In countries where average incomes are low, governments typically struggle to raise sufficient tax revenues to provide public goods and services effectively. Pakistan, for instance, raises only 13% of GDP in tax revenue and registers less than 30,000 tax-filing corporations. The UK, in contrast, raises 38% of GDP in tax revenue and registers over a million tax-filing corporations.1 Improving the mobilisation of domestic resources is a key challenge for developing countries.2 The low tax take in low- and middle-income countries is largely due to weak enforcement. Where informal sectors and the cash economy are dominant, taxable economic activities are easily hidden and don’t leave behind verifiable paper trails (e.g. credit card records). Audits are few in number and poorly targeted – partly because of the weakness of information trails – and tax evaders are rarely caught, let alone punished.
The prescriptions of traditional tax theory
What should tax policy look like in such a weak enforcement environment? In particular, how should businesses be taxed? One of the most celebrated results in public finance theory, the production efficiency theorem by Diamond and Mirrlees (1971), dictates that only production-efficient tax instruments should be used. Governments can tax wages, consumption and profits, but tax policy should not affect firms’ production decisions, ruling out taxes on intermediate inputs, trade or turnover. This theoretical result has been central to much of the tax policy advice given to developing countries. However, the result critically relies on perfect enforcement, an assumption that is clearly violated in countries with large informal sectors. So how should policymakers design tax policy when facing high levels of evasion? In a recent paper (Best et al. 2015), we break new ground in the analysis of this question.
Modeling the trade-off between production efficiency and revenue requirements
We start by developing a model of optimal taxation in which businesses may misreport their tax liability. In the model, the government sets not only the tax rate, but also the tax base. Moving towards a broader turnover tax base (with fewer permissible deductions than a profit tax) gives firms fewer opportunities to evade. This helps the government raise revenues, but also distorts firms’ choices and reduces production efficiency. We show that, in such a context, it is optimal for the government to sacrifice some production efficiency in return for the reduced evasion that a broader tax base offers. We characterise the optimal balance of this trade-off in terms of parameters that we estimate in an empirical application to Pakistan.
Minimum taxes on turnover: An evasion-proof tax instrument?
In the empirical application we study a tax instrument that is clearly production inefficient, but is nevertheless ubiquitous in developing countries – a minimum tax scheme whereby firms are taxed either on profits or on turnover, depending on which tax liability is larger. In the Pakistani minimum tax scheme, the tax rate on profits (that is, turnover minus costs) is equal to 35%, whereas the tax rate on turnover is 0.5% during most of the years we study. The rules of the scheme imply that firms whose profit rate (which equals profits divided by turnover) falls below the ratio of the two tax rates (0.5 divided by 35, equaling 1.43%) switch from the profit tax to the turnover tax. We use the quasi-experimental variation generated by this threshold combined with administrative tax records on all taxpaying corporations in Pakistan between 2006 and 2010 to estimate the extent to which turnover taxation reduces evasion.
The idea of our approach is as follows. The minimum tax threshold creates a discontinuity – a so-called kink point – where firms switch between the profit and turnover tax regimes. At this threshold, both the tax rate and the tax base change discontinuously, giving firms a strong incentive to locate at or around the kink. Consider, for example, a firm with a true profit rate of 3%. To evade taxes, the firm can over-report its costs so that its reported profit rate falls to about 1.4%. At this point, the firm will be required to pay taxes on turnover. Any further exaggeration of costs has no impact on its tax liability. The firm could further reduce its tax liability by also under-reporting sales, but that is usually more difficult than over-reporting costs. For example, while a factory's output may be readily observable, the amount that it pays its accountants is much more difficult to observe. When firms respond to the differential evasion incentives generated by the minimum tax, we should expect to see that a disproportionate number of firms report profit rates close to the kink, where the tax liabilities on profit and turnover are equal, and firms have reduced their tax liability as much as possible.
This is exactly what we observe when plotting the distribution of firms’ reported profit rates around the kink as shown in Figure I. Panel A of the figure shows the profit rate distribution for firms in 2006, 2007 and 2009, three years in which the turnover tax rate was equal to 0.5%. We see that a disproportionately large number of firms report a profit rate close to the kink point at 1.43%. Moreover, this excess mass moves across time and across firms with the location of the kink as shown on Panels B and C. In 2010, when the turnover tax is increased to 1%, the kink moves to a profit rate of 2.86%. Similarly, recently incorporated firms are eligible for a reduced profit tax rate of 20%, moving the kink to a profit rate of 2.5%. In both cases, the excess mass in the profit rate distribution moves to the new location of the kink.
Figure 1. Profit rate distribution of taxpaying corporations
Estimating evasion responses
Bunching of taxpayers around thresholds where the tax rate jumps discontinuously has been used in recent studies to estimate behavioral responses to taxes (see Kleven 2015). Extending the methodology to our setting where both the tax rate and the tax base jump at the threshold, we use bunching at the minimum tax kink to estimate the evasion response to turnover taxation. Note that this sharp bunching around the kink must be driven by evasion rather than real output changes. While being taxed on turnover rather than profits might lead some firms to reduce output, which could also generate bunching at the kink, this effect must be small due to the fact that the turnover tax is levied at a very low rate. Only the firms’ tax evasion behaviour, combined with the differential ease of misreporting profits and turnover, can explain the large and sharp bunching we observe.
Based on the empirical patterns in Figure I, we estimate that the turnover tax in Pakistan reduces evasion by as much as 60-70% of corporate income as compared to the profit tax. Hence, turnover taxes can provide a useful backstop for corporate profit taxes in countries with widespread evasion.
Our model and evidence allow us to rationalise the prevalence of production inefficient policies in low- and middle-income countries. The analysis highlights the dangers of relying on policy prescriptions derived from traditional public finance models, most of which were developed for high-income contexts. Incorporating our empirical estimates in our model, we find that optimal policy in a context like Pakistan is much closer to the production inefficient turnover tax than the production efficient profit tax. Indeed, switching from a pure profit tax to a pure turnover tax can increase revenue by up to 74% without reducing aggregate profits, thus creating substantial welfare gains.
Of course, our model and estimates take the government’s ability to enforce taxes as given. Our large and striking results suggest significant returns to increasing the tax enforcement capacity of developing countries. However, the evolution of fiscal capacity tends to be slow-moving and depends on factors that are beyond direct policy control (Kleven et al. 2015, Jensen 2015). While fiscal capacity is still low, using alternative tax instruments such as turnover taxes can give governments the fiscal space to invest in public goods and develop their economies.
Best, M, A Brockmeyer, H Kleven, J Spinnewijn, and M Waseem (2015), “Production vs Revenue Efficiency with Limited Tax Capacity: Theory and Evidence from Pakistan”, Journal of Political Economy 123(6): 1311-1355.
Jensen, A (2015), “Employment Structure and the Rise of the Modern Tax System”, LSE Working Paper, December.
Kleven, H (2016), “Bunching”, forthcoming in the Annual Review of Economics 8.
Kleven, H, C Kreiner, and E Saez (2016), “Why Can Modern Governments Tax So Much? An Agency Model of Firms as Fiscal Intermediaries”, forthcoming in Economica.
1 Tax revenue as share of GDP – IMF World Economic Outlook 2013, number of tax-filing corporations: authors’ calculations.
2 This is also recognised by the UN’s Sustainable Development Goals, calling governments to ‘Strengthen domestic resource mobilisation, including through international support to developing countries, to improve domestic capacity for tax and other revenue collection’.