Wednesday, 5 October 2011

Kondratieff Waves... Crashed Our Economy!

This article presents a divergent hypothesis from Tyler Cowen's lack of 'low hanging fruit' hypothesis for the current Stagnation (discussed in my previous post).

I was going to make one quick mention of a pure economics based thought that might have contributed to our current depression but it was based on an understanding inversion (see tiny writing below):

The 1970s switch of Government policy away from Keynesian practises towards monetary policies that buffered the economy from depressions and recessions may have precipitated the financial crisis by their very success: economic fluctuations shake the wastage out of the system, forcing inefficient businesses to shut down, reform or lay off unproductive staff. Keeping things steady just allowed more detritus to accumulate on the buckaroo donkey, saving all the pain for one big, inevitable mess... Well, actually that's complete nonsense!; It was *Keynesianism* that involved strong state intervention (like The New Deal). The 70s and 80s saw the rise of neo-liberalist policies promoting deregulated, free markets and privatisation. So, if anything one would have to blame *over* optimisation for the crash. 

+ Introducing Kondratieff waves:

To say that (inter)national economies are complex things is a blatant understatement. Many factors can effect a short term change in GDP growth rate; changes in: money supply, taxation, national interest rate, financial/business regulations (or removal of), public sector redundancies, etcetera. These have each caused quick and apparently large fluctuations in the past, but ultimately such disturbances only manifest for a couple of years at most before return to equilibrium. No amount of fiddling with these factors can stimulate sustained economic growth. Period. Yet on aggregate, since records began, the world GDP has *always* grown, year on year, even during the great depression, world wars and right now.
US GDP per person - "The Singularity is Near"
This continual rise in per-person wealth, standard of living and productivity has come from a succession of  technological innovations that have permeated society. Although, from an historical distance, the long term growth trend of a country looks pretty smooth, innovation uptake by members of a society tends to following a wave of adoption. Very gradual at first, rapid as it gains widespread popularity, but then perhaps never quite reaching *everyone*.
Wikipedia - Diffusion of innovations
There are only a discrete few innovations that are such majorly influential improvements to life as to have become ubiquitous (for example: mains electricity, automobiles, the internet), so they are spaced along our past. Each major innovation stimulated frenzied economic activity, indeed much employment was necessary to build, from nothing, massive infrastructure or industry (e.g. the railway/motorway network, industrial revolution). Lulls occur after each wave of innovation (because science takes time and) because the start of a wave is dependant on the new environment created by the previous one. So one should expect GDP growth history to be a little lumpy; a series of economic revolutions.

Nikolai Kondratieff wrote of his observation of a long wave economic cycle, back in 1925. It earned him Soviet Gulag, death by Stalin, and title to this theory (respectively). He estimated a fixed period of 50-60 years per cycle of: expansion, stagnation, recession. Since then it has been more commonly split into 4 seasons or irruption, frenzy, synergy, maturity (or such like).

I have led this introduction confidently extolling the effect of new innovations writ large, but economists have proposed purely different possible causes for K-waves: credit cycle (excessive debt), a generation-learning model with non-linear dynamic behaviour of information systems (from Tessaleno Devezas, as 60 years ~ 2 generations). However, there are many accounts that emphasis the historical role of innovations, with some focusing specifically on principle infrastructure (Arnulf Grubler "The Rise and Fall of Infrastructures", with 55 years between peak growth of successive transport modes: e.g. canal, rail, road).

+ A More Encompassing Explanation for Our Current Sorry State:

Cowen listed the most dramatic innovations of the previous century and a half, talking about their application to society as if they stimulated perfectly smooth growth across the whole period. He fails to highlight the unevenness each major technological revolution  created. I think he glosses over what we might now call K-wave 4: that that boom of innovation started even more abruptly than it ended; a clear cycle.

