Intergenerational Earnings Persistence in Italy between Actual Father–Son Pairs Accounting for Lifecycle and Attenuation Bias
| Author | Michele Raitano,Francesco Bloise |
| Published date | 01 February 2021 |
| Date | 01 February 2021 |
| DOI | http://doi.org/10.1111/obes.12375 |
88
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 83, 1 (2021) 0305–9049
doi: 10.1111/obes.12375
Intergenerational Earnings Persistence in Italy
betweenActual Father–Son PairsAccounting for
Lifecycle and Attenuation Bias
Francesco Bloise† and Michele Raitano‡
†University of Roma Tre, Roma, Lazio 00154, Italy. (e-mail: francesco.bloise@uniroma3.it)
‡Department of Economics and Law, Sapienza University of Rome, Roma 00161, Italy.
(e-mail: michele.raitano@uniroma1.it)
Abstract
Using a longitudinal dataset built merging administrative and survey data, we contribute
to the literature on intergenerational inequality providing the first estimate of the inter-
generational earnings elasticity (IGE) in Italy based on actual father–son pairs, taking
into account issues related to measurement biases and comparing the size of the lifecycle
bias when sons are selected by age or by potential experience (i.e. the number of years
since the end of their studies). Our findings confirm that Italy is a low-mobility country.
In our baseline estimate, when sons are observed 6 years after the end of their studies, the
IGE is approximately 0.41 and is robust to various measures of fathers’lifetime ear nings.
However, our results might be downward biased by the young age of sons.To measure the
lifecycle bias and correct IGE estimates, we run the ‘forward regression’of yearly earnings
on lifetime earnings on a sample of workers followed for 30 years. We find that selecting
sons by potential experience rather than by age reduces the lifecycle bias at youngages and
the ‘corrected’ IGE is 0.48.The picture of Italy as a low-mobility country is also confirmed
when we measure the intergenerational association through the rank–rank slope.
I. Introduction
Cross-country comparisons of intergenerational inequality – usually measured through
intergenerational earnings elasticity (henceforth, IGE), which is estimated regressing log
children’s earnings (when adult) on log parental earnings (Bj¨orklund and J¨antti, 2009)
– agree on the rankings among developed countries (e.g. Blanden, 2013; Corak, 2013):
Nordic European countries are the most mobile, while the US, the UK and Italy are among
the most unequal countries, with IGEs above 0.40.
To correctly estimate the IGE, panel datasets covering subsequent generations are
needed. However, when proper longitudinal datasets following the two generations for
many years are not available, IGE estimates are downward biased because of attenuation
JEL Classification numbers: J62, D31, D63.
Intergenerational persistence in Italy 89
and lifecycle biases deriving from the fact that point-in-time earnings are far from a good
proxy of lifetime earnings (Haider and Solon, 2006; Mazumder, 2005).
Furthermore, where, as in most countries, longitudinal datasets observing parents in
middle age and their children when adult are not available, it is not possible to directly
link parents’ and children’s ear nings, and the intergenerational association can only be
estimated applying the two-sample two-stage least squares (TSTSLS) method, that is, by
imputing parents’ear nings exploitingrepeated cross-sectional datasets where retrospective
information on parents’ characteristics is available.1The literature has noted that the TST-
SLS method might produce biased coefficients not perfectly comparable to those obtained
from the baseline OLS estimator, thus limiting the cross-country comparability of IGEs
estimated through different methods (Blanden, 2013). The direction of the bias is unde-
termined (Olivetti and Paserman, 2015) even if, according to most authors, the TSTSLS
estimate of the IGE is likely upward biased (Bj¨orklund and J¨antti, 1997).
Regarding Italy, the focus of this article, previous estimates of the IGE – whichdraw the
picture of an immobile country with an IGE between 0.45 and 0.50 (Mocetti, 2007; Piraino,
2007; Barbieri, Bloise and Raitano, 2019) – were based on the TSTSLS method due to
the lack of longitudinal data tracking subsequent generations for at least a portion of their
working careers. Moreover, these studies observed the earnings of the two generations in at
most a few years.Therefore, because of the possible biases that might have affected the pre-
vious estimates, a crucial question concerns the reliability and cross-country comparability
of IGE estimates for Italy.
In this article, exploiting a recently developedlongitudinal dataset, we are able to answer
this question and provide the first reliable estimate of the IGE in Italysince we can consider
actual father–son pairs instead of relying on the TSTSLS method and can carefully take
into account both the attenuation and the lifecycle biases.
This dataset was recently built merging the 2004–08 waves of the Italian component
of the European Union Statistics on Income and Living Conditions (EU-SILC; the Italian
component is called the IT-SILC) with longitudinal social security records managed by the
Italian Social Security Institute (INPS). For all the individuals interviewed in the IT-SILC,
the dataset enriches information available in the survey’s waves with administrative records
about employment and earnings histories from the year the individuals entered the labour
market until 2014. The characteristics of the dataset allow us, on the one hand, to couple
the son and his co-residing father by using IT-SILC variables that link individuals living
in the same household at the moment of the interview and, on the other hand, to track over
time sons’ and fathers’ gross annual earnings (from employment and self-employment)
through the longitudinal information collected in administrative archives.
Specifically, using IT-SILC variables regarding the current educational status and the
year when the highest education was attained, we select a main sample of 718 sons who
achieved the highest degree and ended their studies at most 2 years before the calendar
year of the interview (e.g. in 2002–04 for those belonging to the IT-SILC 2004 wave, in
2006–08 for those belonging to the 2008 wave),that is, we select sons who had just become
1The TSTSLS method was introduced in the empirical literature on intergenerational mobility by Bj¨orklund and
J¨antti (1997) and has then been used in many countries wherepanel datasets covering subsequent generations are not
available (see the reviewby Jerrim, Choi and Simancas, 2016).
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial