These groups have been chosen because they are of special interest in the context of the GST: either they are exempted (food, gold (Centre), power, non-medicine health, and education); or they are taxed at a lower rate (clothing, gold (States), medicines); or they are charged at very high, demerit rates (petroleum, tobacco, and alcohol).

Two concepts provide the starting point for making policy decisions based on evidence: equity and effectiveness.

Equity allows for categorization of goods as merit/essential/sensitive (hereafter “merit”), etc. Goods that account for a high share of expenditure of the poorer households will typically be merit goods; and a related feature will be that this share will decline for richer households.

But even if a good is a merit good, warranting a lower or zero rate, policy makers will want to ask how effective that decision will be based on how well targeted the implicit subsidy will be, where the implicit subsidy is the difference between taxing a good at the standard tax rate and the lower or zero rate: if the poor also account for a large fraction of total expenditure on the merit good, then the subsidy will be well targeted; if, on the other hand, they account for a small share of the total expenditure, then the subsidy decision will come with the cost that most of the benefits of the subsidy will accrue to the relatively better off.²⁹

So, one can think of a commodity-wise benefit-cost analysis for determining the rate structure. The benefit could be thought of as the subsidy rate for the target group, say the bottom four deciles of the population.³⁰ The subsidy essentially measures how much the expenditure of the target group would be increased by exempting a good rather than taxing it at the standard rate.

The cost could be measured in relation to the principle of effective targeting. The cost is simply that proportion of the total subsidy for any particular good that does not reach the target group and instead “leaks” to the non-target group, in this case, the top 6 deciles.

We depict this benefit cost analysis graphically in Figure-1 for different group of commodities. We want to focus on the groups (and related sub-groups) that are going to be the focus of important policy choices in the tioned earlier. The data are from the 2011-12 thick NSS household expenditure survey.

The vertical axis measures the benefit—the effective subsidy rate, which is the subsidy as a share of the total expenditure of the target group. The horizontal axis depicts the costs measured as the share of total subsidy on any given product that leaks to the non-target group.

Three circles are drawn to highlight desirable evidence-based choices: commodities in the north-west corner of the graph circled in red are socially worthy of exemption because the benefits are high and the costs are low. These include cereals, vegetables, pulses, edible oils, and fuel and light (excluding electricity). Conversely, commodities in the south-east corner of the graph, circled in blue are less worthy of being treated favorably in tax terms because the benefits are low and the costs high. These include gold, non-medicine health services, education, and power. In the middle are commodities, circled in green, that lie somewhere in between that are perhaps worthy of being included at the lower tax rate. These include milk, poultry products and perhaps clothing.³¹

In the case of food and fuels, the PDS system helps make the system fair. For example, taking account of the PDS, the effective tax rate on the bottom 4 deciles is -7.4% for food as a whole, - 32% for cereals, and – 5.7% for fuel and light excluding power. The PDS has therefore served as reasonably effective social policy.

Clothing is also an anomaly but not as striking as the other commodities mentioned above. It is taxed currently at about 3-3.5 percent even though it does not constitute as large an expenditure item for the poor. On balance, it warrants being taxed at the lower rate by both the center and the States.

Gold, silver and precious metals

Currently, gold, silver and precious metals face no central excise and most States tax these commodities at the non-standard rate of 1 per cent. There could be two reasons to under-tax these metals: for reasons of equity and to promote savings. Consider each in turn.

It turns out that there is very little achieved by way of equity and a high cost is paid for exempting these commodities from taxation. Figures 3 and Table 2 illustrate these points. Gold as a consumption good constitutes a small portion of the total expenditure of the poor and a much higher share of the expenditure of the rich (Figure 3). It has the characteristics of a luxury good than an essential good. For example, the richest decile spends 3.5 per cent of its expenditures on gold, silver, and precious metals. In contrast the poorest decile spends about 0.03 per cent.

Table 2 highlights how ineffective or unfair is the implicit gold subsidy. It shows the expenditure of these commodities of each decile as a share of total gold expenditures. The top decile accounts for over 63 per cent of total gold expenditure. And this is a serious under-estimate because we know that NSS is very ineffective at capturing the expenditure of the very rich. Cumulatively, the top 2-3 deciles account for an overwhelming share of total gold consumption and therefore appropriate nearly all the subsidy given to gold.  

A second reason for favouring gold and precious metals could be to promote savings. At a time when there were few savings instruments, it may have made sense to incentivize the purchase of gold via a lower rate in order to promote savings. But today, this objective has been overtaken by two developments: on the one hand, the emphasis is on proper financial inclusion via the Jan Dhan Yojana which would also serve as the more effective means of promoting savings; on the other hand, gold far from being a desired savings instrument has become a problem, with large gold purchases and imports becoming a cause of macro-economic instability.

