IMAGEN “Macroeconomic consequences of Gender discrimination: state of the art” Universidad de Santiago de Compostela - July 18-19 th 2013 July 18 19 th 2013 Universidad de Santiago de Compostela Are women in supervisory positions more discriminated against? A multinomial di i i t d i t? A lti i l approach pp - Marco Biagetti * and Sergio Scicchitano * - *Ministry for Economic Development, Department for the Development and the Economic Cohesion. The views expressed in this article are those of the authors and, in particular, do not necessarily reflect those of the Ministry of Economic Development.
Outline Outline 1. Introduction: motivation and goals 2 Previous literature 2. Previous literature 3. Our procedure: 2+1 steps using EU-SILC data 4 4. The model h d l 5. Results 6. Conclusions: possible explanations, policy indications and further research indications and further research
Motivation: gender wage gap at a supervisory level i l l • It is well known that equality between men and women is a worldwide p priority y in the current economic policy, p y, particularly for the European Commission (EC), • Nevertheless, it is also still evident that, despite women’s higher education than men’s and their increasing participation in the labor market, they are still largely under-represented in th hi h hi the high hierarchical positions and suffer from a significant d ff f i ifi t hi l iti pay gap particularly at a supervisory level (EC 2003, 2009). 3
Goals of the paper: 3 questions p p q 1) Are gender wage gap and discrimination against ) g g g g p women higher among supervisory positions than among non supervisory? g p y 2) Is there any selectivity bias for both men and women for participating to the labor market and working as for participating to the labor market and working as supervisors or non supervisors? 3) Therefore is a two stages approach more precise as 3) Therefore, is a two stages approach more precise, as well as more suitable, than a single OLS in order to give us an unbiased answer to the first two questions? give us an unbiased answer to the first two questions? 4
Key-point in the procedure To control for a selectivity bias affecting the sample selection: two main reasons: 1. When general women’s wages are examined, the possibility that unobservable factors influence selection into the sample usually remains a serious obstacle . 2. When the GWG for women in a supervisory position is examined and compared with that of non supervisor women, the selection term becomes much more relevant in terms of unbiased estimates. This fact occurs for a couple of other reasons This fact occurs for a couple of other reasons. 1. First, if we are interested in examining women’s behavior in a particular working condition x, the most appropriate potential outcome is even threefold - working in x, not working in x, not working at all - so that we have to compute two selection terms (Rodgers 2004). 2 2. When that particular working condition x is the supervisory position where When that particular working condition x is the supervisory position, where women are traditionally underrepresented, the Inverse of the Mill’s Ratio 5 (IMR) is expected to be particularly significant.
Previous empirical literature: the GWG among managers GWG among managers • Despite the general GWG having been largely investigated, the analysis of a wage differential among managerial workers has certainly drawn much less attention differential among managerial workers has certainly drawn much less attention. • Bertrand and Hallock (2001) study the GWG in the top-level management of U.S. corporations. The GWG was at least 45%. • Muñoz-Bullón (2010) finds similar results: the GWG was equal to 47%, • Bertrand et al (2010) study gender differences in the career dynamics of MBAs who graduated from a top US business school from 1990 to 2006. The GWG is g p equal to almost 60 log points a decade after completing the MBA. • Baldwin et al (2001), Oaxaca et al. (2013): hierarchical segregation. • Scicchitano (2012) and Biagetti and Scicchitano (2013 forthcoming) apply a Scicchitano (2012) and Biagetti and Scicchitano (2013 forthcoming) apply a counterfactual decomposition analysis to investigate the GWG among managers in UK and Spain respectively: – UK : clear a U-shaped pattern, both sign. sticky floor and glass ceiling effects. UK : clear a U shaped pattern both sign sticky floor and glass ceiling effects – Spain : the glass ceiling constitutes a bigger problem than the sticky floor.
