Secular Stagnation as a Software Glitch

The big noise in the econoblogosphere over the last fortnight has been the reaction to Larry Summers' reintroduction of the concept of "secular stagnation", the idea that all is not well with capitalism and that we may need to get used to low growth and persistent unemployment (or underemployment). Summers first notes a dog-that-didn't-bark oddity of the economy prior to the 2008 crisis: "Too easy money, too much borrowing, too much wealth. Was there a great boom? Capacity utilization wasn't under any great pressure. Unemployment wasn't under any remarkably low level. Inflation was entirely quiescent. So somehow, even a great bubble wasn't enough to produce any excess in aggregate demand". In other words, the new economy was a bit pants.

One could arguably extend Summers' description across the entire period of the "Great Moderation", back to the mid-80s. Though there was volatility in specific assets and interest rates, due to well-known local conditions (e.g. UK house prices and interest rates in the early 90s, the US dotcom boom in the late 90s etc), volatility at the macroeconomic level, i.e. GDP and inflation, was low. The industrial restructuring of the early 80s did not lead to a step-up in GDP growth across the developed world (let alone wealth "trickle-down"), but rather a regression to the postwar mean (2.6% in the UK), while unemployment stayed high. If we manage to hit that rate of growth in the UK by 2018, ten years after the crash, it will be hailed as a triumph.

Summers then turns to another puzzle, the aftermath of the successful attempts in 2009 to "normalise" the financial system: "You'd kind of expect that there'd be a lot of catch-up: that all the stuff where inventories got run down would get produced much faster, so you'd actually kind of expect that once things normalized, you'd get more GDP than you otherwise would have had -- not that four years later, you'd still be having substantially less than you had before. So there's something odd about financial normalization, if that was what the whole problem was, and then continued slow growth". In other words, where was the bounce back once Gordon & co saved the world?

The concept of secular stagnation was originally popularised by the US economist Alvin Hansen in the 1930s as "sick recoveries which die in their infancy and depressions which feed on themselves and leave a hard and seemingly immovable core of unemployment" (he was observing the petering-out of the New Deal recovery in 1937 and couldn't anticipate the impact that the coming war would have). The assumption behind this was that the motors of economic expansion, such as rapid population growth, the development of new territory and new resources, and rapid technological progress, had played out. Consequently, the upswing of the business cycle lacked momentum. This finds an echo in modern "stagnationist" theories like those of Tyler Cowen ("no more low-hanging fruit") and Robert Gordon ("modern technology is rubbish" - I paraphrase).

The origin of Hansen's thinking lay in Keynes's observation that net saving at full employment tends to grow, whereas net investment at full employment tends to fall. This is Keynes's justification for government to act as the investor of last resort, thereby maintaining aggregate demand and full employment. The socialisation of investment is back on the agenda, even if the S-word is to be avoided and pro-middle class projects (like HS2 and Help to Buy) preferred.

An implication of Summers' analysis, spelled out by Paul Krugman, is that "we may be an economy that needs bubbles just to achieve something near full employment", however the track record since the 80s suggests that these bubbles have actually been relatively poor at the job of providing a stimulus, just as QE has been in recent years, hence the persistent unemployment and absence of high inflation. This in turn suggests that there is a very powerful secular trend at work driving stagnation, and that bubbles and monetary policy have been able to do little more than ameliorate its effects. As Krugman says, "we have become an economy whose normal state is one of mild depression, whose brief episodes of prosperity occur only thanks to bubbles and unsustainable borrowing". So what causes this underlying mild depression?

The cause of stagnation in the Keynes/Hansen model is a combination of supply-side deficiencies (an ageing population, declining returns from education, not enough new monetisable technologies) and demand-side deficiencies (not enough consumption and/or productive investment). Supply-siders like Tyler Cowen naturally emphasise the former, with the accent on demography, moral decline and the non-appearance of jet-packs, while demand-siders like Duncan Weldon emphasise the latter, with the accent on inequality and wage stagnation. Some demand-siders, like Yves Smith, also point to the pernicious effects of modern finance: "Companies are not reinvesting at a rate sufficient rate to sustain growth, let alone reduce unemployment ... managers and investors have short term incentives, and financial reform has done nothing to reverse them".

Other commentators have sought moralistic explanations. FlipChartRick suggests that the growth of superstar executive pay has led to the decline in investment, but I think this is confusing cause and effect. Declining investment, along with weakened trades unions, has grown profits at the expense of wages and thus created a larger pot of winnings for distribution among shareholders and executives. Rising inequality certainly has a dampening effect on aggregate investment, because of the greater marginal propensity of the rich to save rather than consume, and save in non-productive forms like property, but it doesn't follow that investment is deliberately curtailed (in concert, across thousands of companies) in order to advance inequality. There must be a structural cause - i.e. something that isn't the result of policy but the unplanned product of changes in the material base.

