3/24/06 Slovak Election Update: Voting and Turnout

Before I begin, a note about file format.  Several readers have requested a “to go” version
and so a full .pdf file containing this post is available here:  Download 3_24_06_slovak_election_update.pdf

Second, a note about background.  For your information I have attached a briefing
document prepared by Tim Haughton and myself during the summer of 2005.  Much has changed since then but the document
offers a brief visual primer of Slovakia’s
political scene.  It is available here: Download slovak_politics_primer_2005.ppt

Now to the heart of the matter.

Voting and Turnout
Last week’s post addresses recent trends in polling
according to party and bloc and the way those poll numbers would translate into
party seats. However it also begins to
address the problem that polling numbers have not automatically translated into
seats: some parties have gotten more votes than polling would suggest while
others have gotten less. This week, I
attempt to take what little data is publicly available and attempt to figure
out what current polling numbers suggest for the final tally of votes.

Polls, Votes and

There are a variety of ways in which we can attempt to
predict actual votes on the basis of polls. The most direct, though not necessarily the most accurate, is to use
past elections as a baseline. Because of
the significant change in Slovakia’s party system in the past six years, and
the unusual configurations of coalitions in elections to regional councils,
only two elections—the 2002 Parliamentary Election and the 2004 European
Parliament Elections—are plausible objects of comparision. Each of these have their problems,
furthermore, since the rapid change of Slovakia’s political scene makes 2002
ancient history (the parties Smer and ANO were relatively new at the time) and
since the very low—17%—turnouts of
the 2004 Europarliament election give that election a rather different
character. Nevertheless, this is the
only electoral data we have to work with and so like the drunk under the
lamppost, I will look for keys where there is the most light. The table below compares the results of
elections in 2002 and 2004 with the results of public opinion polls from major
polling firms that immediately preceded the elections:


The results show better-than-expected performance in black
and worse-than-expected performance in white.  Some parties consistently received more votes
than polls would suggest (SDKU, KDH, SMK) while others consistently received
less (Smer, HZD, SNS), while for others the polls tended to produce fairly
accurate results (HZDS, SF) and another was uneven (KSS). It would appear that despite their stated
preferences, voters from some parties are less likely actually to go and vote
(polling evidence suggest that last-minute vote switching does occur but that
it neither has a systematic bias in favor of any party, nor is it predictable,
so it would be difficult to build it into a model in any case). From these election baselines, it is possible
to construct a variety of models for actual performance..  For the purposes of this post, I have created
a simple model that takes poll results as the baseline for 100% an election
with turnout and then calculates the drop in turnout for each party necessary
to produce the actual results with the actual turnout (69.1% in 2002, 16.6% in
2004).  By these calculations, every 1%
drop in overall turnout in 2002 translated into a loss of nearly 1.5% of turnout
among Smer supporters but only a 0.8% drop of turnout among KDH supporters and no
drop at all among SDKU supporters. Because
of the very low overall turnout rates for 2004, the drop-per-percentage is
actually smaller but with pretty much the same patterns. To draw out the implications of these
calculations, the graphs below extrapolates from the most recent polls to estimate
levels of actual electoral support for all parties at any given level of overall
turnout between 50% and 100%. To read
the graph, simply guess the level of turnout for 2006 and trace upward from
that number to see what share of the vote your party is likely to get. Because the 2002 and 2004 models produce
estimates that are different in degree (though not in direction), I have
included both models here as well as a “combined” model that averages the two. 

Notable here is the weakness Smer in all three models surveys:
at rates of turnout comparable to 2002 its current 33% estimate translates into
a range no higher than 30% and as low as 23% depending on the model. By the same standards, SDKU’s current 11%
looks more like 12% to 16%. 





Polls, Votes and Blocs

Because the parties of the current coalition tend to benefit
from lower turnout, the combined effect for the entire coalition is even
greater, and the three graphs below show the overall effect for ideological
blocs, again according to the three models. I use the “bloc” notion as shorthand for the time being and do not
suggest that parties within these blocs are more likely to seek each other out
as future coalition partners (though their voters do seem to cluster together
in patterns something like these, but that is a story for next week).

In the most extreme of the models, any turnout lower than 80%
actually produces a numerical advantage for Right (SDKU, KDH, SF, ANO) over
Left (SMER, KSS) with the Slovak (HZDS, SNS) and Hungarian parties relatively
unchanged.  It is noteworthy, however,
that even in this fairly extreme model the current coalition and its offshoots would
require an exceptionally low turnout–below 60%–to have a chance at an
electoral majority.





