vendredi 28 avril 2017

Published in Huffington Post, Us Edition.


Not one day goes by without hearing about predictions of massive job losses due to "Uberisation". However, statisticians find it hard to read the effect of said Uberisation outside of a very narrow set of sectors (car transportation, rental of apartments, etc.). Furthermore micro-economic studies tell a quite different story for sectors outside this set. In manufacturing [1], for example, the way production adapts to new technologies is very slow and incremental. Productivity gains may indeed have many causes, such as additive manufacturing (“3D printing"), increased digitization of processes made possible by the price drop of sensors and of information processing, or a simply better organization of the manufacturing chain. As a consequence, each year, industrial productivity naturally increases by a few percentage points, even in the absence of any technological/business model disruption. Actually if something new is happening in the industrial sectors it is rather a deceleration in productivity gains and productivity investments rather than an acceleration!




  
The issue is therefore not to brandish the digital threat some gurus and consultants like to evoke, but rather to understand the nature of such threat to the existing organization of a given value chain. In many cases, the threat is more likely to come from a competitor - who may be quicker than you in using digital in order to increase performance, or to propose innovative services – instead of an Internet giant or a start-up popping up out of the blue. There is real risk of investing too much to counter a theoretical Uberisation, and there is also real risk in not investing enough to reinforce your own productivity and quality of service. These poor investment choices, fueled by hazardous extrapolations and narratives, brings either inspiring optimism to investors or fears to businesses. That is precisely the definition a bubble [2].

Not all sectors are affected in the same way. In transportation, Uber “disrupts” the "interaction" jobs (which are about the relation between a cab and a customer), not the "production" jobs (Uber does not decrease the number of people driving a car). The “design” jobs (management, design, artists…) are still far from being disrupted: at best, artificial intelligence provides them with tools eliminating a small and repetitive part of their job. In the case of Uber, interaction and production jobs are complement, not substitutes: the lower the cost and the complexity of the former, the more plentiful the latter. Cab companies owners, call centers operators, and taxi licenses owners can indeed worry about Uberisation. But drivers shouldn’t as Uberisation leads to a net increase in the number of driver jobs, and this is a net gain for society as a whole after taking into account customers’ surplus in the tally. Please also note that we’re still very far from the driverless cabs – plane “autopilots” have been existing for years, and they still need a human supervision.

In order to estimate the overall impact of Uberisation on employment, we used micro-level employment data broken down across 88 industries (retail, metallurgy, publishing, and building construction ...) and across different functions (production, interaction and design). This data is available for France, but there is no reason why the results would be significantly different elsewhere. In our model, "production" and "design" activities are affected by productivity gains (use of machines, robots for production, artificial intelligence…) but they don’t disappear entierely. We show that “interaction” jobs may however be replaced entirely by the Uberisation process.

Our analysis was completed using industry experts’ assessments, we also took into account the dynamics of demand. For example, healthcare services are expected be digitized, but the increasing structural demand for such services ensures that, in the long run, there is no risk of a decline in employment.

Our model estimates that, in the medium term, 14 % of the total employment could be displaced by digital. This can of course just be an order of magnitude, but it is consistent with other studies (for the same country, OECD’s estimate is 9%)[4].

It shows that the reality is very far from the scary situation where all jobs are eaten away by robots powered by artificial intelligence. It quantifies a reality: some jobs are indeed made redundant by digital, and they are the ones the most visible – the "interaction" jobs (sales staff, call centers...). But millions of "production" jobs (construction, garbage collection,...) are hardly affected. Industrial jobs do indeed see annual productivity gains, but the size of such gains has been decreasing, not increasing, in the last decade.

Structure of the jobs by industry

This figure represents the structure of the jobs of each industry in the economy. The position of each industry relative to the vertex “I”, “P” and “D” represents the share of Interaction, Production and Design jobs in this industry. For example, industries with 100% of Interaction jobs would be represented in the vertex “I” and industries with equal shares of Interaction, Production and Design jobs would be represented at the center of the triangle. Based on the study mentioned in this article, the closer to the vertex “I”, the more menaced for “Uberisation”.



Our model also reveals some winning sectors. They are the ones complementary to interaction jobs: the simpler and the cheaper the interactions, the more customers will consume services sold through theses interactions. Farmers are already experiencing a maximum level of competitive pressure from purchasing groups. An Uber for fruit and vegetables (which would be delivered directly to customers) could hardly reduce their margins even more. On the other hand, it could allow customers to raise the level of quality and increase the farmer’s revenues. As mentioned above, a decrease in interaction costs will lead to an increase in production demands – efficiency gains are not “lost”, but reallocated somewhere. Finally, the economy is not inert: in France, 15% of all jobs are destroyed every year and the same percentage are created because of evolutions related to technology (more renewable energy jobs and less jobs related to coal), competitiveness (exchange rate variations, know-how development...) or consumer tastes (more e-cigarette and less mobile phone shops). The impact of Uberisation (14% of employment over several years) is not much compared to Schumpeterisation (15% of employment each year)!

There is no doubt that digitization will reduce the number of jobs in some sectors and increase it elsewhere. With each technological wave, there are risks for some jobs categories, but also opportunities for jobs and value creation. It is better to start with factual and realistic numbers in order to better deal with such risks and grasp such opportunities.

Vincent Champain, senior executive in a multinational company and president of the Long Term Observatory think tank, and Frederic Benque, investment partner at NextWorld. Many thanks to Joyce Bessis for her help on the translation.

[2] Definition of Professor Robert Schiller, expert on bubbles in « Rising Anxiety That Stocks Are Overpriced », New York Times, 27/8/2015.
[3] Employment Study 2013, Data from INSEE
[4] OECD, « Automatisation et travail indépendant dans une économie numérique », 2016.

0 commentaires:

Enregistrer un commentaire