Dashboard News: Slovak papers see the light

It is absolutely delightful for me to see that after years of publishing one poll after another without much analysis and presenting each difference as “news”, two major papers in Slovakia have begun to present long-term time-series data of polls.  I suspect they will still present each new poll as “news” (though congratulations to them also for doing much more comparison and acknowledgment) but now readers can at least do some comparing on their own.  A few brief thoughts on the two attempts

  • Of the two the less ambitious effort is Pravda’s which is simply a story link containing tables of results from the major polls from January until the present.  It does the job but without much visual impact.  The full chart can be found here http://spravy.pravda.sk/prieskumy-agentur-focus-median-a-mvk-dvb-/sk-volby.asp?c=A100421_160242_sk-volby_p12.  Since this is merely a story page it is not clear whether Pravda intends this as an ongoing feature or merely as a one-off presentation of the data.  The omissions are also interesting: first, there are no smaller parties listed even though FOCUS does list the smaller parties and those parties sometimes rise to the level of relevance, if only in the way they take votes away from others (by this standard, SaS would not have qualified to appear on the list until November of 2009); the other interesting omission is Median, the polling firm with the “you name it” methodology.  I have been critical of this in the past and suggested that the poll is out of line, so its omission is justified here.  I use Median to look at trends, but without graphics (as in Pravda’s case) this is not as relevant and including Median might simply be confusing.
  • Sme’s effort is rather more ambitious, a graphic interface with three parts.  The top one shows politicians standing next to lines representing party support and four buttons for polling firms.  Clicking on the poll button shows the results for the most recent iteration of the poll and the month in which it was taken.  Hovering over the individual politicians and lines shows the numerical level of support for the party.  This is great, and it is great to have a choice, but there are two minor problems: first, the lines are dwarfed by the pictures of the politicians and so it is hard to gauge the relative levels of support by eye, and you have to move the bar across to see numbers for individual parties so it is not easy to ‘eyeball’ the levels either.  Of course this is a relatively minor problem (and one easily fixed).  The second part of the display is a long-awaited (by me at least) effort to put party support for various polls and various time periods together in the same graph.  I’ve been doing this on and off for several years in this blog (most recently in the Dashboard, above) and have been waiting for a long time for Slovak papers to do it so I wouldn’t have to and could focus on the analysis rather than the creation of the graphics.  Unfortunately, while this comes close it does not get close enough for me to be able to shut down my efforts.  There are several minor but significant problems: first, the graphic rounds the numbers off to the nearest percentage point, which is actually fine for most parties but problematic in discerning key trends for smaller parties where rounding up or down may actually mean mischaracterizing party support by up to 10%; second, and most important, the graphic puts lines and dots where they do not belong.  Because it measures in one month increments, it puts dots even where polling firms did not issue polls (and the reader has no way of knowing when without looking at Pravda or my own data); it also puts all polls on the same line when in fact they may have been taken at significantly different periods during a month, something that becomes important as we get into the final months; the biggest problem is that it carries these forward from “present” levels, suggesting, for example, that we have data for Premier polls in April (or MVK in March and April) when we simply do not.   Still, this is such a big improvement over what has been done before that I am extremely impressed, and the problems above could probably be fixed by a clever webdesigner in about 2 hours.   The last part of Sme’s graphics is an assessment of the number of parliamentary seats produced by the results in question using “person” icons.  This is fine, but takes up much more space than it needs to do and does not easily tell the important story (mine doesn’t either, and I should fix that) which is how much that represents for various coalition combinations.

All of this means that I’m not yet ready to leave the dashboard business, but the great news is that the average consumer of news in Slovakia now has a relatively easy method for assessing the meaning of claims about polls and that is fantastic.  It also means that Slovaks, for the moment at least, have pulled ahead of their Czech neighbors who also have an election coming up and whose Lidove Noviny publishes what was, until Sme, the best graphical analysis: http://www.lidovky.cz/ln-volby-ps10.asp?v=preference.  The graphics here are better and more readable than Sme’s but they lack the comparative polling elements (and therefore the implicit “trendfinder” capability) that Sme has introduced.  Perhaps it is time for me to get into the Czech dashboard business as well.  Until the papers catch up again.

