Measuring true sales and underreporting with matched firm-level survey and tax-office data

Abstract

This paper uses firm-level survey data matched with official tax records to estimate the unobserved true sales of formal firms in Mongolia. Taking into account firm-level incentives to comply with taxes and a production function technology linking unobserved true sales with observable firm-level production characteristics, the authors derive a multiple-indicators, multiple-causes model predicting unobserved true sales. Comparing predicted true sales with sales reported to the tax office, the analysis finds that 38.6 percent of firm-level sales are underreported. It also finds evidence that firm-level survey data suffer from significant underreporting. Finally, the paper compares this approach with two alternative approaches to measuring underreporting by firms.