Cowen, I think erroneously, groups "free land" with his technological and social innovations. This comes from the modern American perspective of the book. I would look at it from the impersonal point of view of the land mass, which spontaneously acquired democracy (a constitution) and access to cutting edge, European technology. Once free from the repression of the British Empire, USA took a ballistic growth trajectory, like the one China (and India) has been on since the 80s, just like Japan's recovery after WW2. Because they were/are catching up with to the cutting edge society all their growth S-curves merge together, no need to invent the technology for the next one. As Cowen himself puts it:
 "Borrow and implement the best technologies and institutional ideas of North America, Europe, and Japan." 
But that's exactly what the US was doing until the beginning of the 1900s, catching up with the UK, it's economy inflating to the more dominant status it should already have had, courtesy of it's superior material resources (had modern society blossomed evenly distributed across the globe). Similarly Asia 'should' be economically dominant in the world (already), and it shortly will be at it's current rate (but that's for another post).
Screen Grab of Hans Rosling's entertaining 2009 TED India talk.
Graph shows: UK (orange), USA (yellow), Japan (small red), China (big red), India (blue).
Play with this dynamic graph yourself at GAPMINDER.
So my point is that, from a US biased perspective, economic progress may actually have come in a continuous flow, from the 1700s right up until the the 1970s (the world wars saw the US continue to grow by gaining relative economic advantage over a ravaged Europe). The influence of K-waves was obscured from US history; the 'stagflation' slow-down of the 70s and 80s might look like a isolated event to them. This might have helped hide the presence, and danger, of (the end of) another wave (the 5th). 'Planet finance', was overextended, but did no bank on an external, global economic shock like a K-wave depression, so instead of their housing bubble smoothly deflating, it burst. It was quite similar to the bust of the Japanese housing at the start of the 90s:

Japan had been growing rapidly, at 10% up until 1980, until it reached technological/social parity with US & Europe and it's per capita growth abruptly changed pace. Their housing market apparently failed to notice this change of pace and kept booming right through the 80s until reality set in a decade later, wiping a factor of 3 off the value of many properties, back to 1985 prices. (See "Ascent of Money - 'Safe as Houses'"). Although the losses in US (and European) housing have not been quite so dramatic, USA's global dominance (making it the safest investment) meant it's housing crash had a stronger global impact.

+ Rambling Thought 1 - Why do major technological innovations tend to clump together so as to reinforce a long wave pattern?:

[A] Inventions happen to be clustered together in meme-space (a massively multi-dimensional graph that encompasses all possible ideas, in which we are slowly uncovering the hugely convoluted tree of useful ideas). Spontaneous structure, like the gaps and island of stability in the periodic table; strongly defined features can arise almost mystically from the complexity generated in the interactions between simple, deterministic structures.

[B] Their real world uptake influences the discovery of each other:
[B1] An upswing in growth/productivity from one invention triggers forthcoming innovations immediately, sweeping the rest into a uniform wavefront. 
[B2] The down-swing/stagnation provides a more fertile time for effective research/creativity (better motivation).
[B3] The massive changes in society wrought by a ubiquitously successful innovation/revolution obscure/repress the expression of other significant ideas for a while (memetic incongruity). Like the refractory period after a neuron firing, except it takes a whole generation to get over a successful paradigm.

[C] Each successive technological revolution is entirely dependant on the platform created by the previous one. Kurzweil's series of overlapping sigmoids (logistic functions); In practice there is an invention, design, perfection lag that prevents the next sigmoid starting halfway through the previous (historically). Specifically, completed infrastructure provided by the previous revolution might be a prerequisite for the next revolution to even *start* taking off.

[D] Maybe if I'd finished reading "Bursts: The Hidden Pattern Behind Everything We Do" by Albert-Laszlo Barabasi, then I would have another idea for the apparently 'Bursty'sequence of major innovations.

+ Why are Kondratieff Waves 52 years Long?

This paper by Andrey V. Korotayev: "A Spectral Analysis of World GDP Dynamics" describes the results of a Fourier analysis (basically) of global GDP growth data. It claims to have found statistically significant evidence for K-waves and other cycles.
Fig.2B from Korotayev's paper. 
 Specific cycle frequencies identified by author, as they relate to previously recognised phenomena:

  • 2-3 Years     = Kitchin cycles (inventory lag; slow/speed output).
  • 7-11 Years   = Juglar/business cycles (investment; building factories).
  • 15-25 Years = Kuznets swings (may actually be misidentified 3rd Harmonic, below)
  • 17-18 Years = 3rd harmonic of Kondratieff Wave
  • 52 Years       = Kondratieff wave (technological innovation/paradigm cycle)

