. Two of them were eliminated even before the final polls, but until the last moment it seemed two parties would make the jump: Wilders’s PVV and EenNL, which was mainly a vehicle for Rotterdam-based Fortuyn backers from the circles of Leefbaar Rotterdam, which is still the second party locally.

This turned out not to be the case: not only did EenNL narrowly miss its seat (it lacked about 2,700 votes on a total of about 65,000 for a seat), but the PVV won not the expected five (or six, with EenNL’s one) but nine.

Wilders’s success can be explained by strategic voting: extreme right voters wanted to make as large an impact as possible, and therefore voters of the small three parties switched to Wilders.

But he still won more seats than expected even of the total extreme right; and that’s something that also happened to the LPF in 2003, though not in 2002. Is this a new rule? Do extreme right voters tend to hide their vote from pollsters? If so Wilders might win more than the 17 seats the polls give him (and the VVD would lose some).

Last-minute left->right shift

What came as a total surprise to me was the last-minute shift to the right relative to the polls that took place in all three elections. In the last years I read quite a few Dutch political science articles and I haven’t found a trace of a mention of any such effect, let alone an explanation. (If I missed something, please leave a comment.)

What I did read recently (can’t remember where) is that turnout among right-wing voters is generally a little bit better than among left-wing voters. The article linked this to left-wing voters generally being poorer, which could be true for all I know.

Still, I’m afraid we have to accept this effect. So I predict that the right will win three more seats relative to the last polls, while the left will lose two and the CU one.

Note: this is relative to the last polls, which will probably appear on the Monday or Tuesday before the elections.

<— Even more coalition news | Small fry, 31 May —>

This is the political blog of Peter-Paul Koch, mobile platform strategist, consultant, and trainer, in Amsterdam. It’s a hobby blog where he follows Dutch politics for the benefit of those twelve foreigners that are interested in such matters, as well as his Dutch readers.

Archives:

Comments (closed)