Mixed data methods of estimating undeclared earnings, with application to Latvia

This study develops two methods of estimating the proportion of envelope wage receivers and the share of envelope earnings in the wage bill. Both methods are applicable for the whole economy, as well as by sectors and by socio-economic groups, if some combination of survey and administrative earnings data is available. The Mixed Data Method (MDM) estimates envelope wages by comparing survey and administrative data for the same employee. In addition, MDM applies a matching procedure to produce estimates in cases of survey non-response. MDM is suitable for large survey datasets with integrated (or matched) administrative data, as is the case for the national versions of EU-SILC in many countries, including Latvia. According to the MDM estimates based on the Latvian EU-SILC, the average envelope share across all employees (including those declaring all earnings) dropped from 30% in 2007 to 23% in 2011-2012 and 21% in 2015-2016. The envelope share in the wage bill is higher for low-income workers, but the total amount of undeclared earnings is larger among high-income employees. The Distribution Matching Method (DMM) is less demanding in terms of data but provides only lower-bound estimates. DMM assumes that, for some measure of earnings and some set of intervals, administrative data on distribution of employees by officially declared earnings are available along with representative survey data on distribution of employees by total (declared plus undeclared) earnings. Completely informal employees should be either excluded from the survey or identifiable among respondents. In this case, DMM provides lower bounds for the share of registered employees receiving envelope wages. Moreover, the weighted average of sectoral, regional or age-specific lower bounds usually improves the economy-wide lower bound estimated directly.