art with code

2009-08-31

GDP projections

I took some GDP and population figures from Wikipedia and NationMaster and tried to predict the future with them.

First off, the current situation:

Region2008 nominal GDP/capitaPopulation in millions
United States47,103310
European Union38,390500
Japan38,055127
Russia12,487142
China3,1741,345
India1,0781,173


Grabbing last decade's average growth rates for each, we can do a linear extrapolation to 2020:

Region2020 GDP/capita2020 population
United States72,900341
European Union73,200661
Japan46,000127
Russia91,000137
China13,3811,431
India2,9901,368


These projections do have some issues though:

Russia posted 6x growth over the last decade, continuing that for a second decade sounds unrealistic, because 1998 Russia was suffering from a big financial crisis. A more realistic 3-4x decade growth would put them at $37,500-$50,000 GDP per capita.

160 million person more people in the EU. The EU population growth is mostly driven by enlargement. To grow by 160 million, the EU would need the Balkans (25 million), Turkey (75 million) and Ukraine (46 million.) And the EUR/USD exchange rate is favoring EUR at the moment, while in 2000 the USD was higher. Using an estimate based on 1999 figures, the GDP/capita projection would be around $67000. Though if the EU population growth misses the spot, the GDP/capita will likely be higher as a result, the possible member countries having a GDP/capita a third or fourth of the EU average.

Japan's low growth is from the bubble bursting in 1996, and the following decade of flat growth. Japan's only now reaching mid-90s levels again; the 1994 and 2008 Japan GDP/capita are roughly the same. Assuming that Japan starts posting decade growth rates in the 1.5-1.75x range, they'd be somewhere between the US and EU in 2020.

2040 projections


The linear projections fail even more for 2040. To highlight, consider Russia's GDP/capita. The 2040 projection is 3.7 million dollars. The whole Russian GDP would be 478 trillion dollars, constituting half the global economy. Which is unlikely, to say the least.

My guess for Russia's growth over the next three decades would be something like 4x in 2010-2020, 2x in 2020-2030, 1.5x in 2030-2040. Which'd put Russia at $156,000 GDP/capita. It might seem high today, but consider that the US GDP/capita thirty years ago was $10,000.

At the rate EU and Indian populations were growing in the last decade, EU's population would overtake India in 2080, and they'd have 3.5 and 3.4 billion people, respectively. The 2040 projection for EU would be 1.15 billion citizens, and that'd require all of North Africa and Middle East, plus perhaps parts of Sub-Saharan Africa, and Russia or Pakistan. Which sounds a wee bit far-fetched. Perhaps there will be an union of unions, or some other export mechanism for the EU's rule-of-law-investment-free-trade-combo.

The EU GDP/capita in 2040 would be around $153,000. The US GDP/capita projection for 2040 is $165,000, and Japan might be somewhere around the two. The linear projection for Japan would have them at $78,000, but that doesn't sound too likely.

The Chinese GDP/capita would be at $174,262 at today's high growth rates, which I wouldn't trust to hold for three decades. Today China's GDP/capita compared to the US is at around 1970s South Korea, or late-50s Japan and Spain. If China sticks to Japan's path (4x, 4.5x, 2.7x), they'll be at around US levels in 2040, perhaps followed by a bubble crash. With the South Korean way (4x, 2.3x, 1.6x), the projection is 35% of the US GDP/capita, or $58,000. The Spanish path (2x, 5x, 2x) would take them to half the US GDP/capita.

India would be the most populous country in the world with 1.86 billion, propelling it past China's projected 1.6 billion. India's 3x per decade growth would bring their GDP/capita to around $18,800. The Indian growth might pick up though, three decades of Japan-style growth would bring them to $58,000. Perhaps a good estimate would be somewhere in between: $32,000?

And then the computational power disclaimer: by 2040, you can get a human brain worth of processing power for something like $100 / year, which nicely throws a spanner in these GDP projections. If each person has a dozen virtual humans running on the side, the amount of stuff that one person can get done might well be multiplied by that.

2009-08-26

EU in the UK media

A recent EC poll on Attitudes towards the EU in the United Kingdom [pdf] tells us that the Brits think their printed press is overly negative about the EU. What? Not The Sun, surely? Jokes aside, television is also perceived to be negative about the EU, while radio is seen to give an objective view.

Maybe it's all because... (Check out this York University study on British attitudes towards EU too.)

Ta-dah! Speculation! Conspiracy theories!


I guess the reason why radio is objective is because radio is mostly BBC or commercial channels with their focus on music and advertisements. So either you get a view independent from commercial interests in the case of BBC, or no view at all in the case of commercial radio.

Television in Britain is a mix of BBC, ITV, Channel 4 and satellite/cable/internet channels, including Sky and US channels. Assuming that the domestic channels are EU-neutral and that the satellites have an axe to grind might explain the difference from radio. I don't know if you can draw any conclusions from this, but if you mix the perception of printed press with that of radio in a 4:6-ratio, you arrive at roughly TV numbers.

Not to forget that Britain is a bit of a queer bird in the EU due to it being a secondary country in its language area, while most other EU countries have linguistic primacy, or at least the dominating country is an EU member. The possible exception to this rule is Portugal, though I don't know how much Brazilian media they see. As a result, the British media is strongly influenced by U.S. and Australian (through News Corp.) interests and opinions. And the Australians (New Zealanders even more) really don't much like the EU, ever since the CAP started screwing up their agricultural exports back in the seventies.

