jakke

May 20 2012 ∞

Five new polls for the Greek election in the last four days. Here are predicted seat counts for next month’s election, based on those five polls. The full predictions include nine parties, but these are the four (centrist pro-austerity ND and PASOK and leftist SYRIZA and DIMAR) that have been involved in most coalition discussions to date. Some highlights from the results:

  1. Polls are currently all over the place. The two most recent polls give SYRIZA 28% and 20.8% of the vote, respectively. I’m not sure if this is due to wholly incompatible methodologies or ineffective sampling techniques or ambivalent voters or what but the signal-to-noise ratio is pretty low here.
  2. The fifty-seat bonus to winning party is huge in determining who comes out ahead. It’ll be either the centre-right pro-austerity ND or the hard-left vigorously anti-austerity SYRIZA, and either one will come out with at least 120 seats of the 151 needed for a majority governing coalition.
  3. The probability of a pro-austerity ND/PASOK coalition is currently about 1 in 3. The probability of an anti-austerity leftist SYRIZA/DIMAR coalition is currently about 1 in 14. Obviously other coalitions are possible, but if no agreement can be reached (e.g., if SYRIZA can’t attract a coalition partner and the fascist XA takes up too many seats to form any other viable coalition) then there will be yet another election in July. By this point, the Greek government will almost certainly have literally run out of money and will have no way of operating as a going concern.

So that’s where things are at as of 20 May. The fifty-seat bonus for the winning party looks increasingly ridiculous in such a fragmented electorate; I don’t know how voters will feel if it’s a 0.1% difference between SYRIZA and ND, especially if there’s more suggestion of vote-buying or other impropriety. Although I guess it is kind of necessary seeing as (again) a third election would cause the government to run out of money entirely in the meantime.

7 notes

May 15 2012 ∞

So Greece is having another election. It’s almost certainly going to be 17 June, although there’s an outside chance that it will be 10 June. This has been a foregone conclusion since the two pro-austerity parties failed  get a majority in the election; the other party leaders were totally unwilling to cooperate, but they had to go through the constitutionally-prescribed motions of trying a bunch of different coalition possibilities. Obviously, all the party leaders are blaming everyone else and urging the country to unite and so forth.

How will the next election go? Probably quite a bit differently than the last one. Above I’ve tracked down all the polls I could find and used them to predict how the next election will probably go. (I included the nine largest parties in the prediction, but above you’re seeing the predictions for the pro-austerity PASOK and ND and the anti-austerity SYRIZA and ANEL, which will almost certainly be the largest parties.) The important thing to remember is that the first-place party gets 50 bonus seats (out of 300) so if the vigourously anti-bailout SYRIZA (Coalition of the Radical Left) comes in first then they would have a big advantage in forming a coalition with the other anti-austerity parties. A SYRIZA/DIMAR/KKE is about 80% likely to be feasible. (Although KKE is the Communist Party, and previously they’ve been unenthusiastic about negotiating a coalition.)

As you might expect this is causing some panic among investors. The Athens stock market is dropping steadily (of course) but much much more worrisome is the huge flight of deposits from Greek banks. Depositors (both sophisticated investors and ordinary consumers) are pulling hundreds of euro out of Greek banks. Everyone is worried that the Greek government will pull out of the eurozone and convert all the bank accounts to drachma (which is actually looking increasingly likely) so they want to take their money out and sink it into safer assets (e.g. canned goods). So far, the banks have been able to cover all the withdrawals, but pretty soon they won’t have the cash on hand to do so.

Anyway, Greece is probably going to be a pretty stressful place to be for the next month or so. They have no real government in place and the countries that are paying their bills are getting impatient fast. And then after the election expect some combination of panic, riots, and additional elections. Basically it’s suboptimal all around.

7 notes

May 03 2012 ∞
China’s rapid industrialization over the past few decades has given rise to an economy dependent on emissions-intensive manufacturing with a serious pollution problem. In January images from a sudden onset of haze blanketing northern China quickly made the rounds of the Internet and in February MIT researchers estimated that health damages from pollution cost the Chinese economy over $112 billion in 2005. As the Chinese economy continues to grow at an unprecedented rate, how has this pollution problem changed? Welp, this ended up being way sunnier than I thought, because as you can see from the data there really hasn’t been any trend in pollution levels at all over the last decade. Actually kind of surprised by this. I swear I’m not a lackey of the Chinese government or anything. Anyway, click through to read more.

