If AI lifts productivity, will demand sustain growth?

If AI lifts productivity, will demand sustain growth?

Building on our previous piece yesterday, we now look at the macroeconomic implications of rising labour productivity on the global economy. We make the global economy our canvas to avoid between-country gain/loss that AI may result in since that is not the purpose of this piece. Our basic question is what may happen to the global pie of output and jobs. When looked at globally, demand consists of consumption, investment, and government expenditure because the trade component cancels out (see Box 1A).Since government intervention is outside the sphere of “pure” market economy, we focus on consumption and investment. The question we pose in this piece is how a rise in productivity would affect this demand? By its very nature, this piece is speculative, so we present two opposite cases for the readers to mull over.Crisis of underconsumptionA rise in labour productivity, without a commensurate increase in real wages, means a fall in the share of wages (denominator of the last term rises but the numerator does not rise in tandem in Box 1B). The implication of this is significant for consumption growth.Let us say the total output of 100 is divided equally between workers and capitalists and, for the sake of the argument, workers consume all their wages and capitalists consume half of it. The total consumption demand would be 75; 50 by the workers and 25 (half of 50) by the capitalists. Suppose AI is introduced in this economy which increases workers’ productivity dramatically but the wages rise by less than that, so the wage share falls to, say 30%. The consumption demand of the workers fall to 30, whereas for the capitalists it rises to 35 (half of 70), totalling to 65. Demand for consumption falls from 75 to 65, so demand worth 10 just disappears into thin air.The Luxemburg-Baranovsky debateMore than a century ago, a theoretical debate took place between Rosa Luxemburg, a German Marxist, and Mikhail Tugan-Baranovsky, a Russian economist on the issue of falling share of consumption in total demand and growth in a capitalist economy. It would be apt to revisit that debate in the context of AI and rising labour productivity.Tugan-Baranovsky argued that even if the share of consumption falls, investment can replace the lost demand. The 10 units of consumption goods that vanished into thin air above could simply get replaced by investment demand. He went to the extreme to argue that machines can produce machines which can produce more machines even if the demand for the final consumption goods they produce shrinks over time. The case of AI, imagined as machines producing more machines, comes very close to what he had in mind.Luxemburg argued, on the other hand, that machines are not an end in themselves. If the demand for the final consumption goods they produce is decreasing over time, there will be a surplus of such goods piling in the warehouses. The 10 units of consumption not sold will not be produced next time. And if they are not produced, then the machines which could produce them will lie idle. This would dissuade the capitalists from making more machines. Let alone machines producing more machines for the sake of it, there will be a contraction of investment in machines.The implication of her argument is clear. Of the two sources of demand, investment is unique. It creates both the capacity to produce output (say a factory which produces motorcycles) and the demand for output (construction of the motorcycle factory gives rise to demand for goods like cement, steel etc. which goes into building the factory). Looking at the level of the economy, if the capacity to produce output expands faster than the demand for it, there will be a contraction of investment eventually.Some counter-tendenciesTo be sure, there are other possibilities. None of this may happen if one of the following conditions holds. One, while productivity rises in the sector affected by AI, it may give rise to demand in sectors where labour productivity is quite low. Seen at the macro level, total labour productivity, which is the average of high labour productivity in AI sectors and lower productivity in the rest of the economy, may remain the same as it was in the pre-AI age. Two, even if overall labour productivity rises, wages rise in proportion, in which case the very source of a fall in the wage share disappears. This seems unlikely under capitalism though. Three, even if the wage share falls, investment demand could replace the lost consumption demand because AI requires more intensive capital investment to cater to the final demand.Early signs of AI investmentEven though it is difficult to isolate specifically which component of investment is exclusively associated with AI, there is a broad consensus in the literature that the bulk of it would show up under the category computers and peripherals. Chart 1 shows that investment in AI seems to be off the charts compared to any other component of investment. For comparison with other components of investment and GDP, we have created an index which starts at 100 in 2021 for all the variables under consideration. While nominal GDP has risen 1.5 times, investment in AI has risen 3.5 times within a short span of less than two years.A few caveatsThere are issues we have not touched upon in this piece. One, the ecological crisis that AI may lead to, in terms of electricity generation and water usage, is being considered to be devastating. Two, precisely for reasons of the ecological crisis, the global North may push the crisis to the South by setting up data centres there instead of their own countries, thereby replicating a colonial pattern of trade. Three, even though there may be a global slowdown induced by AI, some countries may gain at the expense of others. Four, this slowdown can be arrested through aggressive government expenditure, for example in the form of universal basic income or military spending, particularly in the global North (see Box 1A). Last but not least, this entire AI phenomenon may be a stock market bubble building to burst, much like the dotcom or subprime bubble in the US financial markets earlier.The last one in particular may be worth commenting on in the context of the unfolding saga of Nvidia. Most of these heavy investments have been made with the expectation of higher profits in AI in the future. And this has been financed in the US primarily through debt currently. There is, however, many a slip between the cup and the lip. One, the further the profit-making phase of AI in the future is, the more debt will pile up with a part of the debt every year being taken to simply roll over past debt. Two, if after all this, the profits of AI, as and when they do arrive, do not match the heightened expectations based on which such binge-borrowing was done, these debts will not be honoured. Actual cash flow may not match the already committed debt repayment path. It may set a domino effect in the financial markets, much like the subprime crisis. Three, if interest rates rise in an effort to target rising inflation induced by the self-inflicted crisis in the Strait of Hormuz, the debt burdens may rise even more.A highly leveraged sector, in the event of either a rising interest rate or failing expectations, or both, may result in what Hyman Minsky, an American economist, called Ponzi finance. Let alone gradually paying off the principal amount of the debt accrued, the firms are unable to even honour interest commitments. They take more loans to pay past loans but this cannot go on indefinitely, so, a spate of bankruptcies takes over. Recall the collapse of Lehman Brothers which triggered the 2008 global financial crisis.Even if we discount these caveats, as far as growth and jobs are concerned, it all depends on whether AI leads to a dramatic and economy-wide rise in labour productivity. Only time can tell who between Luxemburg and Tugan-Baranovsky would come out on top in this AI battle.(The authors are a part of the DevMac [Development Macroeconomics] network, which seeks pluralism in Economics. This article is the second in a two-part series.)

Original Source

Read the full article at Thehindu →

KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.