PreprintPDF Available

Technical change and wage premiums amongst skilled labor: evidence from the economic transition

Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

I present reduced-form and structural evidence that the reorganization of the Russian economy in the post-transitional period increased the demand on law and business graduates. This demand shock provides a novel unified explanation of the Russian wage structure for 1985–2015. I then show that this shock is a common feature of all transitional economies, and it contributed to the transformational recession. The demand behaviour is identified with a new skill-biased technical change model of demand for skills with three production inputs (high school graduates and bachelor-level educations with two majors), showing that a technology shift that favours a particular skill might emerge within the skilled group rather than between skilled and unskilled. This is relevant because similar shifts (e.g., data scientists vs. liberal arts) emerge today in the frontier economies that adopt new general-purpose technologies (e.g., machine learning). Thus, this paper informs policymakers today on tools to counteract a potential drop in economic equality and performance that result from this adoption. Lastly, because of similarities between the mechanics of the transition and the 2022 sanctions to discourage Russia's war effort, my results highlight the importance of additional sanctions against the education system to prevent the regime's structural adaptation and preservation.
Content may be subject to copyright.
Technical change and wage premiums amongst
skilled labor: evidence from the economic
This document is a post-print; please cite the published version:
Alexeev, S. “Technical change and wage premiums amongst skilled labor:
evidence from the economic transition.” Economics of Transition and
Institutional Change (2022)
Sergey Alexeev
The University of New South Wales (UNSW)
22-32 King St, Randwick NSW 2031, Australia
April 26, 2022
I present reduced-form and structural evidence that the reorganization
of the Russian economy in the post-transitional period increased demand
on law and business graduates. This demand shock provides a novel uni-
fied explanation of the Russian wage structure for 1985–2015. I then show
that this shock is a common feature of all transitional economies, and it
contributed to the transformational recession. The demand behavior is
identified with a new skill-biased technical change model of demand for
skills with three production inputs (high school graduates and bachelor-
level educations with two majors), showing that a technology shift that
favors a particular skill might emerge within the skilled group rather than
between skilled and unskilled. This is relevant because similar shifts (e.g.,
data scientists vs. liberal arts) emerge today in the frontier economies that
adopt new general-purpose technologies (e.g., machine learning). Thus,
this paper informs policymakers today on tools to counteract a potential
drop in economic equality and performance that result from this adoption.
Lastly, because of similarities between the mechanics of the transition and
the 2022 sanctions to discourage Russia's war effort, my results highlight
the importance of additional sanctions against the education system to
prevent the regime's structural adaptation and preservation.
Keywords: Wage Level and Structure; Human Capital; Skill-Biased Technical
Change; Economic Transition.
JEL codes: J31; J24; O33; P30.
1 Introduction
It is widely accepted that the IT revolution drove changes in wage structures
and earnings inequality in the United States and other OECD countries in the
1970s. The empirical observation that the deployment of new technology is
accompanied by the creation of better-paid jobs that require higher qualifica-
tions is known as skill-biased technical change (SBTC) (Acemoglu and Autor
2011; Katz and Autor 1999; Violante 2016). To generalize SBTC beyond IT,
economists use the term general-purpose technology to describe technological
advances that pervade many sectors, improve rapidly, spawn further innovations
and induce wage inequalities (Rousseau 2016). Some examples include electric-
ity (Jovanovic and Rousseau 2005), steam engines (Chin, Juhn, and Thompson
2006; Pehkonen, Neuvonen, and Ojala 2019) and organizational change (Bres-
nahan, Brynjolfsson, and Hitt 2002; Caroli and Van Reenen 2001; Dessein and
Santos 2006; Garicano and Rossi-Hansberg 2006; Milgrom and Roberts 1990).
Although SBTC processes have become the leading explanation for changes
in the wage structures of developed countries, in transitional economies, little
consideration regarding the effects of technical changes on the wage structure
has been given to date. This paper shows that the SBTC framework provides
a powerful explanation of the Russian wage structure for the period 1985 to
2015 and consolidates the somewhat disorganized literature on wages in Russia.
The findings also support recent observations that the simplest formulations
of SBTC, in which technological advances raise the relative demand for skilled
workers in every task, overlooks some richer implications of technical change for
the demand for skills (Acemoglu 1999; Chin, Juhn, and Thompson 2006; Ojala,
Pehkonen, and Eloranta 2016; Pehkonen, Neuvonen, and Ojala 2019).