This 52 year wavelength certainly seems to fit the 50s-70s peak well (see below), but the 00s peak was far less pronounced, and the Western world is already plunged deep into recession with apparent stagnation ahead. Also, the author has considered that the data from the World War years are aberrations are so were excluded from most of the power spectrum analyses, so it's little wonder the 1900s-20s peak doesn't look to fit at all. I'm don't think the absolute magnitude of the plots are supposed to match.
Plot 1 is an idealised Kondratieff signal (primary and 3rd harmonic combined).
Plot 2 is the actual World GDP growth rate.
I quite fancy hypothesis [C] for my 'rambling thought 1'. If the periodicity of K-waves is tied into the large amounts of time it takes to build out infrastructure/manufacturing capabilities throughout a whole society. As well as sucking up huge volumes of labour and finance, each new infrastructure could only be built upon the previously completed one. The industrial revolution required the large towns/cities created by freeing much of the population from subsistence via the farming revolution (otherwise England could never have produced >70% of the worlds textiles). The steel age mega-constructions required existing railways. Telecommunications lines ran side by side with railways. Internet required mature telecoms.

In this case, for the next Kondratieff cycle, one should perhaps be looking out for large scale infrastructure, something(s) that will take a matter of decades to deploy (not just some gadget that will provide large productivity gains). So green energy infrastructure perhaps; solar PV (on a smart grid with bits of other renewable projects and electric cars) would fit well here I think, especially as it's growth is currently a rising exponential (likely the start of a sigmoid curve). A tablet PC revolution is not infrastructurish; genetics, nontech and robotics revolutions don't seem to fit well here either (though the practical part of genetics would require years long integration into existing health services).

'Infrastructure' is not exclusively land networks, it can include 'airways' (aeroplanes, airports) and shiping stock (and port fascilities), which are major investments. Arnulf Grübler has carefully plotted real world data showing lovely S-curve adoption, smooth log scale infrastructure growth (perfect exponential expansion, like this one for shipping, below) and switch overs between dominant energy types. What's more the plotted midpoints for principle infrastructure (rail, roads, etc) are spaced 55 years apart. (See full book PDF)

Plot from page 87 - Arnulf Grubler (1990) -
"The Rise and Fall of Infrastructures: Dynamics of Evolution and Technological Change in Transport"
+ Rambling Thought 2: - If the turn of this millennium indeed co-incided with the prosperity phase of a Kondratieff wave, why was it so much less pronounced than the 1950s boom?? (in fig.3D, above):

[A] The 4th K-wave (50s-90s) was unusually pronounced for some reason:

  • The nature of innovations of the 4th wave lent themselves particularly well to making GDP growth figures look good. Perhaps the modern concept of GDP was even conceived during this period (so is biased towards it's specific virtues).
  • Perhaps the 4th wave *was* just that much more significant!; cheap fossil fuel powered air, road, rail and wave transport made a massive change to the physical world which required a huge effort investmenting in the accompanying infrastructure and manufacturing.
    • The Green Revolution massively boosted the output of all developing nations during the 4th wave.
    • Japan's post WII, 10% catchup growth will have significantly contributed to making wave 4 stronger. Likewise China's growth should smooth out the lull between K-waves 5 and 6, then strongly reinforce Wave 6. The first K-wave recession after China reaches cutting edge GDP parity could be a real bitch though: think the problems of the 70s rolled in with a 00s housing boom bust 2 times bigger! :os
The 1st and 2nd wave can not be compared because of the paucity of data from those times, the 3rd wave was mired by WW1 and government naivité to macroeconomics (as well as lesser information collection). Explaining a gentle 5th wave is more difficult.

[B] The 5th wave (90s to present day) looks too small for some reason:
  • The 5th (and later) waves may be simultaneously damaging GDP (while growing it in some ways): e.g. by causing greater unemployment.
  • Maybe false growth in the 80s and early 90s took the edge off the information technology boom. We did have Thatcherist neolibralism after all. 
  • Traditional GDP measures are becoming less appropriate as they fail to include all the free productivity taking place in the digital economy. I don't just mean pirated music/movies, more like all the user generated (useful) content: blogs, wikis & forums, multimedia entertainment (remixes, free unsigned songs on youtube). Hence old fashioned GDP measures, aimed at a mass manufacturing economy, severely underestimate contemporary useful output. 
  • There may also have been massive missed opportunity for traditional economic growth of new media content distribution models and such, due to strangling IP legacy laws and belligerent existing monopolistic business interests.
  • Rising wealth inequality (in the most developed nations) suppressed true economic growth; these countries are productively weaker than they should be; mass infrastructure investment, or such like, was impaired by low median incomes. This could be the long shadow of the Thatcher legacy, extended by further deregulations and financial innovations that promote increasing concentrated wealth. Cowen claimed this, but was not specific as to causes or effects.
Korotayev's spectral analysis shows whole world GDP data, so growth from emerging markets undergoing continual (catch-up) growth, as yet unrestrained by technological limitations, will be superimposed. I think Kondratieff waves only describe the economic growth patterns of the currently most advanced nations. Previous to K-wave 4 the economic activity of the undeveloped world would probably have been negligible (by definition) or else automatically excluded by lack of records.