China’s rapid industrialization over the past few decades has given rise to an economy dependent on emissions-intensive manufacturing with a serious pollution problem. In January images from a sudden onset of haze blanketing northern China quickly made the rounds of the Internet and in February MIT researchers estimated that health damages from pollution cost the Chinese economy over $112 billion in 2005. As the Chinese economy continues to grow at an unprecedented rate, how has this pollution problem changed?

Welp, this ended up being way sunnier than I thought, because as you can see from the data there really hasn’t been any trend in pollution levels at all over the last decade. Actually kind of surprised by this. I swear I’m not a lackey of the Chinese government or anything. Anyway, click through to read more.

3 notes

May 02 2012 ∞
To date, 85 Fringe episodes have aired. This seems to be enough to start some statistical analysis on the quality of those episodes - what makes for good Fringe? In the graph above I’ve used the AV Club grades; I realize these aren’t perfect, but they do seem to track fan consensus pretty closely and they generally identify the best episodes pretty effectively. I’ve converted letter grades to numbers by setting A to 10, A- to 9, and so on. Vertical lines represent the breaks between seasons. As you can see, there isn’t much trend here. Season 2 saw some of the absolute worst episodes, including a couple terrible fillers filmed during Season 1. The ends of the seasons tend to look way better. But other than that, lots of noise. Is there anything more we can get here? To find out, I regressed the quality rankings on five indicator variables that say something about what kind of episode it is: Whether the episode was one of the last five of the season (to capture the effects of the final episode rush) Whether any Observers other than September show up (to indicate special mythology-heavy episodes) Whether David Robert Jones shows up (to indicate Big Bad episodes) Whether alt-Astrid shows up (to indicate the alternate universe, and also because I love alt-Astrid) Whether William Bell shows up in any form (to capture ridiculous William Bell moments, including but not limited to sooooul magnets) Of these half-dozen variables, only three had an effect strong enough to bother writing home about: the indicator for being one of the last five in the season, the indicator for multiple Observers, and the indicator for David Robert Jones. Everywhere else, coefficients were statistically indistinguishable from zero. It looks like it doesn’t matter what’s going on, as long as we get David Robert Jones and lots of Observers and a season finale. I tried playing around with other indicator variables (e.g., for special opening sequences, which indicate time travel, or different ways of measuring when the episode aired) and ultimately nothing mattered as much as David Robert Jones, the Observers, and season finales. This might be an issue with sample size - but David Robert Jones plus Observers plus season finales reliably make for better watching than anything else I could think of. Anyway, what I’m trying to say is this: I am incredibly effing psyched for this Friday’s episode.

To date, 85 Fringe episodes have aired. This seems to be enough to start some statistical analysis on the quality of those episodes - what makes for good Fringe?

In the graph above I’ve used the AV Club grades; I realize these aren’t perfect, but they do seem to track fan consensus pretty closely and they generally identify the best episodes pretty effectively. I’ve converted letter grades to numbers by setting A to 10, A- to 9, and so on. Vertical lines represent the breaks between seasons.

As you can see, there isn’t much trend here. Season 2 saw some of the absolute worst episodes, including a couple terrible fillers filmed during Season 1. The ends of the seasons tend to look way better. But other than that, lots of noise. Is there anything more we can get here?

To find out, I regressed the quality rankings on five indicator variables that say something about what kind of episode it is:

  • Whether the episode was one of the last five of the season (to capture the effects of the final episode rush)
  • Whether any Observers other than September show up (to indicate special mythology-heavy episodes)
  • Whether David Robert Jones shows up (to indicate Big Bad episodes)
  • Whether alt-Astrid shows up (to indicate the alternate universe, and also because I love alt-Astrid)
  • Whether William Bell shows up in any form (to capture ridiculous William Bell moments, including but not limited to sooooul magnets)

Of these half-dozen variables, only three had an effect strong enough to bother writing home about: the indicator for being one of the last five in the season, the indicator for multiple Observers, and the indicator for David Robert Jones. Everywhere else, coefficients were statistically indistinguishable from zero. It looks like it doesn’t matter what’s going on, as long as we get David Robert Jones and lots of Observers and a season finale.

I tried playing around with other indicator variables (e.g., for special opening sequences, which indicate time travel, or different ways of measuring when the episode aired) and ultimately nothing mattered as much as David Robert Jones, the Observers, and season finales. This might be an issue with sample size - but David Robert Jones plus Observers plus season finales reliably make for better watching than anything else I could think of.

Anyway, what I’m trying to say is this: I am incredibly effing psyched for this Friday’s episode.

