Are women over-represented in dead-end jobs? A Swedish study using empirically derived measures of dead-end jobs

Abstract

It has been claimed that women experience fewer career opportunities than men do mainly because they are over-represented in ‘Dead-end Jobs’ (DEJs). Using Swedish panel data covering 1.1 million employees with the same employer in 1999 and 2003, measures of DEJ are empirically derived from analyses of wage mobility. The results indicate that women are over-represented in DEJs, especially in the public sector. The findings are interesting from (a) a methodological viewpoint, as it is indicated that the career opportunities associated with occupations can be indicated using one measure for both men and women, (b) the glass ceiling perspective, which arguably under-emphasizes gender inequality in relation to low positions, and (c) a class perspective, which basically ignores gender and sector in explaining career chances.

Appendix: construction of the AWM scores

The following calculations (i)–(v) are made separately for each combination of occupation (i), sector (j) and sex (k), i.e. theoretically for 452 combinations (113*2*2). However, all combinations do not exist. The purpose of calculations vi-viii is to produce sex-neutral AWM scores, i.e. to calculate one index score for each occupation in each sector, as well as sex and sector-neutral scores, i.e. one index score per occupation.

The general wages for 1999 and 2003 are regressed by the following two equations in the cross-sectional data sets:

• (i)
  $$ \ln ({\hbox{W}}_{99} ){\hbox{ }} = {\hbox{ }}\alpha _1 {\hbox{ }} + {\hbox{ }}\beta _{11} *{\hbox{age }} + {\hbox{ }}\beta _{12} *{\hbox{age}}^2 {\hbox{ }} + {\hbox{ }}\varepsilon $$

  and

  $$ \ln ({\hbox{W}}_{03} ){\hbox{ }} = {\hbox{ }}\alpha _2 {\hbox{ }} + {\hbox{ }}\beta _{21} *{\hbox{age }} + {\hbox{ }}\beta _{22} *{\hbox{age}}^2 {\hbox{ }} + {\hbox{ }}\varepsilon $$

The coefficients from (i) are used to estimate the mean wage 1999 and the mean wage 2003 (as the mean of a person at the age of 30, 40 and 50):

• (ii)
  $$ \widehat{\hbox{W}}_{99} {\hbox{ }} = {\hbox{ }}(\exp {\hbox{ }}(\alpha _1 {\hbox{ }} + {\hbox{ }}\beta _{11} *30{\hbox{ }} + {\hbox{ }}\beta _{12} *(30)^2 ){\hbox{ }} + {\hbox{ }}\exp (\alpha _1 {\hbox{ }} + {\hbox{ }}\beta _{11} *40{\hbox{ }} + {\hbox{ }}\beta _{12} *(40)^2 ){\hbox{ }} + {\hbox{ }}\exp (\alpha _1 {\hbox{ }} + {\hbox{ }}\beta _{11} *50{\hbox{ }} + {\hbox{ }}\beta _{12} *(50)^2 ))/3 $$

  and

  $$ \widehat{\hbox{W}}_{03} {\hbox{ }} = (\exp (\alpha _2 {\hbox{ }} + {\hbox{ }}\beta _{21} *30{\hbox{ }} + {\hbox{ }}\beta _{22} *(30)^2 ){\hbox{ }} + {\hbox{ }}\exp (\alpha _2 {\hbox{ }} + {\hbox{ }}\beta _{21} *40{\hbox{ }} + {\hbox{ }}\beta _{22} *(40)^2 ){\hbox{ }} + {\hbox{ }}\exp (\alpha _2 {\hbox{ }} + {\hbox{ }}\beta _{21} *50{\hbox{ }} + {\hbox{ }}22*(50)^2 ))/{\hbox{3}} $$

The estimated mean wages from (ii) are used to weight the wage (WW) measure for 2003 in the panel data set.

• (iii)
  $$ {\hbox{WW}}_{03} {\hbox{ }} = {\hbox{ W}}_{03} *(\widehat{\hbox{W}}_{99} /\widehat{\hbox{W}}_{03} ) $$

The average wage mobility is regressed by the following equation (panel data set):

• (iv)
  $$ \ln ({\hbox{WW}}_{03} ){\hbox{ }} = {\hbox{ }}\alpha _3 {\hbox{ }} + {\hbox{ }}\beta _{31} *{\hbox{age }} + {\hbox{ }}\beta _{32} *{\hbox{age}}^2 {\hbox{ }} + {\hbox{ }}\beta _{33} *{\hbox{ }}\ln ({\hbox{W}}_{99} ){\hbox{ }} + {\hbox{ }}\varepsilon $$

The coefficients from (iv) are used to estimate the average wage mobility (AWM) (as the mean percent of a person at the age of 30, 40 and 50 with a wage of 20,000 Swedish kronor in 1999; 20,000 is quite close to the mean wage, and the use of other levels only affects the AWM scores very marginally).

