09 February 2019

Business Costs and the robot tax

Do penalties shape corporate behaviour? Optus has been hit with yet another penalty, which a dour legal analyst might construe as a cost of doing business.

The ACCC reports that the Federal Court has ordered Optus pay a $10 million penalty for misleading customers over digital purchases. The telco's customers unknowingly purchased games, ringtones and other digital content through Optus' third-party billing service.  Optus admitted that the company misled consumers and breached the ASIC Act when it billed customers for third party-produced content which they mistakenly bought or subscribed to through its “direct carrier billing” (DCB) service.

The ACCC notes that
 The $10 million penalty is one of the highest imposed by the Court after ACCC action on a consumer matter, and equals the penalty paid by Telstra last year after it admitted to similar conduct. 
Optus admitted that it did not properly inform customers that the DCB service was a default setting on their accounts, and that they would be billed directly by Optus for any content bought through the service, even unintentionally. 
Optus, which earned commissions on items sold through the DCB service, also admitted that it knew from at least April 2014 that many customers were being billed for DCB content they had mistakenly or unknowingly signed up for. The DCB service allowed a purchase or subscription to be confirmed and charged to a customer’s bill after just one or two clicks on a web browser.
Despite receiving over 600,000 enquiries about the service, Optus failed to put in place appropriate identity verification safeguards, and referred customers who sought to query DCB service charges to third parties. 
Many customers then encountered significant difficulties in cancelling the purchases and obtain refunds from the third parties. “In many cases, Optus customers had no idea they were buying anything, and certainly did not need or want the content for which they were being charged,” 
The ACCC goes on to note that
ACCC chair Rod Sims said. “Optus failed to take appropriate action, choosing instead to continue to charge customers and collect commissions on these sales, even after numerous complaints. 
“We are pleased that the Court agreed that this conduct is simply unacceptable, and deserves a significant penalty,” Mr Sims said. 
About 240,000 Optus customers have so far been refunded. The ACCC understands Optus has paid about $8 million in refunds and third party providers another $13 million.   
Optus has committed to contacting potentially impacted customers who complained about the services and have not already received a refund, as well as those customers who Optus identifies as having been incorrectly charged. 
Optus will also review any future complaints in light of this action and deal with those customers in good faith. Given the volume of enquiries to Optus about the service there are likely to be many affected customers that have not yet received a refund.
Optus earned about $65.8 million in commissions for products sold through the DCB service, which launched in 2012. Its customers were charged about $195 million for the content. The ACCC commenced proceedings against Optus in October 2018. Optus admitted that it made false or misleading representations in contravention of the ASIC Act, and agreed to apply jointly with the ACCC for orders from the Federal Court.

Do the maths.