+ Rambling Thought 3 - Filling in the Economic Details of Kondratieff Waves:

- The rise of a wave: A massive new industry 'primes the pumps' (as Maynard Keynes had the US government do in the 30s) of the whole economy. It persuades the monied to part with their savings on mass, investing in new businesses that clearly have a profitable future. For example, US Steel had a bigger turnover than the US government at it's inception in 1901 (the prosperous part of the 3rd wave). More employees are needed and are paid better with apparent job security so spend more too, but I think the big investors are probably of primary importance in determining boom or slump (they are more able to take or leave investment opportunities).

- The fall: After a couple of decades of boosted growth, financial markets are are over-optimistic; the extra employment and capital returns from the revolution start tailing off has been taken for granted. There is a last fling of fake growth with too little real basis and so a bubble grows and bursts. Unemployment rises (as society is no longer undertaking a massive overhaul). Inflation may hit too it seems. Perhaps this is compensating for an inappropriate, but unnoticed, previous increase in the money supply, during the 'fake' growth period, at the end of a wave? Or, more directly, a government might have attempted to prop the slowing economy with spending that essentially boils down to printing money. 'Differential accumulation' might be a major cause: firms aim to beat the profit of competitors and dominant (monopoly) firms are more able to get away with this. They would be selling less product, so hike the price to maintain profit, hence stagflation may result if the downturn in sales is linked to increased unemployment.

+ Are K-waves Truly Fixed Frequency?:

If generation based learning models are genuinely meaningful in describing K-wave length, then one might expect a slight lengthening of their period as populations age, have children later. The observation that the houses in 50s TV shows contain roughly all the same items as contemporary abodes is a weak anecdote of great stagnation (fro my liking). However, Cowen noted that major scientific discoveries are getting more complicated and difficult (including Huebner's graph of slowing innovation per person). Public science funding always seems to be getting a raw deal too. If fundamental scientific discovery were falling behind schedule then this could well lengthen K-waves (at least until China catches up and sorts it all out for us).  

On the other hand, I quite fancy that this current bust/recession is an earlier than expected falling limb of the (5th) cycle. It's a strong indication of the end of a the computer revolution when the biggest manufacturer sells it's hardware business because it's products have become a (less profitable) basic commodity. In this hypothesis the bust could well have been the raising edge overshoot (the 3rd harmonic). So the primary Kondrateiff cycles are speeding up. The idealised waveform might even eyeball fit the raw data plot (of Fig.3D) better with waves 4 and 5 closer and 3 spaced a little to the left.

The spectral analysis result might make it appear that there is only direct evidence for fixed period waves, but as I've pointed out, there is anthropic selection of data (insufficient was collected before mid 1800s) and with the data for the war years removed there was only really one and a half wave lengths of data analysed, and that was dominated by the super-strong K-wave 4. One should not over weigh the significance of evidence purely because it produces a pretty graph or numerical result.

And so to segue onto a Singularity friendly variation on Kondratieff waves...

Daniel Šmihula's - "Waves of technological innovations and the end of the information revolution"

Šmihula's paper (full PDF) is rather wishy-washy, in that it shows no explicit data analysis, it seems to be purely a literary review from all the sources that it quotes. He asserts that successive innovation waves/revolutions are getting shorter in length (see below).

He links his modification of K-waves with the observable economic phenomena, we have so far dealt with, as "strong but not absolute"; he says of the latter, more recent waves, "their mutual cohesion is more visible than in a case of earlier ones". Perhaps the big 4th K-wave I talked about signalled a resonance when the two types of waves overlapped. The shortening innovation wave length matching up with a more purely economic phenomena.