During the Soviet period (the leftmost part of Figure 1), both the college
wage premium (Gregory and Kohlhase 1988) and general monetary inequality
(Novokmet, Piketty, and Zucman 2018) were low. Inequalities existed in the
form of access to better shops, products, or vacation facilities, and jobs that
1In Poland, the Czech Republic and Slovakia, returns to education increased while the
experience premium fell (Chase 1998; Rutkowski 1996). In contrast, returns to both education
and experience fell during the transition in Slovenia (Stanovnik and Verbiˇc 2005) and in
Romania (Andr´en, Earle, and S˘ap˘atoru 2005). Other influential comparative studies for the
period of transition are as follows. Fleisher, Sabirianova Peter, and Wang (2005) concluded
that returns to schooling increased proportionally with the speed of economic reforms and
favored the young. Brainerd (2000) found that the wage penalty for women increased in
Russia and Ukraine but decreased in the rest of Eastern Europe. Krueger and Pischke (1995)
showed that returns to both education and experience fell slightly in eastern Germany after
unification. Sabirianova Peter (2003) used linked employer-employee data and showed that
in Russia, the transition to a market economy was skills-biased because market liberalization
adjusted wages to the true marginal productivity. Gorodnichenko and Peter (2005) compared
wage premiums for schooling between Russia and its closest institutional analog, Ukraine, for
the period 1985–2002 and noted that, once both economies were decentralized, the Russian
wage premium went up quicker than that of Ukraine's and attained a significantly higher value.
Figure 1: Bachelor degree returns and economic output in Russia: 1985–2015
GDP, thousand constant roubles
Returns on bachelor degree
Returns on higher education GDP Regression line
Returns on bachelor degree
Notes: Returns are estimated with Specification (1). Dashed vertical lines separate the peri-
ods. Vertical lines indicate robust standard errors clustered at the individual level. Regression
line is fitted for 1998 to 2015. GDP is in per capita terms calculated in constant prices and
local currency.
Source: RLMS 1994-2015 (1985 and 1990 collected retrospectively in 2000 and 2001); World
Bank for GDP.
required a college degree often provided that access. Unlike other Central and
Eastern Europe countries (CEECs),1many studies show that returns to educa-
tion during the transition period in Russia remained relatively low (the middle
part of Figure 1).2This is often referred to as the market adjustment puz-
zle (Gimpelson and Kapeliushnikov 2011) and is attributed to the abundance
of well-educated workers in an economy in which blue-collar employees were
in high demand (Brainerd 1998). The Soviet system encouraged education in
an unsustainable manner, for example, stipends and tuition were free. Par-
ents could not let their children inherit capital assets but could support them
through education. Education also allowed citizens to escape particularly phys-
ically harmful Soviet blue-collar jobs and provided connections, which ensured
that necessary goods and services could be accessed in case of shortages (Katz
1999). The resulting oversupply of skills (sometimes referred to as the educated
Russian's curse (Cheidvasser and Ben´ıtez-Silva 2007)), was presumed to push
2Other Russian college wage premium studies for the period of transition are as follows.
Newell and Reilly (1996) estimate a wage function at the very beginning of the reforms and
find low (about 4%) returns for schooling. Newell and Reilly (1999) report returns on levels
of education up to 1996 and show an initial increase in returns and a subsequent decline.
Clark (2003) utilizes data from 1994-1998 and notes a significant (6-13%) return on a year of
schooling comparable in magnitude to those in other transition economies.
the returns down in the first post-transitional years. Ultimately, returns peaked
in 1998 (the rightmost part of Figure 1), when the economy started recovery,
and have been gradually decreasing ever since. The decrease is attributed to
the larger involvement of applicants with lower levels of ability during the ex-
pansion of the college education system (Belskaya and Sabirianova Peter 2014;
Belskaya, Sabirianova Peter, and Posso 2020; Kyui 2016). General income in-
equality followed a pattern similar to the college premium. It peaked in 1998
and then gradually decreased (Gorodnichenko, Peter, and Stolyarov 2010).