(The future of the English language is pretty interesting too, with the prospect of an India-dominated English language media during the latter part of the century. You could reasonably expect the percentage of English speakers in India to go up to 40% by then (given improving education) which, coupled with a 50% increase in population, would put the number of Indian English-speakers at around 500 million. Not first-language speakers though..)

Oh, where were we? Right, reasons for anti-EU sentiments. Well. There's the common political strategy of blaming the out-group, celebrating the in-group. I.e. when things go wrong, it's EU's fault, but when things go right, it's because the current ruling party of Britain is the best ruling party in the history of ruling parties and they totally rule. And btw. you should totally vote for them in the next elections. Plus it's kinda hard to get kicked out from the EU, so you can blame it for everything. It's not like it's going to go all America on your ass, ey?

Stay tuned for the next blog post where I use the amazing powers of spreadsheets to create amusing growth projections for the mid-century!

2009-08-22

Why do netbooks cost 40% more than a year ago?

The netbook was touted as a small, slow and cheap second computer. The first netbooks were small, and while the 300€ pricetag wasn't all that small, at least compared to the promises of hundred dollar laptops, they were still the cheapest portable computers around. Today, netbooks are still small and slow, but they are anything but cheap. The average netbook costs 500€, which puts it in the midrange price category, above fully-featured 15" entry-level laptops that regularly sell for 400€ and below. What could be the reasons for the massive 40% netbook price hike?

Let's think like a computer manufacturer for a bit. You have a highly contested market with standardised parts and low levels of product differentiation and customer loyalty. And, worst of all, your product designs become obsolete within a year or two. 

Ok, first you want to set a breakeven period, during which you predict to sell enough copies to recoup your development costs. As computers obsolete quickly, you want a short breakeven period, let's say three months. So, in that time, you need to sell enough copies to make copies_sold * (selling_price - parts_cost) >= development_costs true.

The development costs are pretty much fixed for a run-of-the-mill laptop and copies sold depends mostly on selling price, so the only part that you can control is parts cost. Given the nonlinear curve for copies sold as price increases (demand curve), you need to increase your profit margin as you increase the parts cost. 

If you predict that you'll sell a hundred thousand copies of a 300€ netbook but only fifty thousand copies of a 400€ netbook, and your development costs are fifty million, you can work out the profit per copy required for each price point: 50€ for the cheaper netbook, 100€ for the more expensive netbook. You can then get the parts cost by deducting profit from the price: 250€ parts cost for the 300€ netbook, 300€ for the 400€ netbook. So adding fifty euros to the parts cost brought the sales price up a hundred euros.

Further, as the demand curve is sigmoid, sales of a slightly more expensive 500€ netbook might be only linearly worse than a 400€ netbook, so you could put in 75€ worth of extra parts and go for the low-end luxury segment. And, given the choice between the commodity and luxury ends of a linear segment of the demand curve, you want to go for the luxury end, as it makes your brand more prestigious, allowing you to spend less money on marketing.

My working hypothesis on the cause of the massive netbook price hike is that the parts cost rose by around fifty euros, bunting it from the exponential part of the demand curve into a linear segment, leading netbook manufacturers to start marketing netbooks as luxury items instead of commodity items. 

The basic hardware of netbooks has been fixed by Intel and their bundling policy, where it's cheaper to buy a CPU with the bundled substandard integrated graphics chipset, than only the CPU, making third-party chipsets unable to compete on price. The two major new parts added to netbooks since their introduction are an expensive large hard disk and the similarly expensive installation of Windows software, which probably account for the increase in parts cost.  

The resulting price increase then drove manufacturers to position the netbook as a luxury item. And as luxury items are very much driven by differentiation, but the fixed hardware platform does not allow for differentiation through faster processing speeds or bigger displays, manufacturers were required to differentiate on the basis of battery life and industrial design.    

In conclusion, low-cost commodity software and storage made the original 300€ commodity netbook a working proposition in the marketplace. Replacing the commodity parts with more expensive variants necessitated a disproportionate price hike and repositioning of the netbook as a luxury item, though the amount luxury you can derive from the slow CPU chained to substandard integrated graphics is debatable. 

2009-08-03

EU court rules 11-word snippets can violate copyright


Eleven words, eleven words.. yes.
Let's see.

/usr/share/dict/words has 98569 words. That to the eleventh is 8533827430813090265537515496031670135739632890386559769 different 11-word sentences. Each 11-word sentence takes 17 * 11 bits = 187 bits.

Which gives us a keyspace size of 2*10^44 terabytes. A bit too much.

But if we manage to reduce the freedom of degrees in the sentence through grammatic modeling and some other magic of data mining, while reducing our vocabulary to common words... Let's say 5 free words from reduced vocabulary of 1024.

1024^5 = 1125899906842624 different sentences. Each sentence now weighs 10 * 11 = 110 bits.

So, 15.5 petabytes for the keyspace size.

Assuming that we can generate this stuff as fast as the I/O system allows, we can imagine buying some machine time off the Amazon cloud to bruteforce it. If one machine can manage 50 MB/s write, and we use a cluster of a 1000 machines, the total combined write speed will be 50 GB/s. For a total time of 310000 seconds, or 86 hours, at the cost of 1000*0.1e/h.

Giving us the cost for plausible total copyright control over the English language:
8611 euros for computing time and 1.25 million euros for the 15,500 1TB hard disks to store the resulting document.

(Well, you can compress the document to less than a kilobyte by writing a program that generates every 50-bit number, but what is the legal power of that?)

And now, with our very own copy of bruteforced English, we can sue the pants off every stinking anglo and anglo-wannabe out there. COPYRIGHT BANZAI!

Blog Archive