Šmihula's accelerating progress tallies with Kurzweil's observations of the increasing rapidity of technological uptake by the general public, such as telecomms devices as plotted below:
From "The Singularity is Near"
If there is something in Šmihula's musings (and mine) then the 6th innovation wave is starting now (in the depression phase) and it should be about 20 years long. So the next ~10 years will be a relative slump as green energy infrastructure grows through the slow early phases before finally blossoming (then bubbling and popping, no doubt). Clean energy may be dominated by a couple of technologies that seem to come from nowhere (following exponential growth): cheap, efficient solar PV being a near certainty. Kurzweil puts PV equivalent to entire world's energy requirements just 16 year away, perfect to fuel an economic boom in the early 2020s. After clean electricity production rises past a significant fraction (~30%) of total production, energy storage technology will (need) to start to explode too (to buffer the surplus between day/night, etc). For example via domestic hydrogen storage tanks a-la Dan Nocera.

Perhaps Craig Ventner style artificial oil, brewed using modified algae will be an overlapping innovation between the second part of the green energy boom, linking the 6th K-wave with the genetics revolution propper (predicted in the 2020s by Kurzweil). I think health services are going to see a transition from treating illnesses to treating patients over the 2010s. The advent of cheap, personal genome sequencing will enable this for certain. I'm not so sure about the treatments themselves, but hopefully computation systems modelling will produce personalised treatment plans too. Perhaps in-vivo genome modification will mean cancer is effectively cured and various other ailments can finally be addressed directly. I do not forsee the genetics revolution producing the same kind of employment boon as green energy installation should, but rise in quality of life could be palpable.

Extrapolating along the same progression, the 7th K-wave would be ~14 years long (2030-2045ish), which takes us right up to the forecast of a Singularity. It could well be nanotech dominated with a side helping of superhumanly intelligent robots, which if I allow myself to speculate wildly (for a change) would probably mean the final nail in the mainstream economy as we know it. Any 8th and later waves would have a different meaning for humanity (if they aren't bypassed altogether).

* Concluding Summary:

Nikolai Dmitriyevich Kondratieff
I think: Kondratieff waves are a are real, emergent phenomena. They arise from the interaction of practical, psychological and social (economic) limitations with the largely continuous progress of science. This produces a repeated pattern of societal innovations with invention, refinement and adoption. This last stage, the spread of the innovation throughout society, is the massive signal buried in GDP data, too big and slow to see: the forest behind a screen of trees (recent events). I think the current bust/recession signals the end of a Kondratieff wave and the fully developed countries should expect a relative stagnation period of the order of a decade before the next sustained economic boom.

The issue of the periodicity of Kondratieff waves is more uncertain. Being familiar with, and optimistic of, a Technological Singularity I gravitate towards the Smihula type (shortening exponentially). I do not feel the spectral analysis of global GDP data rules this out. Also, I imagine that most economists would be unknowingly biased the other way, by natural assumption that any repeated oscillation in the economy would be either regularly periodic (or random); the general populace tends to think in terms of a steady state reality or linear progress. But if one takes a cybernetic viewpoint, such assumptions seem arbitrary (and odd).

Looking at human endeavours from high above, outside Earth looking in, society is composed of (many) individual humans that each behave in ways determined wholly by the culture of technologies around them (and their genetic heritage). Society moved far slower in, say the 1600s, because of the way information spread (on horseback at best) and how weakly linked society was; each town was practically isolated from the rest of reality (by current standards). New technologies may primarily have arrived via immigrants to an area. The engines for 'industry' were animals, or human toil so building *anything* took a very long time, let alone re-making a whole nation. One could think about any society having a characteristic resonant frequency: the low density, loosely connected 1600ers like a the undulation of a soft bed, while the present day thumps like the skin on a drum, singing together almost in unison.

I certainly do not think there will turn out to be a Kondratieff wave of exactly repeating periodicity given that society, and every human constinuant of it, has been changed so substantially along the way; I can see no basis for such a phenomena. There could perhaps be an observable Tessaleno Devezas style repeated pattern based around learning patterns of successive generations, it could be superimposed in addition to the other causes; but again, given the statistical uncertainties, this could well turn out to be coincidence. I have most confidence that future history will consider a Smihula style wave pattern to have been most significant, with a sequence that only approximates to a geometric sequence; the impossible emergent complexity of global society and it's specific history always adding extra details.

The real test of course will be the economic data for the next 20 years; I predict we will see a sustained lull followed by a clean-energy + biotechnology dominated boom that causes 'Westen' GDPs to plateau in 2022 (plus or minus 2 years).

No comments :

Post a Comment