43 notes

Apr 22 2012 ∞
France had the first round of its presidential election today. To no one’s surprise, Socialist candidate Francois Hollande leads incumbent Nicholas Sarkozy in the exit polls. However, horrible far-right Marine Le Pen is doing way better than was predicted by polling. What happened here? Was this a random sampling mishap, or are voters lying about their choice? On Friday, five separate polling agencies released polls based on samples taken over Wednesday and Thursday (available here, here, here, here, and here). These should be reasonably close to how people actually voted, and since they’re all polling all of France at the same time they should be sampling the same distribution of voters. So based on those polls, what’s the likelihood of the exit poll outcome we saw today? Oh the graph above, the bell curves represent what the last five polls predict, and the horizontal dashed lines indicate the actual exit poll results. For Hollande and Sarkozy, then polling did a good job; the polls are pretty close to the middle of the bell curve. For (despicable bigot) Le Pen, though, the actual vote share was way higher than what the polls predicted. What happened here? There are three possibilities: The polling agencies all just randomly picked a sample without very many Le Pen voters. As you can see from this graph, this possibility is so far out at the end of the bell curve that it barely even registers. Lots of people changed their votes over the weekend. Millions of French people woke up Sunday morning with their mind totally changed and marched out to vote for Le Pen even though previously they’d been set on another candidate. This is definitely possible, although Le Pen never touched 20% support in any poll in the last two months. Voters are lying to pollsters because they don’t want to admit (even to a stranger) that they are the pathetic small-minded racists who would vote for Le Pen. Almost certain that #3 is what’s going on here. This has scary implications for polling European elections, because it indicates that as voter dissatisfaction with the eurozone and the response of the mainstream parties to the ongoing crisis grows we might see some really unpleasant surprise election results over the next couple years.

France had the first round of its presidential election today. To no one’s surprise, Socialist candidate Francois Hollande leads incumbent Nicholas Sarkozy in the exit polls. However, horrible far-right Marine Le Pen is doing way better than was predicted by polling. What happened here? Was this a random sampling mishap, or are voters lying about their choice?

On Friday, five separate polling agencies released polls based on samples taken over Wednesday and Thursday (available here, here, here, here, and here). These should be reasonably close to how people actually voted, and since they’re all polling all of France at the same time they should be sampling the same distribution of voters. So based on those polls, what’s the likelihood of the exit poll outcome we saw today?

Oh the graph above, the bell curves represent what the last five polls predict, and the horizontal dashed lines indicate the actual exit poll results. For Hollande and Sarkozy, then polling did a good job; the polls are pretty close to the middle of the bell curve. For (despicable bigot) Le Pen, though, the actual vote share was way higher than what the polls predicted. What happened here? There are three possibilities:

  1. The polling agencies all just randomly picked a sample without very many Le Pen voters. As you can see from this graph, this possibility is so far out at the end of the bell curve that it barely even registers.
  2. Lots of people changed their votes over the weekend. Millions of French people woke up Sunday morning with their mind totally changed and marched out to vote for Le Pen even though previously they’d been set on another candidate. This is definitely possible, although Le Pen never touched 20% support in any poll in the last two months.
  3. Voters are lying to pollsters because they don’t want to admit (even to a stranger) that they are the pathetic small-minded racists who would vote for Le Pen.

Almost certain that #3 is what’s going on here. This has scary implications for polling European elections, because it indicates that as voter dissatisfaction with the eurozone and the response of the mainstream parties to the ongoing crisis grows we might see some really unpleasant surprise election results over the next couple years.

107 notes

Apr 19 2012 ∞

In the past two days, four agencies have released polls for the upcoming Greek election: here, here, here, and here. This is a lot of data all at once, and it allows for a relatively clean prediction of the election results were the ballot today: assume all these polls are sampling the same population, and that any variation between polls reflects sampling error. The electoral system is mostly proportional with a few tweaks, so it’s relatively straightforward to convert poll results into seats.

Here are the results. If you like, you can compare them to the same prediction for the polls from March. Some highlights:

  • Thanks to the fifty-seat bonus for the winning party, it’s still likely that a eurozone-friendly bailout-backing coalition (like the one currently in power) will still be viable. The centre-right ND looks almost certain to pick up that fifty-seat bonus, and so PASOK and ND (the only parties that back the bailout) have a very high probability of forming a majority coalition. 
  • New right-wing nationalist party ANEL is gaining ground rapidly, and is close with SYRIZA (the radical left) for third-party status. The KKE (the communists) and DIMAR (a new left-wing antiausterity party) are not far behind. It’s not inconceivable that half a dozen parties will each have at least 10% of the seats, which obviously makes for a totally unworkable legislature. (Note that the KKE will never enter a coalition government, out of principle.)
  • The Green Party continues to hover on the edge of the 3% threshold for representation, while the fascist XA is virtually guaranteed to break into representation. Since the eurozone is desperate for an austerity-backing coalition after the election, horrible fringe parties like these Nazi-idolizing pigs actually have a chance at participating in government.