The explanation of the wage structure during the economic transition in
the current paper hinges on the most characteristic feature of the transition
period – the transformational recession.3A common theme of the theories
of the recession is within (Atkeson and Kehoe 1995,2005) and between firms'
(Blanchard and Kremer 1997; Roland and Verdier 1999) reorganization following
the dismissal of the central planner. In a market economy, a pool of consumers
defines an aggregate consumption profile, and, through the price system, firms
follow demand. The environment requires firms to engage with consumers and
each other; however, the contractual and informational imperfections that arise
from these interactions require a judicial system and a body of contracts. This
accounts for the amount of law and business (LB) graduates in the market
economy (Arrow and Debreu 1954). By analogy, centralized economies of former
Soviet Union (FSU) republics and CEECs did not use price signals to broadcast
values, and politicians defined an aggregated consumption profile. Demand
and supply were connected by the central planner, while the execution of the
plan was carried out by several hundreds of ministries that, through directives,
harmonize enterprises into industries and industries into an economy. Firms
had no need to engage with each other or consumers; thus, there was no need
for personnel trained in LB (Kornai 1979).4
If indeed the economy collectively invested in a reorganization following the
transition from a centralized to a market economy, then the detectable creation
of better-paid jobs should accompany this effort. To this end, Section 2of this
paper reestimates the return on bachelor's degree from 1985 to 2015 and ac-
knowledges that the skills of LB graduates are complementary to organizational
technologies. Using the Russia Longitudinal Monitoring Survey (RLMS), I show
that, during the transition period, the skills provided by LB graduates experi-
enced a substantial increase in returns. In Section 3, I study the role of demand
3The transformational recession is an unexpected and universal decline in output during
the transition from a centralized to a market economy. The decline for 26 transition economies
from 1989 to 1995 was 41% (Ickes 2016). By comparison, output in the United States during
the Great Depression declined by 34%.
4For a more detailed coverage of the nature the Soviet centralized economy and the pere-
stroika period I referee the interested readers to Ericson (1983,1991,2006,2017) and Gross-
man (1962,1963,1966).
and supply in these returns by constructing and estimating a novel structural
model that allows for imperfect substitution between LB and the rest of the
graduates (refereed to as NLB from now on). Taken together, the reduce-form
and structural estimates suggest that indeed the economy went through a reor-
ganization that created an unusually large labor market demand-side shock on
the skills of LB graduates after economic decentralization.
These findings provide a unified explanation for the relatively low college
returns from 1994 to 1996 and the decrease in returns from 1998 to 2015. Low
returns during the transition period are not the result of an oversupply of ed-
ucated workers; rather, it is the undersupply of LB graduates that spike the
returns on specialization and the oversupply of NLB graduates, whose returns
are the same as those of the Soviet Union. The decrease in returns in present-day
Russia is not (at least exclusively) the result of a larger enrollment of applicants
with lower abilities during the expansion of the education system; rather, it is
the effect of the transitory wage differential, which initially elevates the average
returns and then causes them to plunge.
In Section 4, I use the model to show a theoretical connection between the
size of this shock and the transformational recession. To support this connection,
I then show that the extent of transformational recession and the fraction of
students enrolled in LB specializations are positively correlated, suggesting that
the mechanics described above generalizes to all other transitional economies.
In conclusion, I note that the findings of this paper are particularly important
for the developed economies today. The frontier economies are going through a
new adaptation stage of general-purpose technologies (e.g., artificial intelligence
and machine learning). The defining feature of this stage is that a differential
emerges within the skilled labor (e.g., data scientists vs. liberal arts). This is
different from what has been discussed to date for the developed nations, where
the differentials between the high school (i.e., unskilled) and college graduates
(i.e., skilled) have been studied, and much similar to what I show happened to
the Russian economy during the transition.
I now demonstrate the reduced-from evidence.
2 Reduced-form evidence
2.1 Methods
In this section, I present an overview of trends in the college wage premium for
the period 1985-2015 with a human capital equation:
Yit =
βt(dit ×λt) +
it ×λt) + λt+εit.(1)
The variable dit is an indicator for a bachelor degree, Yit is log of wage, iindexes
individuals and tindexes years. The equation also has an array of controls, Xh
It includes standard human capital measures, such as experience and indicators
for other levels of education and gender, and also includes occupational and
residential characteristics. Controls for residency and occupation are standard
for the dataset and captures the country's pronounced agglomeration effects
(e.g., Combes and Gobillon 2015). The parameter λtis year fixed effects. These
effects partial out the effect of inflation and tax reforms on wages. The variance
of the control variables is partialled out with time-variant parameters, allowing
for a change in the variables' composition from year to year.
The vector of year-dependent parameters βtis of interest. Because dit is a
dummy variable, and secondary school is omitted, βtis the controlled percent
difference in mean outcomes between respondents with bachelor degree and
those without.
