More insecure and less paid? The effect of perceived job insecurity on wage distribution

ABSTRACT

This article employs a Counterfactual Decomposition Analysis (CDA) using both a semi-parametric and a non-parametric method to examine the pay gap due to perceived job insecurity over the entire wage distribution of dependent workforce in Italy. Using the 2015 INAPP Survey on Quality of Work, our results exhibit a mirror J-shaped pattern in the pay gap between secure and insecure workers, with a significant sticky floor effect, i.e. a greater effect of job insecurity at the lowest quantiles. This pattern is mainly due to the characteristics effect, while the relative incidence of the coefficient component accounts roughly for 22% up to 36% of the total difference, being more relevant at the bottom of the wage distribution.

KEYWORDS:

• Counterfactual distribution
• semi-parametric methodology
• non-parametric methodology
• job insecurity
• quantile regression

JEL CLASSIFICATION:

• J31
• J82
• C14
• D91
• J24
• C31

Acknowledgments

We thank Giovanni Gallo, Stefanie Gundert, Paolo Severati, Wen Wang, participants at seminar in the University of Modena and Reggio Emilia, at the Annual EAEPE Conference, Nice, 6-8 September 2018, at BSA Work, Employment and Society Conference, Belfast 12-14 September 2018, at Counterfactual Methods for Policy Impact Evaluation (COMPIE 2018) Conference, Berlin 26-28 September 2018, at II Annual Conference of Astril “ Technology, Employement, Institutions” University Roma Tre, 13-14 December 2018, at “Population Days 2019 at a glance”, Italian Association for the Study of Population (AISP), Bocconi University, Milan, 24-26 January 2019, at national conference of Italian Society of Economic Sociology (SISEC), University of Naples, Federico II for useful comments. The views and opinions expressed in this article are those of the authors and do not necessarily reflect those of the Institutions. Corresponding author. E-mail: [email protected]

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

¹ See the Testimony of Chairman Alan Greenspan Federal Reserve’s semi-annual monetary policy report Before the Committee on Banking, Housing, and Urban Affairs, U.S. Senate 26 February 1997. https://www.federalreserve.gov/boarddocs/hh/1997/february/testimony.htm.

² Another type is the ‘detailed’ decomposition, where the interest lies in estimating the contribution of each covariate to the overall difference.

³ Results from the probit model are shown in Table A1.

⁴ Many kernel functions can be used to the scope. In our exercise, we chose the Gaussian kernel evaluated at ($w-w_i$) given the bandwidth h. Our choice of the kernel is due to its property of monotonicity of peaks and valleys w.r.t. changes in the smoothing parameters, which proves to be useful when comparing distributions (Sheather 2004). For what concerns the bandwidth, our choice has fallen on the Cross Validation (CV) method: it is suitable as there is no need to make assumptions about the smoothness to which the unknown density belongs (Loader 1999).

⁵ The effect from category 0 to 1 and from 1 to 2 is increasing), thus suggesting a non-linear relation with the wage.

⁶ The effect of the direct (i.e. of the worker) educational attainment is found to increase salary of about 5.3 percentage points, while the indirect (i.e. of their fathers) increases the salary of 2% percentage points.

⁷ The insight here is that, being the dependent variable a self-perceived JI, it already probably contains a sort of self-selection term: therefore the distortion due to self-selection is low.