Recognizing this, the government has recently tried to wean consumers away from gold via the gold monetization and gold bond schemes. It would be perverse and contradictory to use taxes to incentivize the holding of gold, and undo what the government is trying to do via these gold schemes. At the very least, tax policy should be neutral on consuming precious metals.

Thus, on grounds of equity and effectiveness of targeting, on grounds of consistency of policy, gold should be taxed at the standard rate (bullion can be exempted from the GST). Instead, it is taxed at 1 percent, dramatically highlighting the incongruity of policy.

The final point to make, of course, is that the more rational gold taxation can be, the lower will be the standard rate which will be critical in creating a buoyant and compliance friendly GST. As shown in Table 8, the standard rate could come down to as much as 16.8 per cent if gold is taxed at 12 (6+6) per cent.

There might be concerns that increasing taxes on gold will lead to increased smuggling and evasion. This is a legitimate concern. But there is some evidence on how serious the impact of increased taxes might be. Import duties have been increased several times in the recent past on gold. These too are tax increases. In Figure 4 below, we plot the imports of official gold since 2011-12 and highlight the timing of import duty increases. The chart clearly shows that there is no seriously deleterious impact on gold imports in response to tariff increases. To some extent, there will be declines in consumption and imports if taxes increase but these are modest and manageable. The notion that there will be rampant evasion and smuggling if the taxation of gold is increased is not borne out by the data.

Power, health, and education

Some key sectors have been excluded or exempt from the scope of GST. These include power, health and education probably on the grounds that these are public goods, publicly provided, and of importance to relatively poorer sections of the population. But what evidence do we have on the underlying assumptions justifying such a policy?

Figure 2 suggests that these sectors are perhaps under-taxed currently (They lie well below and to the left of the line of the best fit). The design of tax policy, thus, needs to more carefully take account of evidence. For all three sectors, the benefits of exemptions (even without taking account of any embedded taxes) for the poorer sections is small because these items constitute a small share of their total expenditure. For the top 6 deciles, these sectors are three times as important as they are for the bottom 4 deciles. Moreover, the top 6 deciles also consume such an overwhelming large share of these services (probably privately provided) that nearly all the benefits of the implicit subsidy go to the relatively well off. In the case of education, the current tax structure turns out also to be regressive, with the bottom 4 deciles effectively paying greater taxes than the top 6 deciles. These commodities deserve to be taxed more like standard goods. Yet, today, they face low taxes and they are planned to be excluded from the GST.

Thus, tax policy in the name of the poor turns out to be poor or ineffective social policy. And the cost is a tax base that is narrow, exemptions-ridden, and in the case of power, the cost also includes breaking up the value added chain because it is an important intermediate input. In the medium run, of course, direct benefit transfers or better public provision of essential services would relieve tax policy of the burden of having to meet social objectives. But even in the short run, greater attention needs to be devoted to finding better instruments of social policy, and leaving tax policy to meet broad macro-economic objectives.

Notes:

1 Ideally, of course, if governments had well-designed transfer programs, they would achieve the desired objective of helping poorer households by providing cash transfers and sparing the tax system from having to attain equity objectives. In practice, this is not always possible and in India DBTs are still a work-in-progress. See Keen (2015).

2 The analysis can be re-worked for other target groups, say the bottom 3 or 5 deciles.

3 Strictly speaking, the benefit calculation should deduct the extra burden on the target group because the RNR will go up as a result of the implicit subsidy. The RNR will go up to a greater extent the more the leakage that occurs to non-target households.

4 So, for example, if the embedded taxes on a commodity is e (expressed in per cent), and t is the standard rate, then the effective subsidy rate of exempting a good is (t-e)/E, where E is the total expenditure of the target group. In contrast, if that good is taxed at the lower rate l, then the subsidy rate is (this ratio should be (t-(l-e))/E instead) because input tax credits will be available on the embedded taxes.

5 The calculations in Figure 2 are somewhat tentative and subject to a number of caveats. There remains some uncertainty about the assignment of tax rates to commodities in the data in the National Sample Survey, CPI and the Input-Output table. This would have to be reviewed and refined in future work. Second, a key benefit of the GST will be the ability for producers to claim input tax credits regardless of where their inputs are produced. The calculations have not fully reflected the input taxes (except for petroleum products), and given that not all input taxes can currently be claimed this means that current tax rates are effectively higher than what is reflected. Third, another factor that would increase the effective tax rates is central sales tax on the movement of goods between States. In future work, this will need to be captured. Fourth, the calculations use 2011-12 consumption aggregates but 2015 tax rates. Finally, the calculations assume not only perfect compliance, but also ignored threshold effects. Businesses below the VAT/Excise thresholds are not liable to collect tax, and this leads taxes to be overestimated, especially for the B40 who would be more likely to shop in businesses below the threshold.