Equation (1) implies that the percent mean difference in skilled and unskilled
labor is homogenous within a year. To relax this assumption, I then estimate
the following equation:
Yit =
t(dit ×mq
it ×λt) +
it ×λt) + λt+εit.(2)
The Equation (2) is identical to Equation (1), except dit is now interacted
with an indicator for LB major, m
it, and an indicator for NLB major, m
Secondary school is still an omitted category; therefore, the vector of year and
major dependent βq
tcoefficients shows a controlled percent difference in mean
wage between secondary school students and bachelor graduates with LB or
NLB majors.
This equation demonstrates the main finding of this paper: unusually high
but transitory returns on the LB majors during and post-transition. This pat-
tern of returns is pronounced at the bachelor level of education, which experi-
enced a remarkable expansion following the reforms of the early 1990s (Belskaya,
Sabirianova Peter, and Posso 2020; Kyui 2016).
To confirm the robustness of the results demonstrated with Specification (2),
a quantile regression is then employed. The quantile regression estimators are
notoriously unstable at the tails of conditional distribution (e.g., Chernozhukov
2005; Koenker et al. 2018). To increase the efficiency of the quantile parame-
ters, I estimate them on two adjacent waves pooled together. These waves are
1995/1996, 1998/2000, and 2014/2015. The specified quantile regression model
QkYitτ |Xitτ =
βr,k(ditτ ×pr
itτ ) +
itτ +λk
Here, pr
itτ is a series of indicators that replace the indicator for NLB graduates,
it, with an array of indicators for more refined groups of bachelor degrees
specializations. Ultimately the equation shows the returns for 5 majors (as
opposite to 2 in Equation (2)), which are denoted by rin summation. These
majors are LB, STEM, medical, liberal arts, and public sector. This is done
to confirm that the only returns for LB graduates behave in a unique manner.
Symbol τdesignates a two-wave period, kdesignates a quantile. In relation to
the control vector, Xh
itτ , Equation (3) is similar to Equation (2) or Equation (1).
For all three models, to account for the error term's autocorrelation over
different years for the same individual, I cluster standard errors by individuals
and use a robust covariance matrix (Abadie et al. 2017).
Finally, the purpose of the βparameters is not to identify the causal effect of
the major choices on wage, as in, for example, Kirkeboen, Leuven, and Mogstad
(2016). I follow a voluminous literature studying evolution in the returns to
skills (e.g., Acemoglu and Autor 2011; Katz and Autor 1999). This approach
predates Angrist and Pischke (2010), who emphasized research designs at the
expense of economic theorizing.
An economically meaningful interpretation of parameters of interest is a price
of a particular observable that prevails in the labor market. The parameters
capture the interactions of the demand and the supply sides. Specifically, in the
current context, the estimates capture the competition among firms to obtain
the scarce supply of the skills possessed by LB graduates. Section 3builds up
the structural model to separate the roles of supply and demand.
I now introduce the data.
2.2 Data
This paper uses waves 1994-2015 of the individual questionnaires of RLMS.
RLMS is a high-quality national survey used in dozens of publications each
year across all social sciences. The survey is conducted by the University of
North Carolina at Chapel Hill, National Research University Higher School of
Economics and the Federal Center of Theoretical and Applied Sociology of the
Russian Academy of Sciences. For a history of the survey, an outline of the
sample design and the replenishment of sample designs, the loss to follow-up,
and other key factors see the data resource profile in Kozyreva, Kosolapov, and
Popkin (2016). In total, the dataset spans 109,607 observations or an average of
5,768 observations per wave. The survey contains detailed and readily available
information on employment and education.
The data for 1985 and 1990, the Soviet period, is collected retrospectively
in 2000 and 2001, which is a potential threat to validity due to recall bias.
Luckily, Sabirianova Peter (2003) provides an assessment of the recall bias in
RLMS by comparing answers on occupation with the official enterprise reports.
She concludes that the recall is not significant due to stable salaries and the
strong attachment of workers to one job in the Soviet period.
Importantly, RLMS remains the best data source for the late Soviet period
wages. Prior to RLMS, the only surveys available were of former Soviet citizens
residing in Israel and the United States (the Israeli Interview Project and the
Soviet Interview Project), which were used in high-quality studies (e.g., Gregory
and Collier 1988; Gregory, Mokhtari, and Schrettl 1999). Both are probability
sample surveys restricted to urban families from the European parts of the
former Soviet Union. RLMS is a considerable improvement over those surveys.
Still, I exercise due caution for the estimates for the Soviet period. In particular,
in Section 3the observation for 1985 and 1990 are disregarded while estimating
the structural parameters.
Table 1provides the descriptive statistics for the variables considered in this
study. The upper part of the table reports human capital measures, parts below
report residential and then job characteristics.
