Understanding heterogeneity in tax compliance

On 14 January 2016, we had a talk by Raymond Duch of the Nuffield Centre for Experimental Social Sciences (CESS). The title was Why we Cheat: Experimental Evidence on Tax Compliance [paper, video]. In an experimental setting, they find that high performance ("rich") experimental subjects are more likely to cheat on tax payments.

Understanding the results

I felt a key problem of the setup was the absence of coercion and punishment. Paying taxes is, at heart, about the coercive power of the State. Nobody wants to pay taxes; it is only the fear of punishment which makes you pay taxes.

I would interpret his results as saying: In a cooperative, high performance people are more likely to not pay in a fraction of their output to the common pool. The paper is about the behaviour of people in voluntary arrangements, and not tax compliance.

It perhaps suggests that a poll tax comes more naturally to humans as compared with a tax which is a fraction of the income. Imagine that you were in a cooperative: it's easier to think of everyone in the cooperative putting up Rs.X, rather than of everyone putting in x% of their income.

Heterogeneity in tax compliance

Turning to tax compliance, let's think of a simple setup where there is income $y_i$, a flat tax rate $\tau$, actual tax payment $T_i$, a $p$ probability of getting caught and a punishment $\lambda$ times larger than the tax shortfall $\tau y_i - T_i$. In this setup, the key parameter which will shape compliance is risk aversion. People who are more risk averse will comply more.

In countries where $p$ is high, the outcome will have high compliance. As $p$ becomes higher, the distribution of compliance will collapse into a point mass. When $p$ is low, and for certain kinds of distributions of risk aversion, we will get economically significant heterogeneity in tax compliance.

Low risk aversion is likely to be correlated with high performance. So we may endup with a simple correlation where high performance people are more likely to cheat on taxes. This is perhaps less spicy than meets the eye.

Tax compliance by firms in India

Consider the Indian operations of a multinational corporation versus an Indian family business. There is evidence that tax compliance by multinationals is superior. In my understanding, two things are going on.

The first is that $\lambda$ is not a constant; it is lower for Indian firms, as they are better able to manage the non-rule-of-law environment in the tax administration.

The second issue is risk aversion. MNCs tend to be very risk averse and look for safe interpretations of law. This may be related to multiple layers of bureaucracy and the principal-agent problems between the shareholder and the manager. Global compliance teams have a cover-your-ass attitude and force the local operations to play very safe. In contrast, Indian business houses tend to be take more aggressive interpretations of the law. They know this is risky and they walk into it with their eyes open.

There isn't much of a low-compliance-correlates-with-performance story here, as some of the best run companies in India (the MNCs) have the highest tax compliance. The empirical regularity actually runs in the reverse direction.