Wednesday, August 24, 2016

Marginal cost of public funds: a valuable tool for thinking about taxation and expenditure in India

by Ajay Shah.

In an ideal world, taxation would be done in a frictionless way. The ideal world is a nice place where there are no transactions costs either for taxpayers in compliance, or for the tax authorities in collection. There would be no illegality and criminality surrounding the tax system. Most important, the presence of taxation would not modify the resource allocation in the slightest.

`Resource allocation' is economist-speak for the magnitudes of labour and capital, and the technology through which they are used. `Technology' is economist-speak for both the science and technology, and the business methods through which resources are utilised. In the ideal world, firms would produce based on pure efficiency considerations. Nothing about the questions `What to produce?' and `How to produce?' would be modified in the slightest by the tax system.

The government would collect taxes in this ideal world without imposing any excessive burden upon society. In other words, the cost to society of Rs.1 of spending by the government would be only Rs.1.

This notion is formalised as the `Marginal Cost of Public Funds' (MCPF). This answers the question: When the government spends Rs.1, what cost does it impose upon society? As with most economics, this question is posed `at the margin', i.e. what's the cost to society of the last Rs.1 that the government spent? In the ideal world, the MCPF is 1, but in the real world, it's always worse (i.e. bigger than 1).

The aspiration to get an MCPF of 1 was precisely expressed by Pranab Mukherjee in his July 2009 budget speech where, in para 31, he says:

I hope the Finance Minister can credibly say that our tax collectors are like honey bees collecting nectar from the flowers without disturbing them, but spreading their pollen so that all flowers can thrive and bear fruit.

This happy destination is one where the MCPF is 1, i.e. where the cost to society of Rs.1 of tax collection is 1.

Why do we get MCPF $> 1$?

Why is it that in the real world, we always have an MCPF that exceeds 1? There are costs of compliance, costs of administration, corruption, illegality, criminality. When all these costs are encountered, the cost to society of Rs.1 of marginal tax revenue exceeds 1.

Most important is the issue of a modified resource allocation. People respond to incentives. If income is taxed, people work less. If apples are taxed, people eat more oranges. This results in a distorted resource allocation, which results in lower welfare i.e. lower GDP. When the act of taxation distorts the resource allocation, and thus reduces GDP, the cost to society of the last Rs.1 of taxation exceeds 1.

There are seven sources of MCPF$>1$ in India:

  1. Income tax distorts the work-leisure tradeoff and the savings-consumption tradeoff.
  2. Commodity taxation distorts production and consumption, particularly when there are cascading taxes.
  3. We in India have a menagerie of `bad taxes' including taxation of inter-state commerce, cesses, transaction taxes such as stamp duties or the securities transaction tax, customs duties, taxation of the financial activities of non-residents. From 1991 to 2004, we thought the tax system was being reformed to get rid of these, but from 2004 onwards, things have become steadily worse, starting with the education cess and the securities transaction tax. All these are termed `bad taxes' in the field of public finance because when money is raised in these ways, the MCPF $\gg 1$.
  4. India relies heavily on the corporate tax, and has double taxation of the corporate form. In the last decade, corporate income tax and the dividend distribution tax added up to 35% of total tax collection. The double taxation induces firms to organise themselves as partnerships and proprietorships.
  5. There is the compliance cost by taxpayers and tax collectors, which is a pure deadweight cost. At the extreme, these include the costs imposed upon society by illegality and criminality owing to corruption in the tax system. When some firms get away with tax evasion, this changes the incentives of ethical firms to invest, which imposes enormous costs upon society as the most ethical firms are often the highest productivity firms.
  6. There are the consequences for GDP of the political economy of lobbying for tax changes, which arise when we do not have simple single-rate tax systems. E.g. if there was only one customs duty (e.g. 5%), this is much better than having different rates. Similarly, 80% of the countries which introduced the GST after 1995 have opted for a single rate GST.
  7. At the margin, public spending is actually financed out of deficits which are deferred taxation, intermediated through the processes of public debt management. Hence, in thinking about the MCPF, we must think about deficits and their financing also. Additional deadweight cost appears here, as we do financial repression (some financial firms are forced to buy government bonds). This is akin to a narrow commodity tax and is a bad tax.

All good thinking in tax policy and tax administration impinges upon the MCPF. If we setup a flawless GST, the MCPF will go down. If we reform tax administration, the MCPF will go down. The distortion associated with a tax goes up in proportion to the rate squared: hence the MCPF will be lower at a GST rate of 12% rather than at 24%.

How big is the MCPF in India?

As the discussion above suggests, there is the assessment of MCPF at the level of society as a whole and there is its measurement at the level of one tax at a time where the `bad taxes' will leap out of the page.

Estimation of MCPF is hard. Computable general equilibrium models are useful for thinking about shifting from commodity specific taxes to a single rate VAT. But most of the elements above are beyond the analytical reach of empirical economics.