Table 1: Descriptive statistics
Period 1985-1990 1994-1996 1998-2008 2009-2015 Overall
Mean St. Dev Mean St. Dev Mean St. Dev Mean St. Dev Mean St. Dev
Log of wage 5.211 0.510 7.119 1.190 8.119 1.116 9.550 0.773 9.296 1.531
Bachelor degree 0.210 0.410 0.222 0.416 0.229 0.420 0.291 0.454 0.263 0.440
Experience 20.210 10.829 20.443 10.829 20.044 10.974 20.489 11.250 20.325 11.119
Female 0.534 0.499 0.531 0.499 0.534 0.499 0.529 0.499 0.531 0.499
Federal City · · 0.183 0.387 0.177 0.382 0.162 0.369 0.169 0.375
Regional Center · · 0.476 0.499 0.460 0.498 0.426 0.494 0.442 0.497
Town · · 0.303 0.460 0.289 0.453 0.290 0.454 0.291 0.454
Urban Village · · 0.057 0.231 0.053 0.225 0.062 0.241 0.058 0.234
Rural Area · · 0.164 0.370 0.198 0.399 0.223 0.416 0.209 0.407
Foreign company · · 0.036 0.187 0.040 0.195 0.032 0.175 0.035 0.183
National company · · 0.251 0.433 0.412 0.492 0.494 0.500 0.445 0.497
State company · · 0.691 0.462 0.545 0.498 0.414 0.493 0.484 0.500
Ownership missing · · 0.144 0.351 0.130 0.336 0.137 0.343 0.135 0.341
Size 0-10 · · 0.067 0.249 0.075 0.263 0.092 0.289 0.083 0.277
Size 10-50 · · 0.198 0.399 0.194 0.395 0.220 0.414 0.209 0.406
Size 50-100 · · 0.093 0.290 0.091 0.287 0.092 0.290 0.092 0.289
Size 100-500 · · 0.186 0.389 0.167 0.373 0.143 0.350 0.155 0.362
Size 500-1000 · · 0.051 0.220 0.049 0.216 0.039 0.193 0.043 0.204
Size 1000 · · 0.121 0.326 0.130 0.336 0.063 0.243 0.092 0.289
Size missing · · 0.285 0.451 0.295 0.456 0.351 0.477 0.325 0.469
Sample size 31,308 9,193 39,308 61,106 109,607
Source: RLMS 1994-2015 (1985 and 1990 collected retrospectively in 2000 and 2001).
Residential and job information is not available for 1985 and 1990. Age and
educational levels are also missing but can be reconstructed using the panel
nature of the data. RLMS report the year when each educational level was
conferred, which allows reconstructing the educational levels. The age is recon-
structed using birth year. For all years, experience is calculated as age minus
the number of years of education and the number 6. All other variables readily
available in RLMS
Two features are evident from the table. First, the Soviet wages are relativity
the lowest. This impression is the result of hyperinflation during the transition
period. Second, the largest firms are gradually breaking down into smaller firms,
showing that the economy is reorganizing. For example, the share of workers
employed in the firms with more than 1000 workers is 12% in 1994-1996 and
drop to 6% in 2009-2015.
The preferred dependent variable, characterized in Table 1, is a log of monthly
contractual wages at the primary workplace. This choice of the dependent vari-
able is standard for the dataset and is considered the best choice to proxy wages
(e.g., Alexeev 2022; Brainerd 1998; Carnoy et al. 2012; Cheidvasser and Ben´ıtez-
Silva 2007; Gorodnichenko and Peter 2005; Kyui 2016; M¨unich, Svejnar, and
Terrell 2005). This is due to wage delays during the transition period and low
quality or missing information on hours worked. In particular, hours worked
are completely unavailable for the Soviet period.
Finally, to reduce the biases caused by sample attrition due to the higher
mortality rates of older age groups or labor mobility, I follow the previous au-
thors and restrict the estimation sample to the respondents aged 15–60 years
and use the sample weights.
While the preferred dependent variable is customary and convenient, it poses
two immediate threats to validity that originate from across specializations vari-
ation in hours worked and non-pecuniary payments. The first threat is that
omitting the hours worked variable, which is negatively correlated with edu-
cation, may produce a downward bias. The second threat is that it may be
that respondents who are employed in newly created positions or who work in
a new industry are compensated entirely by wages, while more traditional spe-
cializations are partially compensated by other types of non-pecuniary rewards
typical for the Soviet enterprise (e.g., paid vacation at health resorts). Happily,
the main discovery of this paper – the transitory wage differential for the LB
specialists – is robust to two changes in the definition of the dependent variable.