Polls, Votes and Seats

The graph below translates these vote estimates into seat
estimates for the moderate “combined” model but I’d be happy to supply the
others to anyone who wants to see them.


The Question of

Not only are the estimates in the graphs above fairly crude
and based on only one source of information, but they are dependent on a second
set of estimates regarding turnout.  This,
too, is an highly inexact science, but it is possible to make some rough
assessments.  The first graph below
suggests that on similar polls conducted by UVVM in 2002 and 2006, the
percentage of the population that is not interested in voting in 2006 exceeds
the percentage in 2002 by an average of nearly 6 points.  As the second graph suggests, however, the
share of undecided voters is actually smaller than in either of the two
previous election cycles.  The third
graph below suggests that the sum of survey respondents who had decided not to
vote or had not decided on a party was an excellent predictor of turnout in
2002, but that this measure significantly underpredicted turnout in 1998.  In either case, however, the total number of
undecideds and non-voters for 2006 has exceeded those for 2002 by approximately
2%.  Using 2002 as a baseline would therefore
suggest a turnout of around 67%. This is
not a very well-grounded estimate but it as good as the available public
resources allow. 




What Kind of

Using the most recent polling data along with a turnout of 67% and the “combined” model above–a "best guess" given the information available–would yield parliament that looks like the one below.  While this does not tell us directly about
probable governments, it does tell us that the options for government formation
are manifold. There is virtually no
possibility for any ideological bloc to form a government on its own, and limited
chance that any two blocs could form
a government (the fairly ungainly combinations of Left-Slovak [Smer+KSS+SNS+HZDS]
or Left-Right [Smer+KDH+SDKU+SF], suggesting the strong possibility of a combination
of parties within three ideological blocs (Slovak-Left-Right or
Left-Right-Hungarian) or a minority coalition that involves parties two blocs
with a party from the third other as a silent partner. Data does allow for closer scrutiny of
the cohesion and electoral chances of particular coalition combinations and these will be the subject of the next post.


3/17/06 Slovakia Election Update: Parties and Polls

From now until June I hope to post regular (with luck, weekly) analyses of recent developments in Slovak public opinion as they pertain to the upcoming elections. 

Notes and acknowledgements. 
The analyses I post here would be impossible without the many insights into Slovak public opinion that I have gained from working with Vladimir Krivy, Olga Gyarfasova, Zora Butorova, Jan Luha, Ivan Dianiska, Karen Henderson and Tim Haughton and many others.  The blog on which you read this is a gift from John Gridley and allows me to replace the ordinary web pages I constructed in 2002 for the same purpose with this  more accessible and interactive medium.  If you use Mozilla Firefox and Thunderbird you can set the live feed to inform you when I have posted new pages.  If not, there are ways to make Microsoft Explorer and Outlook do the same thing.   If  these methods are not to your tastes, send an email to dividedsocieties@centrum.sk (or the email address listed on the graphs below) that contains the word "Slovak" and I’ll add you to a "please update" list for the next few months as I add new pages.  In general I welcome feedback.  Let me know what you disagree with and what kinds of analyses you would like to see.

Party Support
I attach below four preliminary graphs to give a sense of change and continuity in Slovak public opinion since the last election.  These begin just over three years ago (though I have data going back for nearly a decade and a half and would be glad to share this for any who would like to see it) and unless otherwise noted, are compiled as averages of the results of monthly public opinion surveys conducted by UVVM, Focus and Markant as published in Slovak newspapers and available through other sources.  Since each firm conducts and announces its surveys with different intervals, the monthly average is a bit arbitrary, but it allows for a cleaner presentation of the data without undue oversimplification.

All Parties
Attached below is the three year track for every major party.  The most notable feature is the size of Smer’s lead over other parties–nearly three time as many voters as the next largest competitor.  Not since the predominance of HZDS between 1992 and 1997 has Slovakia seen a gap of similar magnitude.  Two other factors are noteworthy: first, while Smer’s lead has grown, its level of support has remained relatively stagnant for at least the last six months; second, the size of Smer’s lead has increased in direct proportion to the decline of HZDS from clear second place to a level almost indistinguishable from five other parties (and two sets of surveys in late 2005 showed it in third place which is, I think, unprecedented in its existence). 


