Haldane, Fast and Slow

Andrew Haldane, chief economist of the Bank of England and, apparently, one of TIME magazine’s 100 most influential people in the world, gave a provoking presentation on the past and future of economic growth at the University of East Anglia a week ago, entitled ‘Growing, Fast and Slow’.

In it, - as the title suggests - he mobilized Kahneman’s framework of ‘Thinking, Fast and Slow’  to help approach the topical, or perhaps just millenarian, question of whether the world is about to lapse back into a pre-growth stupor out of which it struggled about 300 years ago.  Since I am interested in economic growth, and  have also found Kahneman's work interesting, I want to make some observations on it.

(And before I start, let me acknowledge that those of us who have not yet been named one of TIME’s 100 most influential people . . . . we do feel the lash of disappointment keenly. )

I do not think it simply professional resentment which drives my conclusion that Haldane’s delivers an  intellectual sugar-rush of covering a lot a ground quickly, which quickly dissipates when you start to digest his his thesis. Haldane fast is fun, Haldane slow is not so  good.

The Argument

Haldane wants to take on ‘one of the key issues of our time’, namely, whether the post-crisis slower rates of growth are a temporary or a longer-lasting valley in our economic fortunes. The way he wants to do this is by looking at long histories, since ‘some growth epochs have seen secular stagnation, others secular innovation.’  If we can understand the sociological and technical determinants of those growth phases, we’ll have a better idea of what’s going on today.

So he contrasts the period of secular innovation and growth  since 1750 with the three millennia prior to the Industrial Revolution, during which, apparently, global per capita GDP averaged only 0.01% a year. The conclusion is important: current growth levels are exceptional, and secular stagnation is far more common than secular innovation.  Having made this distinction, he sharpens it: ‘The short history (the Golden era) and the long history (the Malthusian era) of growth could not be more different.’

He then asks the question: ‘what caused this shift in growth?’  He characterises two possible explanations, claiming they are supported by the common theme of patience. In what he characterises as the the neo-classical explanation, patience supports saving, which in turn finances investment, and today’s investment is tomorrow’s growth.  The model is exogenous, supported by a) the patience of individuals, and b) technological progress, the ‘manna from heaven, a surprise gift which keeps on giving’.

Haldane admits this theory ‘does a decent job of explaining the phase shift in growth after the Industrial Revolution’, but identifies two problems. First, he mistrusts the exogenous role of technology, fretting there may be nothing we can do to encourage the next big wave of technological advance. Second, he is keen to develop an ‘endogenous’ growth theory in which in the key factors are as much sociological as technological - skills and education, culture and cooperation, institutions and infrastructure all work together building in a cumulative evolution fashion, rather than spontaneously combusting.

A fair portion of the paper assembles evidence supporting this second approach. For his argument, this evidence really matters because it supports what is actually his central thesis, that ‘the technological seeds of the Industrial Revolution were sown well before Hargreaves Arkwright and Watt arrived on the scene.’  And this proves that ‘innovation is more earthly endeavour than heavenly intervention’.

Still, even if you accept all this, the question still remains, why did societies suddenly being accumulating capital at particular points in history?  This is where he makes an interesting jump, arguing that ‘those technological and sociological trends may, in turn, have caused a re-wiring of our brains’.

His answer is to propose a mash-up between Kahneman and the history of interest rates. He suggests that the fall in interest rates (and return on capital) prior Industrial Revolution reflected society’s ‘evolving time preferences’ (ie, patience). Partly this reflected the possibility that income and wealth had growth enough to indulge the luxury of for more patient thinking. But also the post-Gutenberg proliferation of books may have ‘re-wired our brains’ in such a way to stimulate the slow-thinking, reflective, patient part of the brain (ie, Kahneman’s Type II thinking).  And in turn that would have supported the accmulation of intellectual capital, and consequently technology. So slow thought will have made for fast growth.

The Implications

Having constructed this framework, he then deploys it to draw implications for our own time. First, he wonders about the fall in real interest rates over the past 30 years, suggesting that perhaps this does mean our patience - capacity for Type II thought, technical innovation and growth - has grown. Perhaps, after all, technology will go on being the gift that keeps giving.