The data on hours worked is of the highest quality for the waves 1998 on-
wards. Using a log of wage rate in those waves (and excluding the top 1% of
declared hours worked to account for values outside of the reasonable range)
changes the magnitude of the estimates marginally but uniformly for all ma-
jors. Thus, the expected differential for LB majors is still present. Another
sensitivity exercise performed is borrowed from Gregory and Kohlhase (1988),
who included hours worked as a covariate. Applying this approach to all waves
further confirms that the differential is still there.
I also verified that the differential is robust to payments in-kind. In partic-
ular, in cases in which the RLMS respondents state that, in addition to their
wages, their firm makes payments in-kind, and those values are known to the
respondents, I added those values to their wages. This approach is somewhat
similar to that of Gregory and Kohlhase (1988), which includes indicators to
address differences related to nonmonetary job privileges. The expected dif-
ferential is also present. The differential is also robust to the inclusion of the
specialization fixed effects. These would correct for potentially contaminating
relevant unobserved heterogeneities in regards to specializations.
2.3 Data on fields of specialization
Table 2: Specialization definitions
Current discussion ISCO 88 Example
STEM 2111-49; 2213; 3111-43; 7241-2. Computer programmer.
Medical 2211-2; 2219-29; 3221-42; 3475. Medical doctor, dentist.
Law or business 2411; 2419-29; 2441; 3411-39. Financial consultant, insurer.
Liberal arts 2412; 2431-2; 2442-60. Philosopher, sculptor.
Public sector 110; 2230-52; 3151-2; 3441-60. Fire inspector, police detective.
Notes: International Standard Classification of Occupations (ISCO) is an International Labour
Organization definition of occupations. An example of the coding used in the middle column is
3113 53113 3115.
I classify bachelor's degree specializations into STEM, medical, liberal arts,
and public sector (professions for which the state is the only employer). They
are collectively referred to as NLB. LB graduates are excluded from all groups
and treated separately. Because the empirical exercise of this paper is sensitive
to the definitions of specializations, Table 2provides the most precise defini-
tion possible. The table maps the defined specializations into the International
Standard Classification of Occupations and provides examples.
Respondents with missing specialization information are excluded from the
sample because they are missing at random, as the rest of this subsection demon-
strates. An alternative to excluding the data point with missing information is
to have a separate dummy for missing information. Doing this changes practi-
cally nothing to the estimates. A slight issue arises, as the respondents are only
asked for their specialization from 1998 to 2001 and 2010 to 2015. Adopting
the approach similar to Belskaya, Sabirianova Peter, and Posso (2020), the in-
formation for the missing years is obtained through a panel component of the
data (the same respondents respond each year).
Figure 2depicts the portion of missing information on specializations among
respondents with bachelor's degrees after the missing data is obtained across the
waves. If information is missing, it is taken from the closest year with informa-
tion, prioritizing later years. This rule of giving priority to later years alleviates
Figure 2: Fraction of missing data on majors
1994 1995 1996 1998* 2000* 2001* 2002 2003 2004 2005 2006 2007 2008 2009 2010* 2011* 2012* 2013* 2014* 2015*
Share of missing information
Notes: Years with a star contain actual information on specializations of bachelor; information
for other years is taken from the closest year with information, with priority given to later
years. My testings shows that no problems arises if less than 31% of data is missing. Horizontal
red line marks 31%.
Source: RLMS 1994–2015.
a potential problem that some respondents might change specialization in re-
sponse to relatively high returns on some of them. The rule is also the reason
why the columns in Figure 2have a somewhat symmetric shape from 1998 to
Luckily, the information on specializations is fully available in the data for
the last six years. Thus, it is possible to pretend that information for those
last six years is missing and replicating an imputation procedure to assess its
accuracy. Figure 3depicts the returns (with the full set of controls as in Equa-
tion (1)) on defined groups of specializations using the actual data and the data
that is obtained from the years 2010 and 2011, pretending that information for
the years 2012 to 2014 is missing. The coefficients are similar. The estimate
for the unspecified major becomes more precise with the imputed information
but has no interpretation. It shows that the information on specializations is
missing at random. The last set of estimates show the results when the missing
information on bachelor's degree specialization is excluded. The excluding does
not result in estimates that are statistically different from when the information
is included.