News and New Resources on Public Opinion in Slovakia

Because of formatting problems with certain versions of Internet Explorer , this page is now available as a .pdf file: 2009_12_30 Pozorblog Public Opinion.

I should begin with apologies for such a long hiatus and also with hearty congratuations to Monika Tódová of SME for writing one of the best articles on the process of public opinion polling in Slovakia: Prieskumy môžu politikov miast. It’s not very long but it finally addresses the problem of dueling headlines–“SaS/Most-Hid poised for success” v. “SaS/Most-Hid has no chance” (often in the same newspaper in the same week) and asks “How is it possible?” The answer lies in the ways that various pollsters survey the public, and it turns out (perhaps I should have known this) that FOCUS and MVK (and previously UVVM) offer a list of parties to choose from whereas Median gives no list. The latter method, of course, will tend to benefit those parties that are best known, while the former can give a boost to parties that are little known (and the process by which firms choose to add parties to their formal lists is fraught with difficulty.

The question now, is whether Todova’s article means a change in approach. Her article is the first instance I can remember outside of election campaigns when an article juxtaposes results from multiple pollsters. I can only hope that this will be the beginning of a trend even if (especially because) it might put this blog out of business. But I’m fairly secure that the time and space constraints of Slovakia’s papers will not let that happen and so there will be room for a blog obsessed with the minutiae of public opinion in Slovakia.

But just so that this blog does not immediately become obsolete, I want to share a new method I’m working on for displaying public opinion results. This is not quite ready for prime time, but using the Timeplot application developed as part of the astounding SIMILE project at MIT, it is possible to create dynamic charts that are fully modifiable and easily updated on the basis of simple text files. Of course that meant that I had to clean up my database, but now that that’s done, I should be able to present updated public opinion results with a minimum of effort (that’s the theory) and can even, theoretically, create a dashboard page on the blog itself for those who want to look at the numbers rather than read my analysis.

Below are examples of what is possible. I’m preparing a year end summary, so you’ll see these graphs again soon with some analysis attached.

Wanted: Effective journalism in Slovakia

Median logoI know Slovakia’s journalists are overtaxed with all that is required of them, but there are things they could do to make things better without too much extra work.  Case in point: today all major Slovak news sources today report, almost word for word, a press release from the firm Median reporting poll results from July:

The reports differ a bit, of course: SME and TA3 have a graph, Pravda has a table with past results from Median, HN relates the results (with no actual evidence) to the effect of recent scandals.

What none of them does is to answer the important question about Most-Hid.  Median, for reasons that are not clear, does not report on parties that do not cross the 5% threshold, and all the papers report that fact.  What they do not do is to say just how well Most-Hid actually did.  Here’s where journalism comes into play.  If the role of journalist goes beyond simply typing (or in this case probably pasting what others send to you) then it may be to try to find answers to important questions.

So here’s my suggestion: call Median and ask them 

If Median won’t answer, ask them why.  Focus reports its polling results with an extremely detailed and effective one-page report.  UVVM used to do the same.  There is no reason the papers should print Median’s results unless they are willing to rise to the same (bare minimum) standards. If Median won’t play that way, write a story about the fact that Median won’t provide necessary information):

If Median won’t answer and you still want to figure out something about Most-Hid’s initial support (a quite important detail), here’s my second suggestion: do some math.

If you add up the total support for the parties listed by Median in past polls, you get a consistent result of about 97% for May and June.  That means other parties are getting about 3% (specifically 3.1% in May and 3.3% in June).  If you add up the total support for July you get 92.3%.  That means other parties got 7.7%.  Of course this doesn’t tell you how well particular parties did, but it’s worth reporting that the “Other” number rose by 4.4% and citing the FOCUS numbers which show that support for below-threshold parties actually did not change systematically during the same period (9.0% in May, 12.5% in June, 10.2% in August), meaning that a large share, of not all of the 4.4% probably went to Most-Hid.