Anyway, the election is set for May 6, so expect more updates as we get closer. In particular, it’s important to pay attention to any gains by ANEL (which are almost certainly at the expense of ND) or DIMAR (which are almost certainly at the expense of PASOK). If the ND + PASOK coalition can’t muster 151 seats, Greece’s economic and political future gets pretty ambiguous.

EDIT: Note that I’ve redone the graphs as initially the standard errors were way too big. This has obviously affected the results substantially. Sorry.

2 notes

Apr 16 2012 ∞
Okay so after my last post all sorts of smart and awesome tumblrs gave me advice on how to embetter my graph. In particular, jrhyley showed me the API and cakesandcourage showed me the supremely useful python-tumblr which together make the task of doing this all properly embarrassingly easy. Serious gratitude to both of them because yeah I had no idea what I was doing otherwise.  Anyway, above please see my post hour frequency count for all 6000 of my posts! As you can see, my posting peaks around lunchtime and then again in the evening. It falls off pretty sharply after midnight, likely because there’s no one around to reply to and also I’m more likely to stick economics posts in the queue (which, I know, not cool). Apparently I tend to be doing most of my sleeping by 5 or 6 am, which actually sounds about right.

Okay so after my last post all sorts of smart and awesome tumblrs gave me advice on how to embetter my graph. In particular, jrhyley showed me the API and cakesandcourage showed me the supremely useful python-tumblr which together make the task of doing this all properly embarrassingly easy. Serious gratitude to both of them because yeah I had no idea what I was doing otherwise. 

Anyway, above please see my post hour frequency count for all 6000 of my posts! As you can see, my posting peaks around lunchtime and then again in the evening. It falls off pretty sharply after midnight, likely because there’s no one around to reply to and also I’m more likely to stick economics posts in the queue (which, I know, not cool). Apparently I tend to be doing most of my sleeping by 5 or 6 am, which actually sounds about right.

8 notes

+ ∞
Six thousand posts! I thought that I should do something exciting to celebrate the big round number, so I was going to post a big histogram of the hour at which I posted each of the last six thousand posts (and laugh about how many of them happen at 3 am and so on). Unfortunately, tumblr won’t let me get any more than the last twenty posts from RSS, and it looks like this is something they’ve set deliberately. I’m guessing they don’t want anyone developing their own tumblr viewing software. Anyway, entirely suboptimal. If anyone knows how to get around this, please let me know. (Obviously I need to uninstall missing-e or something.)

Six thousand posts! I thought that I should do something exciting to celebrate the big round number, so I was going to post a big histogram of the hour at which I posted each of the last six thousand posts (and laugh about how many of them happen at 3 am and so on). Unfortunately, tumblr won’t let me get any more than the last twenty posts from RSS, and it looks like this is something they’ve set deliberately. I’m guessing they don’t want anyone developing their own tumblr viewing software. Anyway, entirely suboptimal. If anyone knows how to get around this, please let me know. (Obviously I need to uninstall missing-e or something.)

7 notes

Apr 08 2012 ∞

Hey so it looks like the Greek election will be happening in less than a month. Here I’ve tried to predict the outcome of the election based on the five polls published in March; the standard error is taken from the variation between polls. Because there’s a 3% threshhold for representation and a fifty-seat bonus for the winning party, some of these histograms look a little strange. Here are some highlights:

  • In basically every possible outcome, the centre-right bailout-backing ND (who, remember, were the party who initially fudged the debt numbers with the help of Goldman Sachs) win the election and get the fifty bonus seats. They’re almost certainly going to be running the government after the election, but it will definitely be a coalition.
  • Two new parties (the centre-left DIMAR and the centre-right ANEL) are almost certainly going to be entering the legislature in a big way. Both are firmly opposed to the austerity measures. 
  • The only two parties that support austerity (ND and the incumbent PASOK) will probably have enough seats to keep the eurozone-backed program on track; odds are around two-thirds that the two parties have enough seats for a majority. In the other one-third of outcomes, there’s no clear path to continued austerity measures.
  • The Greens and the fascist scum XA are very likely entering the legislature, while the nationalist LAOS (who initially backed austerity, then changed their minds) might be losing all their seats thanks to the cutoff.
  • SYRIZA (the Radical Left) and KKE (the Communists) will both be making big gains, and could potentially hold the balance of power. However, the KKE refuses to ever join a coalition.

