ENBIS: European Network for Business and Industrial Statistics

Mining the submerged: estimating the probability of moonlighting in Italian building industry

21 September 2009, 16:05 – 16:25

Abstract

Submitted by

Maria Felice Arezzo

Authors

Maria Felice Arezzo

Affiliation

Sapienza University of Rome

Abstract

In 2008 EU's employment chief said Europe's black market economy "undermines the financing" of national social security programs and hinders efforts to boost economic growth. There are no signs that the phenomenon is decreasing. Indeed in certain sectors and certain forms of work it appears to be growing. We need to step up our approach and take more decisive action across the EU." This paper goes in the direction of shedding light on the phenomenon by using statistical models to detect which companies are more likely to hire irregular workers.
Moonlighting could be referred either as a multiple job worker or as a single job worker in an irregular position because he is unknown to the Government since he was not regularly registered as an employee. In this paper we focus our attention on the latter.
The idea is that it is possible to estimate the probability of using irregular workers by relating Italian National Social Security Institute inspections output (expressed in terms of presence/absence of irregular workers) to a wide variety of firm’s characteristics (such as localization, labour productivity, labour cost, and so forth).
The probability of moonlighting is estimated according to three different models: logistic regression, classification trees (CART) and random forests. The main results obtained are:
a) territorial variables (region and province) are the most important to predict moonlighting;
b) Legal structure is also a crucial variable; in particular sole proprietorships have a higher moonlighting probability;
c) Labor cost systematically have an impact on moonlighting probability;
d) It's much easier to classify correctly firms which don't use irregular workers rather than companies that do.

Keywords: Logistic regression, CART, Random forests, Moonlighting

View paper

Return to programme