One uniquely Indian problem is that on an international scale, most of these problems have been abolished. Most mature economies do not do financial repression, have low corruption in tax administration, do not have political economy of lobbying for tweaking of tax rates, do not have any of the bad taxes (taxation of inter-state commerce, cesses, transaction taxes, customs duties, taxation of financial activities of non-residents). Most mature economies have moved to a commodity neutral VAT or sales tax. As with most parts of the Indian macro/finance environment, India is an outlier with extremely poor institutional mechanisms in taxation. Nobody does the things that we do, and hence there is no international literature which helps us measure the MCPF in India. All we can know is that the Indian MCPF is very large.

Let's look at the international literature. Many papers find values from 1.25 to 2 in OECD countries. For an example of values from advanced economy, Dahlby and Ferede, 2011, find that the Canadian income tax has a marginal cost of public funds of 1.71, their personal income tax yields a value of 1.17 and their general sales tax has a value of 1.11. Feldstein, 1999, estimates a value of 2.65 for the US. Ahmad and Stern, 1984, estimate that the marginal cost of public funds in India for excise is between 1.66 and 2.15; for sales tax it is between 1.59 and 2.12; for import duties it is between 1.54 and 2.17.

When compared with conditions in India, the values seen in these existing papers are small, as the full blown distortions of the Indian tax system are not found in those countries. The one paper on India (Ahmad and Stern, 1984) only addresses a small part of the distortions associated with the Indian tax system. Indian policy thinkers who read Feldstein, 1999, which estimates a value of 2.65 in the US, would be delighted to achieve the conditions described in that paper.

Putting these considerations together, Vijay Kelkar, Arbind Modi and I believe that the true MCPF in India may exceed 3.

Further research on this question is very important. However, we are not able to visualise a research strategy that could put all the seven sources of distortion into one estimate, using today's knowledge of public economics. We are forced to form a guesstimate, and we propose the value of 3.


A central objective for tax reform should be to modify tax policy and tax administration so that the MCPF comes down. A central consideration in expenditure policy should be to narrow expenditure down to the few things where we can be convinced that the marginal gains for society exceed the MCPF, i.e. the last rupee of spending gives benefits to society exceeding the hurdle rate of Rs.3.

Spending on private goods. The value to society of gifting me Rs.1 to buy private goods that I like is Rs.1. Subsidy / transfer programs do not meet the test.

Leakages in private goods. If government intends for person $x$ to be the recipient of a private good of Rs.1, but owing to inefficiencies and corruption, Rs.0.5 reaches person $x$ while Rs.0.5 reaches an unintended beneficiary such as an official or a politician, this still yields gains for society of Rs.1, except that the gains are being allocated differently from what was intended. This does not change the fact that the aggregate gains to society was Rs.1, which is below the threshold of 3.

Spending on public goods. Spending on many public goods does yield gains that exceed the hurdle rate. We spend Rs.400 crore a year on running SEBI which produces the public good of financial markets regulation. SEBI does this poorly, and there are many ways in which SEBI can better utilise this money. But there is little doubt that the gains to society exceed Rs.1200 crore. If you imagined a world without SEBI, Indian GDP would drop by more than Rs.1200 crore.

Similarly, once we build a government agency to control air pollution, this will yield gains to society (in terms of the reduced burden of respiratory illness) vastly greater than the direct expenditures for running the agency. The same can't be said about public spending that makes the private goods of health care for people with respiratory ailments. In anticipation of the October 2016 epidemic of Dengue, let's fight mosquitoes. The gains to society from vector control easily exceed the hurdle rate, while the services of hospital beds and crematoriums are private goods.

Leakages in public goods. With public goods programs, we will sometimes get the situation where inefficiencies in the expenditure profile damage the marginal product to below Rs.3. Sometimes, these can be salvaged by improving spending efficiency. At other times, we should just admit that we have low State capacity in India and shut down the spending program.

How big should the State be? We should increase spending as long as the marginal gains to society exceed the MCPF. As the MCPF in India is high, this implies that the optimal scale of spending in India should be lower. Evaluating the possibility of a large State is the luxury for people who live in countries where the MCPF is low.

The free rider problem with sub-national governments. In India, the dominant source of resourcing of sub-national governments is central funds. In this case, if I am one state in India, it's efficient for me to advocate bigger expenses, as I don't pay the full cost of distortions experienced by the country. My marginal gains from spending one more rupee are mine, while the MCPF is imposed on the full country. To say this differently, Greece always wants more expenditure by the EU. Hence, sub-national governments should not have a say in the overall size of government. This perspective implies that in the new GST Council, the two-thirds vote share of states will generally favour a higher GST rate.

No comments:

Post a Comment

Please note: Comments are moderated. Only civilised conversation is permitted on this blog. Criticising me is perfectly okay; uncivilised language is not. I delete any comment which is spam, has personal attacks against anyone, or uses foul language. I delete any comment which does not contribute to the intellectual discussion about the blog article in question.

Please note: LaTeX mathematics works. This means that if you want to say $10 you have to say \$10.