You are on page 1of 6

Applied Economics Letters

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rael20

Did wage inequality increase in Portugal? Yes, and


for good reasons

João Manuel Pereira

To cite this article: João Manuel Pereira (2020): Did wage inequality increase in Portugal? Yes,
and for good reasons, Applied Economics Letters, DOI: 10.1080/13504851.2020.1789057

To link to this article: https://doi.org/10.1080/13504851.2020.1789057

Published online: 02 Jul 2020.

Submit your article to this journal

Article views: 37

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=rael20
APPLIED ECONOMICS LETTERS
https://doi.org/10.1080/13504851.2020.1789057

ARTICLE

Did wage inequality increase in Portugal? Yes, and for good reasons
João Manuel Pereira
Department of Economics and CEFAGE-UE, University of Évora, Évora, Portugal

ABSTRACT KEYWORDS
As in Anglo-Saxon countries, wage inequality increased in Portugal from the beginning of the Inequality; reweighting;
1980s to the end of the 2000s. Using data from a matched employer-employee data set and wages; wage distribution
reweighting methods, we show that the increased wage inequality in Portugal is fundamentally JEL CLASSIFICATION
explained by compositional changes in the employment structure related to the modernization of J00; J31; D63
the labour market.

I. Introduction Previous studies pointed out several explanations


for the increase in wage inequality in Portugal,
Wage inequality started rising in the early 1980s,
namely, skill-biased technological change (Cardoso
mainly in Anglo-Saxon countries such as the
1998), increasing and unequal returns to education
United States, Canada and the UK. In continental
over the wage distribution (Machado and Mata
European countries (France and Germany, etc.) and
2005), changes in the employment structure, mini-
Japan inequality also increased but less. The con-
mum wages and employment polarization (Centeno
sensus that emerged from the economic literature
and Novo 2014).
was that economic shocks arising from globalization
Most previous studies cover a short or medium-
and technological change had increased the relative
term period of time, with no long-term perspective
demand for skilled workers and, consequently, wage
of the causes of the inequality change in Portugal,
inequality (Lemieux 2008). However, as labour mar- about 40 years after it started rising. This time
kets of Anglo-Saxon countries are more exposed to period is long enough for labour demand and sup-
market mechanisms than those of continental ply to adjust to economic shocks (technological
European countries, inequality increased more in change, globalization, institutional changes) and
the former group of countries than in the latter. for a long-term explanation for the emergence of
Portugal also experienced an increase in wage increased wage inequality. Therefore, the aim of
inequality, comparable to those of the Anglo-Saxon this article is to provide a long-term view of the
countries (Cardoso 1998), which continued at least causes of the increased wage inequality in Portugal.
until the end of the 2000s (Centeno and Novo 2014). Our findings suggest that the upward movement of
However, the Portuguese labour market is very dif- wage inequality is fundamentally explained by
ferent from those of Anglo-Saxon countries. In fact, compositional changes in the Portuguese employ-
the Portuguese labour market is highly regulated ment structure, such as those linked to workers’
and the wage bargaining process is centralized. So, educational level and average age and the increased
according to the consensus explanation, wage share of women in employment.
inequality in Portugal should have followed the
trend of continental European countries and not
that of Anglo-Saxon countries. In that sense, the II. Data and econometric methods
analysis of the evolution of wage inequality in We use data from Quadros de Pessoal, a matched
Portugal is very different from that of other coun- employer-employee data set. This is a mandatory
tries, as it combines elements of continental census undertaken by the Portuguese Ministry of
European countries and of Anglo-Saxon countries. Employment, which includes information on the