Previous empirical lit: sample selection p p • A couple of papers examine the GWG for the managerial position after controlling for sample selection after controlling for sample selection • Holst and Busch (2009) demonstrate that in Germany, after taking into account a twofold selection effect for a leadership position, only p p , y one third of the gender wage differential could be explained. • Watson (2009) studies the GWG among managers in Australia by g g y also controlling for a threefold selection bias in the labor market participation as a manager or non manager. He finds that female managers earn about 25 per cent less than their male counterparts and b t 25 t l th th i l t t d somewhere between 70 and 90 per cent of this wage gap cannot be explained explained. • However, there is no study comparing GWG while at the same time analyzing the discrimination against women, in both leadership and a a y g t e d sc at o aga st wo e , bot eade s p a d non leadership positions. 7
Previous empirical literature in Italy • Rustichelli (2005) evaluates a GWG between women and men’s daily wages equal to 39% on average, 26.9% of which is attributed to differences in rewards. • Picchio and Mussida (2010) find that, when a control for sample selection is made, the GWG significantly increases at the top of the wage distribution, where it jumps from 17.3% to 24.5%. • Biagetti apply l a counterfactual t f t l Bi tti and d Scicchitano S i hit (2012) (2012) decomposition approach using QR to the wage distribution of managerial workforce in Italy Evidence of both significant sticky managerial workforce in Italy. Evidence of both significant sticky floor and glass ceiling effects • However, although in this country the empirical literature on the , g y p general GWG is somewhat extensive, the comparison between hierarchical positions after controlling for a possible sample selection bias is almost completely ignored . 8
Procedure • As we noted above a crucial point is to control for the selectivity bias • In doing so, we choose to adopt a two ( +1) step procedure . In the g , p ( ) p p first step, we correct for a threefold selectivity bias through a multinomial logit . This selectivity may affect the participation to the labor market as non supervisory or supervisory for both males and females. • In the second step we estimate the wage equation by taking into I th d t ti t th b t ki i t ti account the eventual significant selectivity terms. • This method makes us able to verify the unadjusted and the • This method makes us able to verify the unadjusted and the (multiple-selection) adjusted GWG between supervisory and non supervisory jobs. Furthermore, we can verify whether or not the supervisory jobs. Furthermore, we can verify whether or not the discrimination against women is higher among managerial positions compared to the non managerial through the B-O decomposition . 9
Data Data • We use last European dataset, the Community Statistics on Income and Living Conditions g survey (EU-SILC), whose last wave, for which the reference year is 2007 is available which the reference year is 2007 , is available since march 2009. • To the best of our knowledge, this dataset has never been used yet in studying the analysis y y g y of the GWG between supervisory and non supervisory positions supervisory positions. 10
Definition of manager/supervisor g p • Following Acemoglu and Newman (2002) and Beaudry and Francois (2010) we define the manager/supervisor as the employee whose job p y j g p description is associated to the responsibility in the the organization organization and and monitoring monitoring of of other other employees. • EU-SILC identifies such a working condition with a supervisory p y role where supervisory p y responsibility includes formal responsibility for coordinating a group of other employees coordinating a group of other employees 11
Methodology: first stage gy g Threefold potential outcome, first stage (selection stage): (NW, W, WM) => that is why we divide our dependent variable into 3 categories (i.e. multinomial logit) , to get the following probabilities and compute IMRs h f ll i b bili i d Z is matrix containing personal categorical ( ) 1 1 variables: age, level of variables: age level of = = + ε Pr 0 ( ) ( ) education attained, Y + β + β consensual union, ij ij 1 exp exp Z Z health, household type 1 2 ij ij (linked to the number (linked to the number ( ) of persons and β exp ( ( ) ) Z dependent children = = + + ε ε k ij Pr Pr living in a family) and ( ( ) ) ( ( ) ) Y Y k k citizenship citizenship. + β + β ij ij ij ij 1 exp exp Z Z 1 2 ij ij i=ith individual k=1,2 i.e. W and WM and j=males, females ε ~N(0,1)
Recommend
More recommend