But his heart doesn’t seem in it: the last and most minatory part of the paper gloomily logs evidence that the key inputs to the endogenous growth theory are in decline: specifically, he worries that the rise in inequality (as measured by the Gini coefficient) will slow growth. Rising inequality, he fears, is leading to a deterioration in human capital. ‘Inequality may retard growth because it damps investment in education’.

And from there it is but a short step to worrying about short-termism, and proposing that the ability of the internet to cause a neurological re-wiring like that he proposed for the impact of Gutenberg.  ‘But this time technology’s impact may be less benign.’  He speculates that this time ‘an information-rich society may be attention-poor’. ‘It may cause fast-thinking, reflexive, impatient part of the brain to expand its influence. If so, that would tend to raise societal levels of impatience and slow the accumulation of all types of capital.  Fast thought could make for slow growth.’

Hence, it is not unreasonable to worry about a relapse back into pre-Industrial Revolution economic stupor.

Criticisms

There are many things unsatisfactory about this paper. I won’t even begin to list the number of questions I have about the historical data he relies upon. Similarly, there are a number of points at which he reaches for conclusions which, upon reflection, seem unnecessary and even arbitrary. But there are three areas which are central to his thesis, and which seem to me to be extremely contentious - eventually to the point of culpability.  In ascending orders difficulty, with the least-difficult first, they are:

The Role of Interest Rates. First, it is quite remarkable that he wishes to interpret interest rates, even real interest rates, solely and merely as an indicator of ‘patience’. It is remarkable, for example, that he does not acknowledge the role of inflation expectations, and their rate of formation, in the decline of bond yields over the past 30 years. It is remarkable, too, that he has nothing to say on the impact (or otherwise) of monetary policy and monetary institutions throughout the ages.  How is this possible? Surely he does not envisage monetary institutions and policies as merely far derivatives of underlying changes in neurological structures affecting our degree of patience?

Technological Teleology. Second, underlying his entire paper is a search for a technological teleology. Offering a theory of that teleology derived from an underlying stipulation of, or assumption of, increasingly broad and unquantifiable categories of ‘capital’  lies at the heart of the ‘endogenous’ growth theories he proposes. These attempts to categorise and quantify these purported types of capital are, I think, inherently implausible. (Two reasons: i) they can’t be counted and ii) even if they could, they’d suffer the same catastrophic problems with ‘real prices’ as all other capital stock estimates do.)  But much more dramatically, they dismiss the quite plausible possibility that technology has its own teleological aims and trajectories.  Haldane writes: ‘innovation was an earthly creation, not manna from heaven’, but that hardly exhausts the list of possibilities of what’s going on with technology. It seems quite likely that technological innovation implies and develops its own trajectory, and that that internal logic has a more powerful causative impact on the development and distribution of various types of capital, than vice versa.

The book I’d recommend for this is Kevin Kelly’s ‘What Technology Wants’. One of the ironies of Haldane’s presentation is that at some stages, he comes awfully close to acknowledging the possibility that technology in fact has its own teleology independent of human intention. He admits, for example, that ‘empirical evidence suggests a high degree of history-dependence, or hysterisis, in technological transfer’.  But not, apparently, in technological development itself, only its transfer. Really?

The Rise of China and Asia. But the worst fault, the most glaring absence, is the way Haldane has wiped the rise of China and Asia over the last 30 years from his assessment of the world’s current situation. This surfaces at every argument he advances for the worries about the state of the conditions allowing for ‘endogenous’ growth. For example, he worries about the decline of social capital, specifically that the rise in inequality (as measured by the Gini coefficient in the US) will slow growth.  He also worries that this will get worse as middle-class jobs continue to be ‘hollowed out’ by technology.

But you cannot exclude China from any argument about inequality and growth.  All analysis agrees that China’s historically-unprecedented growth has been accompanied by a large rise in inequality (so maybe a rising Gini coefficient doesn't automatically result in a lower growth rate, as claimed elsewhere?). Conversely, the rise of China’s working masses has also led to an unprecedented reduction in global poverty. It seems very likely that, when measured in global rather than national terms, the last 30 years has seen a narrowing of inequality rather than its accentuation.

Simply  noting the that the Gini coefficient in the US has risen at the same time as its growth rate has slowed hardly even classifies as argument, let alone as demonstration.