• (v)
  $$ \begin{aligned}{} {\hbox{AWM }} = & \quad \quad {\hbox{ }}if{\hbox{ }}N\, > \,99 \\ & \quad \quad {\hbox{ }}((((\exp (\alpha _3 {\hbox{ }} + {\hbox{ }}\beta _{31} *30{\hbox{ }} + {\hbox{ }}\beta _{32} *(30)^2 {\hbox{ }} + {\hbox{ }}\beta _{33} *{\hbox{ }}\ln (20.000)){\hbox{ }} + \\ & \quad \quad \exp (\alpha _3 {\hbox{ }} + {\hbox{ }}\beta _{31} *40{\hbox{ }} + {\hbox{ }}\beta _{32} *(40)^2 {\hbox{ }} + {\hbox{ }}\beta _{33} *{\hbox{ }}\ln (20.000)){\hbox{ }} + \exp (\alpha _3 {\hbox{ }} + {\hbox{ }}\beta _{31} *50{\hbox{ }} + {\hbox{ }}\beta _{32} *(50)^2 {\hbox{ }} + {\hbox{ }} \\ & \beta _{33} *{\hbox{ }}\ln (20.000)))/3)/20.000) - 1)*100 \\ \end{aligned} $$

The AWM equal weight score (AWM_e) for each occupation (separately for the private and the public sector) is calculated from the sex-specific AWM scores (_w: women; _m: men). N is the number of men and women in the panel and N ₀₃ is the number of men and women in the cross-sectional data set from 2003.

• (vi)
  $$ \begin{aligned}{} & {\hbox{AWM}}_{\hbox{e}} = {\hbox{ }} {\hbox{AWM}}_{\rm{w}} {\hbox{ if }}N_{\rm{m}} \, < \,99{\hbox{ or if }}N_{03\_{\rm{m}}} /(N_{03\_{\rm{m}}} {\hbox{ }} + {\hbox{ }}N_{03\_{\rm{w}}} )\, < \,0.05 \\ & {\hbox{AWM}}_{\rm{m}} {\hbox{ if }}N_{\rm{w}} \, < \,99{\hbox{ or if }}N_{03\_\rm{w}} /(N_{03\_{\rm{m}}} {\hbox{ }} + {\hbox{ }}N_{03\_{\rm{w}}} )\, < \,0.05 \\ & {\hbox{else }}({\hbox{AWM}}_{\rm{w}} {\hbox{ }} + {\hbox{ AWM}}_{\rm m} )/2. \\ & \\ \end{aligned} $$

The AWM different weight scores (AWM_d) for each occupation (separately for the private and the public sector) are calculated from the sex-specific AWM scores and from the number of women and men in the cross-sectional data set 2003:

• (vii)
  $$ \begin{aligned}{} & {\hbox{AWM}}_{\hbox{d}} {\hbox{ }} = {\hbox{ }} {\hbox{AWM}}_{\hbox{w}} {\hbox{ if }}N_{\hbox{m}} \, < \,99 \\ & {\hbox{ }} {\hbox{AWM}}_{\hbox{m}} {\hbox{ if }}N_{\hbox{w}} \, < \,99 \\ & {\hbox{ }} {\hbox{else }}(({\hbox{AWM}}_{\hbox{w}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{w}}} ){\hbox{ }} + ({\hbox{AWM }}_{\hbox{m}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{m}}} ))/(N_{03\_{\hbox{w}}} {\hbox{ }} + {\hbox{ }}N_{03\_{\hbox{m}}} ) \\ & \\ \end{aligned} $$

Both AWM_e and AWM_d (here denoted as AWM_{e_pr} AWM_{e_pu} AWM_{d_pr} AWM_{d_pu,} where _pr is private sector and _pu is public) are made sector neutral (AWM_{e_sn} and AWM_{d_sn}) by number of women and men in the cross-sectional data set from 2003 in the private vs. public sector (N _pr and N _pu):

• (viii)
  $$ \begin{aligned}{} & {\hbox{AWM}}_{{\hbox{e\_sn}}} {\hbox{ }} = \quad {\hbox{AWM}}_{{\hbox{e\_pr}}} {\hbox{ if }}N_{{\hbox{pu}}} \, < \,99 \\ & {\hbox{ }}\quad {\hbox{AWM}}_{{\hbox{e\_pu}}} {\hbox{ if }}N_{{\hbox{pr}}} \, < \,99 \\ & {\hbox{ }}\quad {\hbox{else }}(({\hbox{AWM}}_{{\hbox{e\_pr}}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{pr}}} ){\hbox{ }} + {\hbox{ }}({\hbox{AWM}}_{{\hbox{e\_pu}}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{pu}}} ))/(N_{03\_{\hbox{pr}}} {\hbox{ }} + {\hbox{ }}N_{03\_{\hbox{pu}}} ) \\ & \\ \end{aligned} $$

  and

  $$ \begin{aligned}{} & {\hbox{AWM}}_{{\hbox{d\_sn}}} {\hbox{ }} = \quad {\hbox{AWM}}_{{\hbox{d\_pr}}} {\hbox{ if }}N_{{\hbox{pu}}} < 99 \\ & {\hbox{ }}\quad {\hbox{AWM}}_{{\hbox{d\_pu}}} {\hbox{ if }}N_{{\hbox{pr}}} < 99 \\ & {\hbox{ }}\quad {\hbox{else }}(({\hbox{AWM}}_{{\hbox{d\_pr}}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{pr}}} ){\hbox{ }} + {\hbox{ }}({\hbox{AWM}}_{{\hbox{d\_pu}}} {\hbox{ }}*{\hbox{ }}N_{03\_{\hbox{pu}}} ))/(N_{03\_{\hbox{pr}}} {\hbox{ }} + {\hbox{ }}N_{03\_{\hbox{pu}}} ) \\ & \\ \end{aligned} $$