Total cost of these important details: one phone call and/or 2 minute 3o seconds of addition and subtraction.

If this lack of effort applies in this specific area, it’s probably also true in areas where I do not read.  It’s time for Slovaks to demand better journalism.

August 2009 Poll Results: Bending The Mold

Poll results, last 12 monthsThe pattern looks familiar and the lines haven’t changed much, but there are a few ways that this month’s results by FOCUS change our understanding of the current political competition.

I have actually been fearing this change for a long time–not because of the actual politics involved but because of the emergence of a new party.  The graphs in this blog are the product of a long and complex process of fighting with Excel to produce results that can be read by Google’s chart API (about which I understand little, but which is quite remarkable).  That process has produced an elaborate set of calculations which are, unfortunately, based on the presumption of a certain set of parties (and only those parties) gaining election to parliament.  This month holds quite a few big changes in public opinion and one of those–the emergence of Most-Hid with just over 5% of the vote–means that my old systems won’t work anymore.  This is good news, in a sense, because it gives me the impetus to find a solution that will not require as much work (calculating in Excel, creating the google charts, posting separately to Google Docs), but for the moment it makes things more difficult.  I will therefore resort to Excel charts for awhile.  And now, after that pointlessly detailed introduction (I buried the lead again), the graphs and then some thoughts on anybody should care:

Poll results, last 12 months

And the same graph without the distorting scale effects of Smer:

Poll results, last 12 months minus Smer

1. Most-Hid might make it into parliament
This may not be a surprise (Bugar is quite popular among Hungarians) but it is important, and the way the numbers fell is important in several ways:

  • Most-Hid v. SMK is not exactly a zero-sum game.  This month’s 5.3 score for Most-Hid came at relatively little cost to SMK which has dropped only about 1 percentage point in the last 4 months.  Of course SMK has dropped quite a few percentage points since Csaky became party chairman (and even before while Bugar was still chair) but it would appear that the party has brought disaffected Hungarians back into the political system rather than stealing directly from SMK.  As a result, Hungarian parties combined scored the best public opinion result that Hungarian parties have received in almost 5 years (since January 2005).  All of the opposition’s gains this month can be traced to that single re-mobilization.
  • Both can get into parliament.  There has been some discussion about whether the two parties might split the vote down the middle (as SNS and PSNS did in 2002) and lose representation altogether.  The results from today suggest that a 50-50 split is actually an ideal result for the Hungarian population in Slovakia.  More worrisome would be a 60-40 split, cutting the Hungarian representation nearly in half.  Of course there are some who suggest that infighting among Hungarian parties could disaffect enough to push Hungarian turnout so low that a 50-50 split would deny representation to both, but this month’s good results come after bitter conflict, so it is hard to imagine how bitter the conflict would need to become to provoke the worst case scenerio.
  • Things are far from over.  It may be that these two parties split the Hungarian vote.  It is more likely that one will tend to prevail over the other, either Most-Hid because of more dynamic leadership or SMK because of stronger organization and tradition.  This is one of the keys to the outcome of the next election so it bears considerable watching.

2. SaS has a (small) chance
This is something of a stretch because the party is only at 3.4%, but unlike the other small parties on the “right” it shows a positive trajectory.  KDS has stalled below 0.5%, and Liga appears stillborn (in eight months of polls the party has racked up a total–not average, total–of 1.4%).  OKS and ANO are effectively dead and DS and Misia21 exist only on paper (and barely there).  The big loser in this is probably Slobodne Forum which looked to be doing well in late spring, but SaS’s much better performance in the Europarlament elections appears to have given it the edge.  We have too little data to tell if it is a meaningful pattern, but SaS’s growth so far has been almost perfectly proportional to SF’s decline.