Similarly, Haldane worries about a possible decline in human capital. Why? The evidence he offers is that ‘the US is slipping down the international league tables of education attainment’ and ‘the UK could be following suit’. So what? Tell it to the Chinese!  On a global scale, you cannot conclude that human capital is eroding simply by noting that US and UK students no longer always rule the roost.

And finally, he frets about infrastructural capital where, it is said, ‘investment trends are not encouraging’. Really? First, he equates infrastuctural capital with the size of public sector investment to GDP - can he really believe this to be a useful proxy? Second, of course, once  again, the only relevant data is taken to be from the West. It’s not as if there’s been any significant infrastructural investment in Asia over the last 30 years, is it?

To be frank, this sort of cherry-picking of the data, this consistent willingness to edit from the picture that part of the global economy which is not based in the US or Western Europe, ruins the latter part of his paper. Maybe 20 or 30 years ago, the omission of more than half of humanity from the story might have been overlooked, or at least excused as an inevitable result of a lack of data. But not now - extrapolating global growth trajectories from the narrow and difficult recent experience of the Western middle classes is simply dumb.

My Conclusions

Having made these criticisms, I feel duty bound to offer my own conclusions on the future of economic growth. I have three.  First, any competent interpreter of Solow growth models will recognize that the underlying assumption is that there is no reason why with sufficient capital back-up an Asian (say) cannot be as productive as (say) a European or American.  So if capital flows are truly global, the ineluctable tendency will be for the world to become less unequal, and, specifically, that the unusually privileged position enjoyed by Europeans and Americans will be constantly under threat. Second, that in these circumstances, the immediate response will be for those privileged to seek to entrench their privileges by securing monopolistic or oligopolistic market positions in any way they can. In this scenario,  widening national inequality measures in the West will almost certainly reflect rising  economic rents, and usually market failure. Third, technology has its own teleology; it is infinitely more likely to surprise and amaze us than to sputter out because of our own ‘lack of social capital’. 

How the Oil Price Shock Is Hitting US Industry

It's reasonable to attribute the recent disappointments of domestic demand (retail sales down 0.9% mom in December, down 0.8% in January) to the impact of the oil-price shock.  During the initial price shock (which does not hit before November 2014), households are surprised and so do not adjust the rest of their spending patterns, with the result that the savings ratio ticks up swiftly and inadvertently, even as overall retail sales soften.  Later, of course, households get habituated to the lower price of energy, and spend the windfall. In short, you get a J-curve impact on domestic demand, so can afford to be relatively relaxed about the initial weakness in sales.  

But such complacency may be misplaced when it comes to the industrial sector.

The headline numbers don’t look disastrous: January’s industrial output rose 0.2% mom, and capacity utilization rates were unchanged at 79.4%, and the preliminary reading of February’s Markit manufacturing PMI showed a mild acceleration to 54.3. Nevertheless, by the end of 2014 there were some imbalances opening up between manufacturing supply and demand which would normally be enough to set alarm bells ringing. I track headline industrial momentum by looking at industrial production, inventory/shipment ratios, capacity utilization and exports.

As the chart shows, this combination of measures produced the sharpest fall in momentum in December 2014 since the financial crisis.  The two most obvious sources of weakness were a 0.3% mom fall in industrial output which was 1.5SDs below trend, and a 2.3% yoy rise in exports which was 0.7SDs below trend.

But this was not really what did the damage: rather it was a rise in the manufacturers’ all-industry inventory to sales ratio to 1.34, which was the highest since August 2009, and represented a very sharp rise from a ratio which for the previous five years had average 1.29 with a standard deviation of just 0.01pt.

Which sectors were doing the damage?  Obviously, petroleum and coal inventory/sales ratios have risen sharply,  presumably generated by a fall in speculative demand as prices have fallen: the ratio rose to 0.74x in December from a low of 0.68x in September. Even so, this level of inventory mismatch is not unprecedented: we saw the ratio at this level in mid-2012, in mid-2011 and throughout 2010.  But that spike would perhaps have been unremarkable had the overall total not been boosted by a steady climb in the ratios for computers/electronics, chemicals, textiles, printing and possibly beverage and tobacco.  One way of interpreting the modest build-up in printing, textiles, beverage and possibly even computers/electronics is that these are sectors which are likely to have been hit by the unexpected shortfall of retail demand in 4Q.