CONTACT João Manuel Pereira jpereira@uevora.pt Department of Economics and CEFAGE-UE, University of Évora, Évora 7000-803, Portugal
© 2020 Informa UK Limited, trading as Taylor & Francis Group
2 J. M. PEREIRA

characteristics of workers, firms and establish- increased. Finally, there was also an increased share
ments in the private sector in Portugal. This is of females in employment, which is in accordance
a widely used data set for research about the with the international trend (Olivetti and
Portuguese labour market (Card et al. 2018). Petrongolo 2016).
Table 1 displays the evolution of wage inequality To analyse the evolution of wage inequality we
in Portugal in the private sector over the period use several inequality indices (Gini, the Theil and
1985–2017 based on hourly wages computed from log wage gap differences) and the reweighting
the raw data. Overall inequality increased from approach of DiNardo, Fortin, and Lemieux
1985 to the middle/end of the 2000s and decreased (1996). By using this approach, we can divide the
after that (Gini, Theil, 90/10). However, the evolu- change in inequality into two parts, one explained
tion of wage inequality was not uniform over the by the changing prices of workers’ characteristics
wage distribution. In fact, from 1985 to 1995, (price effect) and the other explained by changes in
inequality increased both at the bottom (50/10) the employment composition (composition effect).
and at the top (90/50) of the wage distribution, That division is implemented by carrying out
but more strongly at the top. After the mid- a counterfactual exercise that combines the char-
nineties, it started decreasing at the bottom of the acteristics of workers from some period t* with the
wage distribution, but continued increasing at the returns of these characteristics in another period
top. This trend ended at the middle/end of the t. The reweighting factor (w) that allows such coun-
2000s, when wage inequality also started decreasing terfactual exercise to be implemented is given by:
at the top of the wage distribution.
PrÖT à t⇤ jxÜ 1 PrÖT à t⇤ Ü
In parallel with these changes in wage inequality, wà ⇥ (1)
the Portuguese labour market underwent an in- 1 PrÖT à t⇤ jxÜ PrÖT à t⇤ Ü
depth transformation in terms of workforce com- PrÖT à t⇤ jxÜ is the probability of a worker with
position from 1985 to 2017 (Table 2). First, the characteristics x being observed in year t*;
average level of education of Portuguese workers PrÖT à t⇤ Ü and (1-PrÖT à t⇤ Ü) are the sample pro-
increased greatly. Second, the average age level also portions of the years t* and t, respectively. We
estimate PrÖT à t⇤ jxÜ with a logit model using
a similar set of covariates as in Lemieux (2006),
Table 1. Wage inequality ratios and indexes in Portugal, gross
hourly earnings, 1985–2017.
namely, the education and age dummies present in
1985 1995 2000 2005 2007 2010 2015 2017 Table 2 and all possible interaction terms between
Gini 0.30 0.36 0.35 0.36 0.35 0.34 0.33 0.32 them; in addition, we also use tenure and tenure
Theil 0.16 0.25 0.25 0.27 0.26 0.25 0.24 0.23 squared, five dummies identifying the kind of col-
90/10 3.48 4.21 3.96 4.06 3.97 3.73 3.51 3.26
90/50 2.29 2.66 2.61 2.64 2.59 2.58 2.50 2.42 lective agreement according to which wages are set
50/10 1.52 1.58 1.52 1.53 1.53 1.44 1.40 1.35
(including no collective agreement) and seven
Source: author computations based on Quadros de Pessoal, several years. The
wage inequality ratios and indices were computed with the absolute value dummies controlling for workers’ skill level. The
of nominal hourly earnings. dependent variable is a dummy for 1985 pooling
data for 1985 and 2017.
Table 2. Means of selected variables, Portugal, 1985–2017.
1985 2007 2017
III. Results
Education (Lower secondary) 0.14 0.22 0.26
Education (Upper secondary) 0.05 0.21 0.29
Education (University degree) 0.03 0.13 0.21
Table 3 displays the change of a set of inequality
Age (16–25 years) 0.17 0.11 0.08 measures between 1985 and 2017. In line (1), we
Age (26–35 years) 0.34 0.34 0.25
Age (36–45 years) 0.25 0.29 0.32 display the change computed with the raw data. This
Age (46–55 years) 0.17 0.20 0.24 change simultaneously measures the effect of chan-
Age(≥ 56 years) 0.07 0.07 0.11
Female 0.03 0.43 0.46 ging prices and quantities of workers’ observed and
Sample size 1,294,006 2,354,460 2,320,057 unobserved characteristics. However, if we apply the
Source: idem Table 1. Note: these means represent the share of each reweighting factor (1) to the 2017 data, we obtain
characteristic in the private sector employment in Portugal, given the
selected sample. the inequality change that would be observed if the
APPLIED ECONOMICS LETTERS 3

Table 3. Wage inequality change in Portugal 1985–2017, hourly earnings (log).


