The Effects of Immigration on Labour Tax Avoidance: An Empirical Spatial Analysis

Abstract

We investigate whether the geographic concentration of non-EU immigrants in the various Italian provinces affects labour tax avoidance (LTAV) practices adopted by firms located in the same provinces, as well as in the neighbouring provinces, and operating in construction and agriculture industries that mostly employ immigrants in Italy. For this purpose, we develop a LTAV proxy based on the financial accounting information of a sample of 993,606 firm-years, disseminated throughout the 108 Italian provinces, over the period 2008–2016. Our results, based on a Spatial Durbin Model panel regression, reveal a statistically significant positive association between the concentration of non-EU immigrants and LTAV at province level, as well as the presence of spillover effects among neighbouring provinces. Our findings are robust to several additional analyses, including instrumental variable estimations. Our study provides empirical support to previous structuralist or marginalization theories holding that socioeconomically marginalized groups, such as non-EU immigrants, are more likely to be involved in labour exploitation practices, which could underlie our LTAV outcomes. Furthermore, it supports the need for tax authorities to strengthen labour inspections, coordinated at national level, especially in those contexts where non-EU immigrants are mostly employed. On the other hand, a greater social integration, assistance, and recognition of rights of immigrants may help to alleviate their situation of weakness that makes them more vulnerable to LTAV practices. Finally, tackling LTAV, associated with the underemployment of immigrants, may prevent its negative effects for society arising from the reduction of public resources to sustain the social welfare and finance public goods and services.

Appendix

Definition of Variables

Variable definition of Eq. (3)

$$AbSSCs_{i,t} = \beta_{0} + \mathop \sum \limits_{r} \beta_{r} PROVINCE_{i,t}^{r} + \mathop \sum \limits_{k} \beta_{k } CONTROLS_{i,t}^{k} + \mathop \sum \limits_{s} \beta_{s} INDUSTRY_{i,t}^{s} + \varepsilon_{i,t}$$

AbSSCs abnormal SSCs equal to residuals from Eq. (2) simultaneously estimated with Eq. (1), PROVINCE dummy variable for each of 107 Italian provinces, CONTROLS firm-level control variables of Eq. (3) regression model: SIZE natural logarithm of total assets in thousands of euros, AGE age of the firm in years, LEVER total debt divided by total assets, CAPINT net fixed assets and net intangible assets divided by total assets, ROA net income divided by total assets, LOSS dummy variable that takes a value of 1 if the firm had two or more consecutive years of negative income including the current and 0 otherwise, GROW percentage change in net sales relative to previous year, DAC discretionary accruals estimated based on the performance-adjusted modified Jones model (Ravenda et al. 2018), AbMATL abnormal material costs equal to residuals from the following Eq. (5) with material costs (MAT), including both raw materials and merchandise, as dependent variable, estimated cross-sectionally for each two-digit NACE industry-year

$$\frac{{MAT_{i,t} (SERV_{i,t} )}}{{\ln (TA_{i,t - 1} )}} = \beta_{0} + \beta_{1} \frac{1}{{\ln (TA_{i,t - 1} )}} + \beta_{2} \frac{{S_{i,t} }}{{\ln (TA_{i,t - 1} )}} + \beta_{3} \frac{{\Delta S_{i,t} }}{{\ln (TA_{i,t - 1} )}} + \varepsilon_{i,t }$$

(5)

AbSERV abnormal service costs equal to residuals from Eq. (5) with service costs (SERV) as dependent variable, estimated cross-sectionally for each two-digit NACE industry-year, CASHTA cash and cash equivalents divided by total assets, ETR abnormal effective tax rate equals to industry- and size-matched GAAP ETR minus firm’s GAAP ETR, where GAAP ETR is the total tax expense divided by pre-tax income. Industry- and size-matched GAAP ETR is the average GAAP ETR for the portfolio of firms in the same quintile of total assets and the same two-digit NACE industry-year, SD_ROA standard deviation of ROA over the past four years, INVENTA inventory divided by total assets, INDUSTRY dummy variable for each three-digit NACE industry.

Variable definition of Eq. (4)

$$LTAV\_PROV_{i,t} = \rho W LTAV\_PROV_{i,t} + \beta_{1} IMMIGR_{i,t} + \mathop \sum \limits_{k} \beta_{k} CONTROLS_{i,t}^{k} + \theta_{1} W IMMIGR_{i,t} + \mathop \sum \limits_{k} \theta_{k} W CONTROLS_{i,t}^{k} + u_{i} + v_{i,t}$$

LTAV_PROV LTAV measure at province level equal to the estimated coefficients on PROVINCE in Eq. (3), W inverse distance spatial weight matrix with a threshold distance of 57.14 km, spectrally normalized, IMMIGR non-EU immigrant concentration, computed as the fraction of non-EU residents per 1000 residents in each province and year, restricted to the population between 18 and 59 years of age, and spatially differentiated from the province of Rome (source: ISTAT), CONTROLS province-level control variables of Eq. (4) regression model: DENSITY province population per km², spatially differentiated from the province of Rome (source: ISTAT), CRIME natural logarithm of crimes reported by police forces to judicial authorities per 1000 residents, spatially differentiated from the province of Rome (source: ISTAT), UNEMPL annual unemployment rate, spatially differentiated from the province of Rome (source: ISTAT), HGRSAL employee hourly gross salary (CPI deflated, 2016 equivalents), spatially differentiated from the province of Rome (source: ISTAT), ∆GDP gross domestic product growth rate, spatially differentiated from the province of Rome (source: ISTAT), u province fixed-effect (PROVINCE FE).