If so,  then a continuing deterioration in inventory position usually results in production cut-backs in the short to medium term. The suspicion that such a disequilibrium quietly opened up during 4Q is confirmed when one looks in more detail. In particular, one can compare the momentum of manufacturing output with the momentum of sales and inventories. The red line in the chart below subtracts sales and inventories momentum from output momentum: when the line is positive, it tells us supply momentum is outpacing demand.

If this disequilibrium becomes extreme as it did in 2001 and 2008, it heralds recession.  However, quite clearly, we're not there yet, and in addition such a negative result can be sidestepped if export momentum picks up sufficiently to soak up the ‘excess’ industrial production. This is what happened in 2003-2004 and again in 2009-2010.  The bad news is that currently, the rise in the dollar coupled with the fact that the US is leading the world economic cycle means that US exports are under pressure (falling 1.1% mom sa in November, and falling 0.8% in December), so this option is not available.

Conclusion? The impact of lower energy prices is having an unexpectedly significant impact on US manufacturing in the short term. The up-sweep of the J-curve in the retail sector therefore will be met with considerable cyclical relief.

China's Continuing Credit Squeeze, Impact on Trade

China’s January monetary data has to be read with extra caution, since the available data can be read either as:
a) disclosing a significant loosening of policy, with M1 growth up 10.6% yoy (from 3.2% in Dec), and new bank lending of Rmb1,470bn, which was the highest since the credit splurge of early 2009, and up 11.4% yoy; or
b) showing not significant improvement in liquidity conditions, since M2 growth slowed to 10.8%  (from 12.2% in Dec), and new aggregate financing, (which as well as bank lending includes ‘shadow banking activities’, foreign lending, bond and equity market financing) came to a less-striking Rmb2,050bn, which was 21.2% less than in January 2014.

On balance, the dour conclusion is probably nearer the truth. The key development driving January’s seeming expansion of bank lending was PBOC’s late-December decision to broaden the definition of deposits which are counted in bank’s loan to deposit ratio, which is subject to a regulatory ceiling of 75%. Specifically, PBOC now includes in its definition of commercial banks’ deposits those savings held by banks for non-deposit-taking financial institutions – such as stockbrokers, for example. Not only are these savings now included in the deposits calculation but in addition,   for the time being banks are not compelled to hold reserve ratios against them.  This rule-change matters, since these types of savings accounted for approximately 8% of the total deposits of listed banks, and consequently, they allow and encourage a significant increase in bank lending. 

But there is a cost: if these deposits are used to fund commercial bank lending, they are not available to fund other forms of financing. Hence whilst January’s bank lending rose sharply, this was paid for by the virtual annihilation of entrusted loans and trust loans.  In December, Rmb 668bn of these were issued/created; in January that total collapsed to just Rmb86bn.  So whilst growth of bank lending accelerated to 14.3% yoy in January, with a monthly gain which was 1.5SDs above historic seasonal trends,  the stock of total aggregate financing, by my estimate at least, slowed to 13.3% yoy on a monthly movement which was 1.2SDs below historic seasonal trends. And that broader financial aggregate is the one which matters. 

The unrelieved financial stress on China’s industrial economy also shows up in January's trade data. The trade surplus hit a record US$60.1bn, but this reflected the dramatic weakness of imports (down 19.9% yoy), not any strength in exports, which fell 3.3% yoy and were 0.6SDs below historic seasonal trends. This is a record surplus born out of shocking weakness in import demand, not an export-machine grabbing market share, or even a terms of trade benefit granted by falling commodity prices. More policy stimulus is urgently needed  to bolster working capital and cashflow in China's industrial sector.

January shouldn't have been a particularly weak month for China's trade - the calendar disturbances around Chinese New Year should have flattered January's totals slightly, since Chinese New Year doesn't fall until the middle of February this year, whilst coming at the end of January last year. So, if anything, the 19.9% yoy fall in imports, generated by a 21.2% mom fall which was 1.2SDs below historic seasonal trends, was even worse than it seems.  What's more, this collapse was not simply a reflection of commodity prices falling. Indeed, in volume terms, there was surprisingly steady demand for copper, iron ore, steel products and refined oil, whilst imports of crude oil continued to climb. Of the main industrial commodities, only coal took a real battering in January. Rather, it reflected a real gap-down in inter-Asian trade, with the worst hits showing up in Hong Kong, Japan and Asean. 