Investment as a share of retained income has been trending down since the late 80s, yet profits have held up. One perspective on this, put forward by Ben Bernanke in 2005, is that the "dearth of domestic investment opportunities" produces an increase in lending abroad, the so-called "global savings glut", reflecting higher rates of return for capital in emerging economies. A second perspective is an "investment strike", i.e. capitalists are choosing to depress capital expenditure, despite growing profits in emerging economies, leading to an aggregate fall in global investment levels. But how can declining investment be sustained beyond the short-term? Surely lower levels of investment will lead to lower profits in future, and thus a "crisis in capital accumulation"?

A possible answer, according to L Randall Wray, is that the problem is neither a savings glut nor an investment dearth, but rather an excess of capacity due to "the productivity of capitalist investment in plant and equipment. To put it in simple terms, the problem is that investment is just too damned productive. The supply side effect of investment (capacity creation) is much larger than the demand side effect (the multiplier), and the outcome is demand-depressing excess capacity. We call that a demand gap". The importance of Wray's analysis is the focus on the material base, i.e. technological productivity.

Paul Krugman appears to be receptive to the idea that we may be living through a technological revolution, despite the naysayers: "What Bob Gordon (pdf) is predicting is disappointment on the supply side; what Larry Summers and I have been suggesting is that we may face a persistent shortfall on the demand side". He is also sceptical (as a good SciFi fan) about the assumed triviality of modern technology: "I know it doesn’t show in the productivity numbers yet, but anyone who tracks technology has a strong sense that something big has been happening the past few years, that seemingly intractable problems - like speech recognition, adequate translation, self-driving cars, etc. - are suddenly becoming tractable. Basically, smart machines are getting much better at interacting with the natural environment in all its complexity." Krugman's list of wonders is significant because what he is talking about is essentially software, the machine "smarts".

A paradox of eras of rapid growth is that they are also periods of great waste. This is the core truth of Schumpeter's "creative destruction": for every successful idea there must be a long tail of failures. But this is not a problem in macroeconomic terms as any spending helps boost aggregate demand, regardless of the return on investment, hence Keynes's suggestion to bury old banknotes in mines and let the private sector dig them out. The peculiar feature of the dotcom boom of the 90s was that it was insufficiently wasteful, despite the best efforts of venture capitalists, stock-boosters and a seemingly infinite supply of bonkers business plans. The reason for this, I think, was the shift in investment from hardware to software.

The 120 years from 1870 to 1990 can be thought of as the era of hardware. Technological advance accelerated because of three institutional features (this is a key premise of innovation economics). The first was the expansion of state-funded universities and technical institutes in the late nineteenth century, which provided the foundation for systematic R&D. The second was the growth of private-sector labs in large industrial companies in the early twentieth century (e.g. IBM and Xerox), which boosted the returns to applied research. The third was the growth of international standards bodies, particularly after WW2 (e.g. ISO, IEEE and IETF), which encouraged the widespread adoption of new technologies. You can see the ideological legacy of this institutional approach to innovation in endogenous growth theory, the lionisation of instrumental education, the fashion for "innovation clusters", and in the search for "synergies" between business, academia and the public sector.

An area that benefited from this approach was logistics, which is the unsung hero of the modern economy. In the century before 1960, there had been few major changes to the technology beyond the growth of road haulage (i.e. lorries) at the expense of rail. International trade was still dependent on cargo ships and predominantly manual docks. Containerisation (based on ISO standards) was the revolutionary change, leading to the closure of the old city docks, a vast increase in trade volumes, and a consequent fall in commodity prices. But there was a second efficiency gain in the 80s, as a result of the impact of ICT (mainframes, mini-computers, private datacoms networks) on inventory management, which led to the development of just-in-time inventories and lean manufacturing. These improvements in logistics appear to have been a major factor in the reduced volatility of GDP and the chief cause of the "labour supply shock" that we call globalisation.

This points to the increasingly transformative impact of software over the last 30 years. While the early phases of the ICT revolution were hardware-heavy, by the mid-80s software was becoming the dominant element in business productivity growth. From episodic capital-labour substitution (e.g. machine installations), industry moved towards continuous improvement and optimisation, hence the growing importance of process management and statistical control, and latterly data analysis. This didn't just improve productivity, it also made production more modular and portable (necessary to be measurable), which was an important factor in facilitating offshoring and outsourcing. Software also has a high "spillover" value, i.e. its adoption by one business can also benefit others (e.g. improved inventory management by suppliers reduced inventory costs for retailers as well).