90–10 90–50 50–10 Gini Theil
(1)Δ overall inequality −0.066*** (0.002) 0.054*** (0.001) −0.121*** (0.001) 0.023*** (0.001) 0.065*** (0.003)
(2) Δ overall inequality, x’s 1985 −0.382*** (0.002) −0.170*** (0.002) −0.212*** (0.001) −0.043*** (0.002) 0.018** (0.008)
(3) Δ overall inequality, x’s 2017 −0.324*** (0.003) −0.002 (0.003) −0.323*** (0.002) −0.019*** (0.001) 0.032*** (0.003)
(4) Δ Predicted inequality 0.069*** (0.002) 0.074*** (0.002) −0.005*** (0.001) 0.004*** (0.000) −0.001*** (0.000)
(5) Δ Predicted inequality, x’s 1985 −0.267*** (0.003) −0.189*** (0.003) −0.081*** (0.002) −0.061*** (0.001) −0.035*** (0.000)
(6) Δ Predicted inequality, x’s 2017 −0.153*** (0.004) 0.088*** (0.003) −0.241*** (0.003) −0.017*** (0.001) −0.001 (0.001)
(7) Δ Residual inequality −0.000 (0.001) −0.019*** (0.001) 0.018*** (0.001) ... ... . . . . . ..
(8) Δ Residual inequality, x’s 1985 −0.122*** (0.002) −0.074*** (0.001) −0.048*** (0.001) . . . . . .. . . . . . ..
(9) Δ Residual inequality, x’s 2017 −0.223*** (0.003) −0.110*** (0.002) −0.112*** (0.002) . . . . . .. . . . . . ..
Source: idem Table 1. Notes: bootstrapped standard errors (100 reps.) are in parentheses. (***), (**), (*) significant at 1%, 5% and 10% significance level,
respectively.

level of workers’ characteristics was that of 1985 If workers’ characteristics were fixed at the 1985
(line 2); in line 3, we compute the inequality change level and, therefore, we evaluate the effect on
that would be observed if the level of workers’ char- inequality of changing returns to workers’ observed
acteristics in 1985 was that of 2017. The inequality and unobserved characteristics (lines: 2, 5 and 8),
change computed in lines (2) and (3) is explained by inequality would have decreased considerably
the changing prices of workers’ observed and unob- (except the Theil’s index). So the price effect is nega-
served characteristics (price effect) in different tive and what pushed up wage inequality in Portugal
employment structures. The part of the inequality was the positive composition effect linked to the
change explained by changing the level of workers’ changes in the employment structure (Table 2).
characteristics (composition effect) is the difference If, instead, the 2017 employment structure is
between line (1) and line (2) or line (3), according to adopted (lines 3, 6 and 9), the general picture is
the employment structure that is adopted. not very different. Indeed, the 90–10 difference and
Likewise, in lines (4) to (6) similar estimates are the Gini index continue to indicate that inequality
displayed, but for the predicted inequality change would have decreased, but the Theil index goes in
based on regression models for the log of hourly earn- the opposite direction for overall inequality. This is
ings explained by the same set of covariates as the logit not a surprising finding as both indices weight
model applied to estimate the reweighting factor, w. different parts of the wage distribution differently.
Predicted inequality evaluates the part of inequality So, in general, we have negative price effects and
explained by workers’ observed characteristics and positive composition effects, both for observed and
their prices. Finally, in lines (7) to (9), residual unobserved workers’ characteristics. The positive com-
inequality (hourly wages predicted hourlywagesÜ, position effects on observed characteristics are directly
or the inequality within groups, is analysed. To access related to the changes in the employment structure of
the statistical significance of the estimated changes, the Portuguese economy. Indirectly, these changes may
bootstrapped standard errors were also estimated. have also given rise to positive composition effects on
Except in the case of the 90–10 difference, the other unobserved characteristics as their dispersion typically
inequality indices (Gini and Theil) indicate that wage increases with age and education (Lemieux 2006). As
regards the negative price effects on the observed char-
inequality increased in Portugal from 1985 to 2017.
acteristics, several studies show that the returns to
That increase is fundamentally explained by changes
education have been decreasing since the middle of
in the upper part (90–50) of the wage distribution, as
the nineties (Campos and Reis 2019); there is also
in the lower part (50–10) inequality decreased. The
evidence that the returns to age had decreased from
joint effect of observed characteristics and prices on
1985 to 1995 (Machado and Mata 2005).
the change in inequality is, in general, positive,
To better understand the dynamics of the
excluding the lower tail of the wage distribution and
inequality change in Portugal from 1985 to 2017,
the Theil’s index (Table 3, line 4). The changes in the
we divided our analysis into two periods: the first
residual wage inequality are not significant, as the
from 1985 to 2007 and the second from 2007 to
changes in the lower and upper part of the wage 2017. 2007 is the year that precedes the beginning
distribution cancel each other out (7). of the greatest impacts of the last financial crisis; it
4 J. M. PEREIRA