 'Should Robots Be Taxed?' - a study of the automation tax or robot tax by Joao Guerreiro, Sergio Rebelo and Pedro Teles comments 
The American writer Kurt Vonnegut began his career in the public relations division of General Electric. One day, he saw a new milling machine operated by a punch-card computer outperform the company’s best machinists. This experience inspired him to write a novel called “Player’s Piano.” It describes a world in which school children take a test at an early age that determines their fate. Those who pass, become engineers and design robots used in production. Those who fail, have no jobs and live from government transfers. Are we converging to this dystopian world? How should public policy respond to the impact of automation on the demand for labor? 
These questions have been debated ever since 19th-century textile workers in the U.K. smashed the machines that eliminated their jobs. As the pace of automation quickens and affects a wide range of economic activities, Bill Gates re-ignited this debate by proposing that robots should be taxed. 
In this paper, we use a simple model of automation to compare the equilibrium that emerges under the current U.S. tax system (which we call the status quo), the first-best solution to a planner’s problem without information constraints, and the second-best solutions associated with different configurations of the tax system. 
Our model has two types of workers which we call routine and non-routine. Routine workers perform tasks that can be automated by using intermediate inputs that we refer to as robots. We find that robot taxes are optimal only when there is partial automation. These taxes help increase the wages of routine workers, giving the government an additional instrument to reduce income inequality. Once there is full automation, it is not optimal to tax robots. Routine workers do not work, so taxing robots distorts production decisions without reducing income inequality. 
 Under the current U.S. tax system, modeled using the after-tax income function estimated by Heathcote, Storesletten and Violante (2017), full automation never occurs. As the cost of automation falls, the wages of non-routine workers rise while the wages of routine workers fall to make them competitive with robot use. The result is a large rise in income inequality and a substantial decline in the welfare of routine workers. 
The level of social welfare obtained in the status quo is much worse than that achieved in the first-best solution to an utilitarian social planner problem without information constraints. But this first-best solution cannot be implemented when the government does not observe the worker type. The reason is that the two types of agents receive the same consumption but non-routine workers supply more labor than routine workers. As a result, non-routine workers have an incentive to act as routine workers and receive their bundle of consumption and hours worked. 
To circumvent this problem, we solve for the optimal tax system imposing, as in Mirrlees (1971), the constraint that the government does not observe the worker type or the workers’ labor input. The government can observe total income and consumption of the two types of workers, as well as the use of robots by firms. We assume that taxes on robots are linear for the reasons emphasized in Guesnerie (1995): non-linear taxes on intermediate inputs are difficult to implement in practice because they create arbitrage opportunities. This assumption, which is standard in a Ramsey (1927) setting, restricts the outcomes that can be achieved when robot taxation is optimal. 
A Mirrleesian optimal tax system can improve welfare relative to the status quo. In fact, it can yield a level of welfare that is close to that of the first-best solution. Unfortunately, Mirrleesian tax systems are known to be complex and potentially difficult to implement in practice. 
For this reason, we study the optimal policy when the tax schedule is constrained to take a simple, exogenous form. Specifically, we consider the income tax schedule  proposed by Heathcote, Storesletten and Violante (2017) and linear robot taxes. We compute the parameters of the income tax function and the robot tax rate that maximize social welfare. We find that income inequality can be reduced by raising marginal tax rates and taxing robots. Tax rates on robot use can be as high as 30 percent and full automation never occurs, so routine workers keep their jobs. But this solution yields poor outcomes in terms of efficiency and distribution. 
We consider a modification of the Heathcote, Storesletten and Violante (2017) tax schedule that allows for lump-sum rebates that ensure that all workers receive a minimum income. We find that this modification improves both efficiency and distribution relative to a tax system without rebates. 
In the three best systems in terms of welfare, the first-best, Mirrleesian optimal taxes and simple income taxes with lump-sum rebates there is full automation once the costs of automation are sufficiently low. These solutions resemble the world of “Player’s Piano.” Only non-routine workers have jobs. Routine workers live off government transfers and, despite losing their jobs, are better off than in the status quo. 
One might expect that optimal robot taxation would follow from well-known principles of optimal taxation in the public finance literature. We know from the intermediategoods theorem of Diamond and Mirrlees (1971) that it is not optimal to distort production decisions by taxing intermediate goods. Since robots are in essence an intermediate good, taxing them should not be optimal. 
The intermediate-good theorem relies on the assumption that “net trades” of different goods can be taxed at different rates. In our context, this assumption implies that the government can use different tax schedules for routine and non-routine workers. We study two environments where there are limits to the government’s ability to tax different workers at different rates, Mirrlees (1971)-type information constraints and a simple exogenous tax system common to both types of workers. We find that it is optimal to tax robots in both environments when there is partial automation. 
We know from the work of Atkinson and Stiglitz (1976) that when the income tax system is non-linear it is not optimal to distort production decisions by taxing intermediate goods. But, as stressed by Naito (1999) and Jacobs (2015), Atkinson and Stiglitz (1976)’s result depends critically on the assumption that workers with different productivities are perfect substitutes in production. This assumption does not hold in our model. Taxing robots can be optimal because it loosens the incentive compatibility constraint of non-routine workers. 
We extend our model to allow agents to switch their occupations by paying a cost. In the first-best solution, agents who have a low cost of becoming non-routine workers do so. Those with a high cost become routine workers. Once the costs of automation are sufficiently low, there is full automation; agents for whom it is too costly to become non-routine lose their jobs and live from government transfers. In the Mirrlees solution to the model with occupational choice it is optimal to use robot taxes to loosen the incentive compatibility constraint of non-routine workers. The planner can use the income tax schedule to redistribute income or to induce more agents to become non-routine workers. When the cost of becoming non-routine are high (low), the planner resorts more (less) to using the income tax schedule to redistribute income. 
The paper is organized as follows. In Section 2 we describe our model of automation. Section 3 describes the status-quo equilibrium, i.e. the equilibrium under the current U.S. income tax system and no robot taxes. Section 4 describes the first-best solution to the problem of an utilitarian planner. In Section 5, we analyze a Mirrleesian secondbest solution to the planner’s problem. In Section 6, we study numerically the optimal tax system that emerges when income taxes are constrained to take the functional form proposed by Heathcote, Storesletten and Violante (2017) both with and without lump-sum rebates. In Section 7, we compare the different policies we consider both in terms of social welfare and of the utility of different agents. Section 8 discusses the model with endogenous occupation choice. Section 9 relates our results to classical results on production efficiency in the public finance literature. Section 10 concludes. To streamline the main text, we relegate the more technical proofs to the appendix.