We conducted a mail questionnaire survey using both the randomized response (RR) technique and the direct questioning (DQ) technique to directly estimate the prevalence and type of income tax evasion. We also assessed the effectiveness of the RR technique in reducing response and non-response biases and examined the relationship between tax evasion and key demographic variables. Of the respondents completing the RR survey instrument, 5.5% admitted tax evasion by under-reporting income, and 6.5% admitted tax evasion by over-claiming deductions. The corresponding proportions obtained from the DQ survey instrument were 1.7% and 4.2% respectively. The RR technique was ineffective in reducing non-response bias, but the estimated proportions of tax evasion obtained by the RR technique are higher than those obtained by the DQ technique. A relationship was found between the demographic variables examined and tax evasion. However, interpretation of the results was restricted by the lack of statistical significance of the differences.