The assessment shows that no significant problem arises even if the figure
reaches 31% (this is how much data is missing in 2014). Somewhat unfortu-
Figure 3: Comparing estimates with actual and imputed information
Law or business
Liberal artr
Public sector
Law or business
Liberal arts
Public sector
Law or business
Liberal arts
Public sector
Mincerian function coefficient
Actual Imputed Dropping
(share of unspecified 20%)
(share of unspecified 26%)
(share of unspecified 31%)
Notes: Last 5 years have information on specializations which permits imputation assessment.
Coefficients are estimated with Equation similar to Equation (2), but with more refined groups
of specializations, as in Equation (3). Black vertical lines are 95% confidence intervals created
using robust standard errors.
Source: RLMS 2010–2014.
Table 3: Assessment of specialization imputation in aggregate
Law or STEM Medical Liberal Public Unspecified
business arts sector
2012 (share of unspecified 20%)
Actual 31.4% 34.3% 5.9% 9.2% 17.9% 1.4%
Imputed 27.3% 26.5% 4.2% 8.3% 13.6% 20.2%
Dropping 33% 34% 5% 10% 18% 0%
2013 (share of unspecified 26%)
Actual 33.2% 33.0% 5.9% 9.5% 18.0% 0.4%
Imputed 26.2% 23.8% 4.2% 7.7% 12.4% 25.7%
Dropping 34% 32% 6% 10% 18% 0%
2014 (share of unspecified 31%)
Actual 34.1% 33.0% 5.4% 9.1% 17.8% 0.6%
Imputed 25.3% 21.9% 3.6% 7.3% 11.0% 30.9%
Dropping 36% 32% 5% 10% 17% 0%
Notes: Actual proportions are taken from the data; Imputed contains proportions
generated from information on specialization contained in 2010 and 2011, ignor-
ing actual information on specializations in years 2012, 2013 and 2014; Dropping
represents proportions after dropping category Unspecified after information was
taken from 2010 and 2011.
Source: RLMS 2010–2014.
Figure 4: Share of graduates with higher education in the labor force
1985 1990 1994 1995 1996 1998*2000*2001* 2002 2003 2004 2005 2006 2007 2008 2009 2010*2011* 2012*2013*2014* 2015*
STEM Medical Public sector Liberal arts Law or business Without bachelor degree
Notes: Years marked with * do not use specialization imputation. Workers with unspecified
specialization are excluded. Percentage of workers with higher education corresponds to the
mean of variable Bachelor degree in Table 1.
Source: RLMS 1994-2015 (1985 and 1990 collected retrospectively in 2000 and 2001).
nately, more than 31% of information on specializations is missing for the 1994
and 1995 waves. Fortunately, for the 1996 wave, only 27% is missing, which
permits probing of the transition period with confirmed certainty.
As another test of the validity of imputation, Table 3shows the sample
aggregate structure of bachelor graduates if the unspecified category is excluded.
For example, the share of LB in 2012 is 31.4%, when we impute information
from 2010 and 2011, the share is 27.3% (and there is an increase in unspecified),
but when we exclude missing information after the imputation, the share of
LB goes back to 33%. As the composition remains almost identical (data on
specializations is missing at random), the unspecified category can be excluded,
allowing tracking the composition of specializations over time, which is presented
in Figure 4.
Figure 4shows the structure of all employed workers. The labor force com-
position demonstrates an obvious tendency: in 1994, LB graduates occupy ap-
proximately 3% of the employed worker, whereas, in 2015, they occupy 12%.
This recomposition is unlikely to happen without a price signal in a market
The next subsection shows that the price has indeed governed the change.
2.4 Results
Figure 1, in Introduction, depicts the bachelor degree returns estimated with
Specification (1). As the literature review discusses, returns during the Soviet
period are low but increase during the transition period. At the end of the 1990s,
the premium for bachelor's degree doubles and then drops until it reaches about
the same level as that in the middle of the transition period.
Figure 5: Returns for LB and NLB: 1985–2015
≤ ≤
Notes: Returns are from Specification (2). Years marked with * do not use specialization
imputation. Dotted lines are locally weighted scatterplot smoothers. Robust standard errors
clustered at the respondent level are shown. Vertical grey dashed lines separate the periods.
Source: RLMS 1994-2015 (1985 and 1990 collected retrospectively in 2000 and 2001).