Parties in the Middle
The clustering of six major parties around the 10% mark makes it difficult to see trends among these parties with any clarity, so it is necessary to use the graph below which shows only the narrow band between 5% and 15% and draws a trend line as a moving average of the previous five months to smooth out monthly variations.  Several trends reveal themselves.  First,  the drop in HZDS support is almost perfectly complemented by the rise in SNS support.  Likewise,  the data shows an almost identical pattern (much more rapid, but with a 3 month lag) in replacement of ANO by SF.  As above, the data is not enough to demonstrate that the voters leaving one are the same who gravitate to the other, but the patterns suggest at least an overall stability.  Amid these major shifts, KDH, SDKU and especially SMK have remained quite stable over time, with only tiny variations, especially since mid-2004.  Whether the recent coalition split will change this is unclear but the two post-split polls show very little overall change.  Finally, KSS has also remained relatively stable since mid-2004 as well (hovering just above 5% after dropping from the 8% level  in late 2003), but four of eight recent polls show the party’s supporter under the 5% threshold.  Whether this represents the actual level of support is an open question (almost no polls in 2002 showed the party above the 5% threshold and yet it managed a result of 6.3%) and will hopefully be the subject of future posts.


"Bloc" Voting
With the large number of parties and the frequent shifts in support–including the emergence and disappearance of relatively significant parties–it is easy to overlook certain markers of underlying stability in Slovakia’s electorate.  This is not to minimize the raw effects of party changes, but it is noteworthy that the voting within recognizable ideological blocs remains quite stable.  The use of "bloc" here, however, implies only that the parties grouped together share some degree of ideological affinity, but not that these are the only kinds of affnities or that it is these affinities that will shape future coalitions (a question for a future post).  Indeed Smer is unlikely to form a coalition with KSS, and the most significant recent political conflicts in Slovakia have occurred within the bloc that I label here as "right" (a reference to economics rather than culture as the "right" bloc spans the full spectrum from libertarian to orthodox Catholic).  Despite considerable fludity of its party membership "right" bloc has remained between 25 and 30% for the past two years.  The "Slovak" bloc (by which I mean parties that think primarily about Slovak national questions) has likewise remained stable between 20% and 25%.   Even the "left," the most volatile bloc, has remained mostly within a window between 32% and 40%.


Translating Votes into Seats
Slovakia’s parliamentary deputies are elected from a single, nation-wide district with a highly proportional electoral system, so the percentages of seats are almost always in direct correspondence to the percentages of votes.  Nevertheless, it is often helpful to estimate the number of seats that a party would get if it received the share of votes indicated in a survey.  The graph below gives a monthly tally of seat totals.  To the extent that party support has remained relatively stable, the graph stays relatively stable as well, with the biggest effects emerging when parties fall below or climb above the 5% electoral threshold, as happens with ANO (falling below consistently after mid-2005), SF (climbing above consistently after mid-2005), KSS (falling below intermittently), SNS (falling below intermittently before  2005) and HZD ( climbing briefly above  in spring of 2004).  The linear arrangement of parties in this graph is highly limiting, however, and does not do a good job of revealing all possible majority coalitions.  A better visual representation of these possibilities is a project for future posts.


(complete copies of these graphs are also available here
Download slovak_election_data_2006_week_13.pdf


Translating Preferences into Votes
As the 2002 Parliamentary election and the 2004 Europarliament elections demonstrated, preferences expressed in public opinion polls may not accurately reflect the actual votes received by a party.  I hope to take this question up in future posts, but it is worth noting that the difference arrises both from polling methods (in which some firms perform better than others) and in the degree to which respondents who profess a preference for a party are actually willing to vote for that party.  These likelihoods appear fairly stable and future posts will attempt to create a model for estimating actual votes based on stated preferences.  For the moment, however, it may be helpful simply to begin with a brief overview of the accuracy of polling duing the 2002 election.


Subdividing the data by individual polling firms yields the following results:

In 2002 there was a systematic tendency to underestimate support for SDKU and (to a much smaller extent) KSS and to overestimate support for Smer (and to a lesser extent) HZD.  We need more data to determine whether this is still true, but the results of polls before the 2004 Europarliament elections showed a nearly identical (though even more extreme) relationship to final results.  More on this later.

I look forward to reading your posted comments or reading your emails (use dividedsocieties@centrum.sk or the address listed in the graphs).  Thanks.