Table 4. Wage inequality change in Portugal 1985–2007, hourly earnings (log).


90–10 90–50 50–10 Gini Theil
(1)Δ overall inequality 0.132*** (0.002) 0.122*** (0.002) 0.009*** (0.001) 0.058*** (0.000) 0.096*** (0.001)
(2) Δ overall inequality, x’s 1985 −0.021*** (0.002) 0.014*** (0.002) −0.035*** (0.001) 0.017*** (0.000) 0.042*** (0.002)
(3) Δ overall inequality, x’s 2007 −0.030*** (0.002) 0.055*** (0.002) −0.085*** (0.001) 0.027*** (0.001) 0.069*** (0.001)
(4) Δ Predicted inequality 0.174*** (0.002) 0.140*** (0.002) 0.034*** (0.001) 0.036*** (0.000) 0.033*** (0.000)
(5) Δ Predicted inequality, x’s 1985 0.006*** (0.002) 0.012*** (0.002) −0.006*** (0.001) 0.005*** (0.000) 0.013*** (0.000)
(6) Δ Predicted inequality, x’s 2007 0.010*** (0.003) 0.076 (0.003) −0.067*** (0.002) 0.017*** (0.001) 0.029*** (0.001)
(7) Δ Residual inequality 0.102*** (0.001) 0.025*** (0.001) 0.077*** (0.001) ... ... . . . . . ..
(8) Δ Residual inequality, x’s 1985 0.028*** (0.001) −0.012*** (0.001) 0.040*** (0.001) . . . . . .. . . . . . ..
(9) Δ Residual inequality, x’s 2007 −0.028*** (0.002) −0.038*** (0.002) 0.010*** (0.001) . . . . . .. . . . . . ..
Source: idem Table 1. Notes: bootstrapped standard errors (100 reps.) are in parentheses. (***), (**), (*) significant at 1%, 5% and 10% significance level,
respectively.

Table 5. Wage inequality change in Portugal 2007–2017, hourly earnings (log).


90–10 90–50 50–10 Gini Theil
(1)Δ overall inequality −0.199*** (0.001) −0.068*** (0.001) −0.131*** (0.001) −0.036*** (0.001) −0.031*** (0.003)
(2) Δ overall inequality, x’s 2007 −0.390*** (0.001) −0.198*** (0.001) −0.193*** (0.001) −0.072*** (0.001) −0.053*** (0.004)
(3) Δ overall inequality, x’s 2017 −0.364*** (0.001) −0.163*** (0.002) −0.201*** (0.001) −0.067*** (0.001) −0.063*** (0.003)
(4) Δ Predicted inequality −0.105*** (0.002) −0.067*** (0.002) −0.038*** (0.001) −0.033*** (0.000) −0.034*** (0.000)
(5) Δ Predicted inequality, x’s 2007 −0.275*** (0.002) −0.188*** (0.001) −0.086*** (0.001) −0.066*** (0.000) −0.053*** (0.000)
(6) Δ Predicted inequality, x’s 2017 −0.269*** (0.002) −0.161*** (0.002) −0.107*** (0.001) −0.062*** (0.000) −0.053*** (0.000)
(7) Δ Residual inequality −0.102*** (0.001) −0.044*** (0.001) −0.058*** (0.001) ... ... . . . . . ..
(8) Δ Residual inequality, x’s 2007 −0.183*** (0.001) −0.085*** (0.001) −0.098*** (0.001) . . . . . .. . . . . . ..
(9) Δ Residual inequality, x’s 2017 −0.185*** (0.001) −0.083*** (0.001) −0.101*** (0.001) . . . . . .. . . . . . ..
Source: idem Table 1. Notes: bootstrapped standard errors (100 reps.) are in parentheses. (***), (**), (*) significant at 1%, 5% and 10% significance level,
respectively

