Fri. Apr 16th, 2021


India’s GST system will not work better if carried out through blockchains — they can be as easy to manipulate.

We read with great consternation an article that appeared in ThePrint —India’s answer to tax-evasion lies in blockchain—suggesting that the country should store GST tax records on a blockchain. Such a move, authors Susan Ostermann, Jarek Nabrzyski and Ian Taylor write, would reduce cases of tax evasion, corruption, and streamline payments in the GST system. Even in today’s world, where wrong-headed policy proposals and proposals demonstrating an incorrect understanding of technology are commonplace, this article stands out as an impressive tour de force in being egregious on both counts. We felt compelled to write this response to caution against the adoption of such shallow and misleading proposals, seemingly generated out of a comprehensive misunderstanding of blockchain technology and its applicability in solving socio-economic problems.

A Herculean task that emits CO2

The authors’ exposition of the technical properties of blockchains suffers from two serious errors.

First, they fail to consider the problem of scale in blockchain solutions, claiming that blockchain data “reside in millions of places”. The largest permission-less and public blockchain ecosystem – Bitcoin — has about 10,000 full nodes because the size of the data on the Bitcoin blockchain is over 320 GB and not all nodes that participate in the platform have the computational capabilities to maintain all the data.

With only 10,000 full nodes, the Bitcoin network takes up to 10 minutes to process a single transaction, since creating a block requires solving a time-consuming math puzzle and all nodes must agree that the transactions in a block are valid transactions. The Bitcoin network as a whole processes between 300k-400k transactions every day, with each transaction consuming about 900 units of electricity, approximately the same as the electricity consumption of the average Indian household in a year, according to 2014 data. So, once we put the GST on the blockchain, as the authors suggest, we will be able to enjoy the added convenience of having to wait several minutes to process each one of the lakhs of invoices processed in India on any given day and make the sweltering Indian weather hotter still with all the heat and CO2 emissions generated by the system.

Second, the authors present a view of blockchain that makes sense only if we consider all blockchains to be permission-less and public. But public and permission-less blockchains must have their entire transaction data available publicly so any node has a chance to validate them. Is it realistic for the government to expose its GST collection data to the entire world, including details of payments for classified military purposes? If not, they would have to use cryptography – which would necessarily restrict the set of possible nodes to ones with whom secret keys can be shared.  Ensuring cybersecurity for such a critical system would be a Herculean task, and even a single failure would compromise extremely detailed economic information about all public and private tax-paying entities in the country irreversibly.

Thus, we find it surprising that the authors ignore the plain fact that e-governance projects are only viable as permissioned private blockchains. For a GST administration system on a blockchain, this would likely be a solution where the clients could be businesses who participate in GST collection and payments – and the government GST administration would be distributed between nodes that work as peers, orderers, and validators. From a purely technical perspective, this is probably viable.

Not really an open, trustable solution

But then we confront a much more important issue. The premise of permission-less public blockchain is that it allows trust to emerge in a system even if individual participants in the system may be untrusted. Setting up a permissioned, private blockchain of the form outlined above requires us to presuppose trust in a limited number of opaque entities within the government’s control. So, if we must presuppose trust, one might as well trust a central entity managing a GST database than a conglomerate of entities managing a blockchain. Creating a blockchain that exists entirely within the State’s control defeats the whole point of using blockchains, which is to generate trust even if the State’s instruments are untrustworthy.

Thus, a private, permissioned blockchain is distributed in the technological sense, but not so in terms of the agency of human and institutional stakeholders. The nodes are in control of a few entities and therefore, the blockchain becomes just a distributed computing application with certain desirable properties, such as redundant storage, no single point of failure or corruption, and certain undesirable properties, such as high response time and energy consumption. If the entities involved are few, or under the authority of a single power, as would be the case of the GST system, which would run under the Indian Revenue Service’s aegis, then those entities may well collude and manipulate the blockchain just as easily as a database.

We also note that the authors claim without evidence that Indian insurance agencies settle payment with blockchain-based transactions. Having been deeply involved in the Indian blockchain technology landscape for the past three years, we can safely say that this is not true. The company named in the article “Vitraya Technologies” does not even claim to have any customers yet – their website just says that it is building a product for insurance settlement over blockchain. Such a gap between promise and reality is quite characteristic of blockchain-based claims. While the underlying technology behind blockchains is extremely interesting and inspiring, they are yet to demonstrate any advantages over centralised databases in situations beyond cryptocurrencies and escrow mechanisms. Permissioned blockchain can also be useful where multiple competing entities do not trust each other but need to cooperate for business purposes such as contract compliance and service-level agreement compliance. In such cases, the peers, validators, orderer nodes are from all the competing entities – thus allowing the creation of trust between untrusted entities. If all these important nodes are under the control of a single entity, apparent trust claimed is no trust at all.

Not the system India needs

The authors’ understanding of the economic problem they claim to be solving is also very sparse, focusing primarily on solving corruption problems prominently picturised in fictional series such as The White Tiger. The only genre of tax evasion problem that a blockchain-based solution could possibly solve is the one seen in this film, where people running an illegal mining business pay off bureaucrats in Delhi for unspecified favours. Extrapolating to taxation, the authors imply that people handling the GST IT systems may be accepting bribes to surreptitiously alter entries in the GST database. In a blockchain-based GST system, someone would have to bribe at least half the nodes participating in the blockchain to have database digits changed.  This much is incontrovertibly true.

But there is very little evidence that such skullduggery occurs at all, let alone at a large enough scale to warrant concern. In the first place, India’s tax-to-GDP ratio is quite comparable to other nations at the same level of economic development. In the second, to the extent that anomalously low tax collections are a concern, the concern rests more heavily with respect to direct taxes like income tax than indirect taxes like GST. India’s low levels of income tax collection, in turn, arise from the facts that nearly half of the population declares agricultural income on which taxes cannot be assessed, most non-agricultural income arises from informal, low-scale businesses whose proprietors have no incentive to declare their incomes, and the low level of per capita income of the country in general. It is as far-fetched to think that using a distributed ledger to store tax records instead of a simple database will move the needle with respect to tax collection as to believe that shifting from a gearless bicycle to one with gears will allow one to cycle to the moon.