The sheer scale of this fall in import demand tells us unambiguously that major parts of China's economy are still very weak. Which parts and why? The fact that the worst hits were taken by inter-Asian trade strongly suggests that China's distributors are unwilling or unable to keep supply channels stocked at the levels previously taken for granted. This was also the message hidden away in December's fall in industrial profits, when despite topline pressures, companies cut their holdings of inventories and accounts receivable more than expected. Northeast Asia's trade with China is, after all, focussed on capital goods and intermediates, so it is these which must be taking the brunt.  Overall domestic demand indicators in China are weak but not spectacularly so, but the tightening of monetary conditions imposed in 2H14 has not yet been successfully reversed. It seems that directing some financial relief on cashflow and working capital is becoming a more urgent priority.

Petrol Prices and the J-Curve Impact on US Retail

They key shock of the week was the 0.8% mom fall in January’s total retail sales, or 0.9% if you exclude autos. How much of this was simply a reflection of lower gasoline prices, and if gasoline prices are responsible for the fall, what is the future trajectory of retail sales likely to be over the coming months? Gasoline prices and gasoline sales were the key to the whole retail slowdown in January. Sales at gasoline stations fell 9.3% mom sa, which is hardly surprising given that the average price of regular unleaded fell by 11.3% mom. Excluding gasoline sales (and autos), sales rose 0.2% on the month. On a 3ma basis, sales excluding autos are falling sharply averaging a 0.5% mom fall, but excluding gasoline, sales growth is running at 0.3%, dipping very slightly from the average run-rate of 0.4% which has been seen since 2011.

Consumption of goods accounts for 23% of US GDP, so at first sight this slowdown threatens growth. But since falling petroleum prices have been responsible for almost all of the slowdown of the last few months, the deflators will take care of that impact as far as GDP growth is concerned. The key word, however, is 'almost': beyond the simple price impact on the headline nominal numbers, the sudden fall in petroleum prices - down 25% since September - has another less obvious effect on consumer behaviour.  If the fall in gasoline prices is experienced as an unexpected windfall by the consumer, one would expect an initial phase in which the fall in prices is an unexpected windfall which initially (and inadvertently) saved,  only to be spent subsequently when the consumer adjusts to his/her new and larger budget.  In other words, one would expect a ‘J-curve’ effect.

Is this happening? The first crucial question to be addressed is: at what point would one expect the fall in gasoline prices to  take consumers by surprise?.  If the answer to this is: ‘when it falls significantly below the range recently experienced, then this is not difficult to spot. The chart below shows a fairly clear range sustained between January 2011 and October 2014, in which the price of regular conventional gasoline averaged $3.48 a gallon, with a standard deviation of 20 cents. In this chart the dotted lines show the two standard deviation level, which at the lower boundary comes out at $3.07 a barrel. One might speculate that above this level, the consumer would be unlikely to react to price fluctuations, but when it dived sharply below that, it represents a clear break from recent experience. According to Energy Information Administration that happened only in the last week of October. So it is only in November, and more obviously in December that one would expect any J-curve effect to be developing.

How big an impact? In the year to September, spending on petroleum accounted for 13% of retail sales (ex autos): a 25% fall in prices since then therefore amounts to a windfall gain equivalent to 3.3% of the retail budget.  This windfall looks to have been unexpected and to have been saved in December at least, when, reversing a decline in the trend visible since the middle of the year, the savings rate jumped 0.6pps to 4.9% in December (vs 4.1% in Dec 2013).  We should expect to see a similar or even higher rate in January.

But later, that savings rate is likely to retreat again as households adjust their spending to their newly-expanded budget. When petroleum prices stabilize, so will that portion of retail sales. Meanwhile, as consumers adjust their spending to reflect the new lower petroleum prices the personal savings rate falls and ex-petroleum sales accelerate beyond the current 0.3% mom run-rate, and probably, for a while, beyond the longer-term 0.4% rate. Looking back, we will see the J-curve effect at work.