3. Smer’s recent decline continues
There is no real cause for gloom in the party (it is still almost three times the size of the next largest alternative) but it has dropped by nearly 10 percentage points from its (admittedly unrealistic) peak of early this year (when it was four times the size of the next largest alternative).  Since the 2006 election the party has peaked and waned five times, so the variability is nothing new, but this is the first time that the party has dropped sharply from a plateau rather than from a peak.  Of course it is safe never to rule out the possibility of recovery to new heights, but it is more likely that the weight of a poor economy and a large number of corruption scandals, some perhaps not so minor, have begun to take away some of the luster.  This may not be a huge loss for the party as many of those shifting away are likely the supporters who wouldn’t bother to turn out to vote for it (as they didn’t in the Europarliament elections).

How it adds up (Smer’s threshold for success)
The big question is the intersection of the points above:  the emergence of Most-Hid creates two parties that may or may not pass the 5% threshold.  SaS adds a third.  HZDS is the fourth (the party got a slight reprieve this month but even with that the 6-month, 12-month and 48 month trendlines show it dropping below the 5% threshold by spring and only the 24 month trendline puts above by about 0.5 percentage points, though the party’s loyal base also makes it necessary to adjust the numbers upward a bit in its favor).  Since each of these parties could dispose of between 3% and 6% of the overall vote and since the magical 5% makes or breaks the party’s parliamentary representation, a lot will be riding on the results.  My preliminary calculations suggest that in a worst case scenario for Fico–if SaS, Most-Hid, SMK and HZDS all made it into parliament–Smer would need 41% to be able to form  a two-party government with SNS (assuming that SNS’s preferences do not also continue to decline), though it could also settle for a three-party coalition identical to the current one (and it could sustain that coalition even if its own preferences dropped as low as 31%).  If HZDS failed to pass the threshold, Smer would gain some seats from the redistribution but not enough to overcome the loss of a potential coalition partner:  if HZDS falls and both Hungarian parties and SaS survive, Smer would need all of its current 38% to form a two party coalition with SNS.  Of course it is unlikely that all three of the smaller opposition parties would succeed.  If one of them fails, Smer could get by with 33% and if two of them fail, the Smer could form a majority two-party coalition even if it got only 28%.

This all deserves more thought and calcluation.  With any luck I will have opportunity to do just that.

Less than perfect public opinion coverage in Slovakia

 [Citajte post aj po slovensky]

“Less than perfect” could be the title of nearly post on polling coverage in Slovakia (hence, in many ways, this blog), and I shouldn’t be too critical but today both major papers went out of their way to mangle polling data.

In SME: “New party Most already clearly drawing Csaky voters” cites a Polis poll showing SMK at 5.7%.  Not only does the paper fail to assess whether Polis does good polls (they use telephone surveys which are problematic and they haven’t released a party poll for a long time so we just don’t know for sure–but more on that in a later post) but it suggests a trend–and a causal relationship even–on the basis of one data point.  We simply don’t know how SMK polled in previous Polis polls this year because there haven’t been any (at least I haven’t seen any and the paper certainly doesn’t provide them, using the 2006 elections as a frame of reference, as if that told us about polling trends).  Other polls did show SMK higher in previous months, but all polls do not sample populations the same way, and even UVVM had trouble with even samples of the Hungarian population (http://www.pozorblog.com/?p=110), and showed a result for SMK as low as 6.7% in September of 2008.  I’m not saying that Most-Hid is not pulling away SMK voters–in fact I suspect it is–but there’s no way of knowing it from the information we’ve got here.

In some ways worse, both SME and Pravda use the same data set to talk about the woes of HZDS.  SME simply makes the same mistake as above, suggesting that “HZDS continues to decline” on the basis of only the one Polis poll, without attending to the fact that different polls show different levels for parties and ignoring the fact that this month’s FOCUS poll show’s HZDS rising by more than a point, from 4.2 to 5.3.  Pravda’s story, “Meciar not bothered by polls does something rather worse, citing not only the Polis poll but also the numbers from FOCUS for April (4.8) and May (4.2) without feeling obliged to report the big increase reported in June.  It is clear to me that HZDS is in trouble, (and these microtrends usually don’t add up to anything, so I wouldn’t argue that HZDS support is actually increasing), but citing the first two numbers of a sequence of three seems a bit beyond the pale when the third number tells the opposite story.