is also the end of a period of growth and the start of in general, decreased over the period and this move-
a period of macroeconomic adjustment and labour ment speeded up after 2007. Hence, the increase in
market liberalizing reforms in Portugal (Távora wage inequality in Portugal occurred fundamentally
and González 2016). That analysis is presented in due to the modernization of the labour market and
Tables 4 and 5 and was carried out in a similar way not to increasing wage differentials either between or
as that displayed in Table 3. within groups of workers.
It is evident from the analysis of both tables that
the reduction of the price of observed and unob-
served characteristics had started in the first period; Data availability statement
it is also clear that this movement speeded up after The data used in this research were obtained from INE
2007, with the economic crisis and with the labour (Institituto Nacional de Estatística) under an agreement
market reforms. Finally, the positive composition between the Ministry of Science and Technology of Portugal
effects are present along the entire period, so they and INE. Further information on how to access these data can
are a long-term trend and not an episodic event. be found at: http://www.dgeec.mec.pt/np4/46/

Disclosure statement
IV. Concluding remarks
I declare that there are no conflicts of interest involved in this
This article shows that the increased wage inequality research.
in Portugal since the middle of the 1980s is funda-
mentally explained by the changes in the structure of
employment, such as the increased educational level Funding
and the average age of the employed and the This paper is financed by National Funds of the FCT –
increased share of women in employment. The Portuguese Foundation for Science and Technology
returns to observed and unobserved characteristics, [«UIDB/04007/2020»].
APPLIED ECONOMICS LETTERS 5

References Lemieux, T. 2006. “Increasing Residual Wage Inequality:


Composition Effects, Noisy Data, or Rising Demand for
Campos, M., and H. Reis. 2019. “Ainda Compensa Investir Skill?” American Economic Review 96 (3): 461–498.
Em Educação?” https://www.bportugal.pt/sites/default/ doi:10.1257/aer.96.3.461.
files/anexos/cep_2_1_pt.pdf Lemieux, T. 2008. “The Changing Nature of Wage
Card, D., A. Cardoso, J. Heining, and P. Kline. 2018. “Firms Inequality.” Journal of Population Economics 21 (1):
and Labor Market Inequality: Evidence and Some Theory.” 21–48. doi:10.1007/s00148-007-0169-0.
Journal of Labor Economics 36 (S1): S13–S70. doi:10.1086/ Machado, J., and J. Mata. 2005. “Counterfactual
694153. Decomposition of Changes in Wage Distributions Using
Cardoso, A. 1998. “Earnings Inequality in Portugal: High and Quantile Regression.” Journal of Applied Econometrics 20
Rising?” Review of Income and Wealth 44 (3): 325–343. (4): 445–465. doi:10.1002/jae.788.
doi:10.1111/j.1475-4991.1998.tb00285.x. Olivetti, C., and B. Petrongolo. 2016. “The Evolution of
Centeno, M., and A. Novo. 2014. “When Supply Meets Demand: Gender Gaps in Industrialized Countries.” Annual Review
Wage Inequality in Portugal.” IZA Journal of European Labor of Economics 8 (1): 405–434. doi:10.1146/annurev-
Studies 3 (1): 1–20. doi:10.1186/2193-9012-3-23. economics-080614-115329.
DiNardo, J., N. Fortin, and T. Lemieux. 1996. “Labor Market Távora, I., and P. González. 2016. “Market Regulation and
Institutions and the Distribution of Wages, 1973-1992: Collective Bargaining in Portugal during the Crisis:
A Semi Parametric Approach.” Econometrica 64 (5): Continuity and Change.” European Journal of Industrial
1001–1044. doi:10.2307/2171954. Relations 22 (3): 251–265. doi:10.1177/0267323116643210.

You might also like