So yeah, that’s how things are looking right now. Of course there is still a month until the election, and things are very volatile - but for now, it looks that the most likely outcome is the same ND/PASOK austerity coalition returning to power with a greatly reduced share of the vote, and several new parties entering the legislature. If ND and PASOK can’t muster a majority, then the prospects for other stable coalitions are really murky.

3 notes

Mar 31 2012 ∞
Hey so here’s my prediction of how things are going to go in the Greek election, based on the most recent available polls. Since their system is almost entirely proportional (with a 3% threshhold and a 50-seat bonus for the winning party) it’s straightforward to go from polls to seats. And here are the results. Bear in mind that there are 300 seats in total, so the magic number for a government is 151. Unsurprisingly, things are going very terribly for PASOK (the party that won the last election) and ND (the party that got Greece into this mess by forging debt numbers with Goldman Sachs in the first place). However, as long as ND gets the most votes, they’re still going to be the biggest party thanks to the 50-seat bonus. In fact, it looks like PASOK and ND will be just on the edge of forming a 151-seat coalition with less than 35% of the vote.  From the debt crisis point of view, it’s absolutely essential to the deal that PASOK and ND form a stable governing coalition after the election. All of the other parties are vigourously opposed to the austerity package. In particular, DIMAR was formed a few months ago as a leftist alternative and has been gaining steadily, while ANEL wasn’t even included in the polls until March but is looking likely to form the main opposition. I’d be very surprised if ANEL didn’t make substantial gains against ND; they’re both centre-right, but ANEL is more vigourously nationalistic with a strong anti-austerity stance (and some rhetoric about Germany still owing money for World War Two, which is probably inevitable). If ANEL were to gain, oh, two percentage points relative to ND, then there would be absolutely no way that PASOK and ND could continue to maintain a coalition government that kept up the existing eurozone deal. Things would get very messy very quickly. Rounding out the parties (in case you’re wondering) are the communist KKE, the leftist SYRIZA, the nationalist LAOS, and the fascist scum XA. Note that the KKE refuses to participate in any coalition government as a matter of principle, so the leftist anti-austerity parties (KKE, DIMAR, and SYRIZA) have zero chance of forming a governing coalition even if ND collapses. I’m also not sure how much social-desirability bias there is in people not wanting to admit they’re the kind of disgusting filth who would vote for XA. Seriously hoping that there’s not a big surge there. I have zero doubt that the rest of the eurozone would gladly accept them into the governing coalition if they held the balance of power, despite the fact that they are literally and explicitly neo-Nazis.

Hey so here’s my prediction of how things are going to go in the Greek election, based on the most recent available polls. Since their system is almost entirely proportional (with a 3% threshhold and a 50-seat bonus for the winning party) it’s straightforward to go from polls to seats. And here are the results. Bear in mind that there are 300 seats in total, so the magic number for a government is 151.

Unsurprisingly, things are going very terribly for PASOK (the party that won the last election) and ND (the party that got Greece into this mess by forging debt numbers with Goldman Sachs in the first place). However, as long as ND gets the most votes, they’re still going to be the biggest party thanks to the 50-seat bonus. In fact, it looks like PASOK and ND will be just on the edge of forming a 151-seat coalition with less than 35% of the vote. 

From the debt crisis point of view, it’s absolutely essential to the deal that PASOK and ND form a stable governing coalition after the election. All of the other parties are vigourously opposed to the austerity package. In particular, DIMAR was formed a few months ago as a leftist alternative and has been gaining steadily, while ANEL wasn’t even included in the polls until March but is looking likely to form the main opposition. I’d be very surprised if ANEL didn’t make substantial gains against ND; they’re both centre-right, but ANEL is more vigourously nationalistic with a strong anti-austerity stance (and some rhetoric about Germany still owing money for World War Two, which is probably inevitable). If ANEL were to gain, oh, two percentage points relative to ND, then there would be absolutely no way that PASOK and ND could continue to maintain a coalition government that kept up the existing eurozone deal. Things would get very messy very quickly.

Rounding out the parties (in case you’re wondering) are the communist KKE, the leftist SYRIZA, the nationalist LAOS, and the fascist scum XA. Note that the KKE refuses to participate in any coalition government as a matter of principle, so the leftist anti-austerity parties (KKE, DIMAR, and SYRIZA) have zero chance of forming a governing coalition even if ND collapses.

I’m also not sure how much social-desirability bias there is in people not wanting to admit they’re the kind of disgusting filth who would vote for XA. Seriously hoping that there’s not a big surge there. I have zero doubt that the rest of the eurozone would gladly accept them into the governing coalition if they held the balance of power, despite the fact that they are literally and explicitly neo-Nazis.

2 notes

Page 1 of 9