Tax evasion, by its very nature, is difficult to observe. In this paper, we present a case study of tax evasion in China. The novel feature of our approach is that at a very disaggregated level of individual products, we can measure evasion relatively precisely, by comparing the values that China reports as imports from Hong Kong, with what Hong Kong reports as exports to China. We can match up this ‘evasion gap’ with the tariff (and VAT tax) schedule at the product level. The result is striking: using the data in 1998, we find that on average, a 1 percent increase in the tax rate results in a 3 percent increase in evasion. The result is similar when a first-difference specification is used with data in 1997 and 1998. This relationship is nonlinear: the evasion elasticity is larger at high tax levels. Furthermore, the evasion gap is negatively correlated with the tax rates on closely related products, suggesting that part of the evasion takes place by mis-reporting the type of imports, in addition to under-reporting the value of imports. This effect is even more pronounced when the evasion gap is measured using quantities rather than values.
JEL Classification: H2, F1