Though LANs and email had arrived by the early 90s, the mass adoption of ICT only came in the late 90s with the second wave of Internet technologies, notably the Web and SMTP email, and the deployment of Windows 95/98 PCs on every desk. Parallel to this, the corporate data centre was transformed by the replacement of expensive mainframes and minis with commodity Wintel and Unix servers, the development of application-independent RDBMSs (which allowed you to build custom applications cheaply), and the growth of off-the-shelf ERP and CRM systems (boosted by Y2K) that centralised corporate data.

The result of all this was a simultaneous explosion in the utility of software and a fall in the price of hardware. This was masked initially because total budgets remained high during the 90s - i.e. what was once spent on a single mainframe was now spent on hundreds of PCs - but it became apparent that this was a one-time bonanza, even before the dotcom bubble burst. Though some technology providers sought to move their profit margins from hardware to software and ancillary services, the impact of freeware and opensource (whose roots go back to the 70s), plus the democratisation of software development, meant that the days of huge, year-on-year capex budgets were over. The more recent arrival of SaaS (software as a service) and the "cloud" is merely confirmation that the technology is now pervasive and practically abundant (i.e. very cheap if not yet free). In the 80s, only the biggest companies could afford programmers. Now, many SMEs can afford their own "Web guy", and a tech startup is by definition a business with minimal capital. The cost of entry for high-tech innovation has not been lower since the evolution of insitutional R&D.

According to the US Information Technology & Innovation Foundation: "Between 1980 and 1989, business investment in equipment, software and structures grew by 2.7 percent per year on average and 5.2 percent per year between 1990 and 1999. But between 2000 and 2011 it grew by just 0.5 percent per year... Moreover, as a share of GDP, business investment has declined by more than three percentage points since 1980". They attribute this decline to two main factors, a loss of US competitiveness and increasing market short-termism (the primacy of shareholder value). I think both of these play a part, but I also think a crucial factor is software, which has become the dominant element in business technology since 2000. Costly high-tech hardware is still central to the manufacturing sector (e.g. robots), but as that has declined in the US and UK from 25% of GDP in 1980 to around 12% now, it is clearly not the dominant element in the wider economy. Software, on the other hand, is extensively used by every industry sector.

Technology is cumulative in nature, i.e. it builds on prior knowledge and seeks to constantly improve its operation, but software is particularly efficient in this regard because it can be augmented and edited, rather than requiring complete redesign and replacement, and because a lot of the knowledge is publicly shared, notably through the incorporation of opensource, so patents are less of a restriction on the spread of techniques. While a machine might only be replaced every 3 or 4 years, software can be upgraded weekly. This means that software tends to improve at a quicker (and smoother) rate than hardware. We are distracted by Moore's Law to think ICT productivity growth is all about faster processors, when in reality it is about better exploitation of increasingly cheaper hardware resources.

This two-way movement, accelerating utility and commodity deflation, is not historically unknown, but it has traditionally relied on economies of scale. What is unprecedented is that this dynamic can now apply at very small scale levels. An SME can exploit ICT and logistics to create a new market with minimal capital outlay. It should hardly be a surprise then that the demand for capital is falling at a time of widespread innovation. The madness of the dotcom boom, both the desperate search for something to throw capital at, and the promiscuity of VCs who realised you could afford to back an entire field of losing horses, was telling us something profound.

For some, like Frances Coppola, the coming abundance is problematic: "As the productivity of both labour and capital increases, the need for them diminishes. This is why the economy is creating bubbles. Those with assets are desperately looking for yield, and governments are desperately trying to generate jobs. Like animals in a drought, investors and governments crowd into the last remaining waterholes, as the water in them gradually evaporates". The modern economy is generating a lot of profit but insufficient employment (i.e. well-paid jobs) because it has become very efficient. The problem then is one of transmission, i.e. the distribution of this profit, not a malfunctioning engine. It is a matter of political economy.

It is a commonplace that banking by 2008 was no longer fit for its social purpose, but the implication of secular stagnation may be that the traditional model of industry - based on capital accumulation, high employment and the leverage of institutional innovation - may have already run its course as well. The solution may not just require the socialisation of investment, and the corollary of a job guarantee, but the socialisation of capital and its remittance as funded free time (i.e. a basic income). If you think that throwing money at a crowd of people in the hope that it may produce an aggregate return sounds mad, I would suggest you need to take a closer look at how software development is already funded.