Figure 5shows the estimated returns for the LB and NLB specializations
from Specification (2). The results show that, during the transition period, the
returns on NLB specializations are not statistically different from returns in
the Soviet period (with the exception of 1994, which has the least amount of
reliable information on specializations; see Figure 3). Conversely, the returns
for LB are about 4–5 times higher. Further, in 1998, once economic growth has
returned, the returns on the skills of LB graduates are 6–7 times higher than
returns on a bachelor's degree in the Soviet period. In the following years, the
returns slow down, presumably in response to an increase in supply. Naturally,
the overall wage premium for a bachelor's degree, a weighted average across
all specializations, partially mimics this massive transitory differential. This
manifests as a decrease in the bachelor's degree wage premium. The point
where the returns on skills of LB graduates gravitates to in 2015 appears to
Figure 6: Returns for LB and other majors: 1985–2015
1985 1990 1994 1995 1996 1998*2000*2001* 2002 2003 2004 2005 2006 2007 2008 2009 2010*2011* 2012*2013*2014* 2015*
STEM (smoothed) Medical (smoothed) Law or business (estimates)
Liberal arts (smoothed) Public sector (smoothed) Law or business (smoothed)
Notes: Returns are from Specification (2). Years marked with * do not use specialization
imputation. Dotted lines are locally weighted scatterplot smoothers. Robust standard errors
clustered at the respondent level are shown. Vertical grey dashed lines separate the periods.
Source: RLMS 1994-2015 (1985 and 1990 collected retrospectively in 2000 and 2001).
be one supported by the new post-transitional economic realities. The same
level of returns is shown by NLB specializations starting in 1998. The economic
reforms of the early 1990s introduced the market price system. Companies took
advantage of this new system's opportunities and signaled that they require a
new mix of skills in the labor market.
Figure 6shows the estimates when the NLB category is expanded. Medical
professionals show the lowest pay in the Soviet period. This is similar to what
other Soviet period studies show. One of the highest ratios of applicants to
admissions was in schools of medicine, despite the meager pay of physicians
(Katz 1999). Choosing a medical professional or other professions that provided
basic state services (note that the Public section is as low as Medical) was
optimal in the environment where personal connections had a larger influence on
economic well-being than money. The unstable results for Liberal arts probably
reflect the scarcity of this group in the sample (see Figure 4).
Now the quantile results from Specification (3). Figure 7depicts the esti-
mation for 1995 and 1996 during the transformational recession and shows that
companies do indeed compete for LB graduates, which drives their wages up.
Figure 8depicts the estimations for 1998 and 2000. The shape of the line again
confirms that the market treated LB graduates differently from other gradu-
ates. The coefficient for the median income group for LB graduates is slightly
higher than the OLS estimate (horizontal red line is below the thick blue line in
Figure 7: Distribution of returns on bachelor degree: 1995 and 1996
Notes: Returns are from Specification (3). Black vertical lines indicate a 95% confidence
interval. Red dotted line is a corresponding OLS estimator for the LB returns. LB graduates
occupy 2% of the labor force. Standard errors are bootstrapped and clustered at the
individual level. Clustering is performed using the method of Parente and Silva (2016).
Source: RLMS 1995-1996.
Figure 8: Distribution of returns on bachelor degree: 1998 and 2000
5 10 20 30 40 50 60 70 80 90 95
Percentile (log of wage)
STEM Medical Law or business Liberal arts Public sector
Notes: Returns are from Specification (3). Black vertical lines indicate a 95% confidence
interval. Red dotted line is a corresponding OLS estimator for the LB returns. LB graduates
occupy 3% of the labor force. Standard errors are bootstrapped and clustered at the
individual level. Clustering is performed using the method of Parente and Silva (2016).
Source: RLMS 1998-2000.
Figure 9: Distribution of returns on higher education: 2014 and 2015
5 10 20 30 40 50 60 70 80 90 95
Percentile (log of wage)
STEM Medical Law or business Liberal arts P ublic sector
Notes: Returns are from Specification (3). Black vertical lines indicate a 95% confidence
interval. Red dotted line is a corresponding OLS estimator for the LB returns. LB graduates
occupy 12% of the labor force. Standard errors are bootstrapped and clustered at the
individual level. Clustering is performed using the method of Parente and Silva (2016).
Source: RLMS 2014-2015.
Figure 8), suggesting that the outliers push the premium down rather than up
(this is important as one might think that LB returns are driven by higher-level
occupations such as managers that are likely to be LB graduates). Conversely,
as Figure 9shows, the labor market for 2014 and 2015 rewards LB graduates no
differently than it does any other specializations (although LB is still the best
paid major).
It seems that wage returns have over time become less progressive (note how
the downward slope of returns on Figure 7at the far left-hand side disappears),
which is what one would expect if income is related to skill and skill enhances
the returns to education. This makes sense given what is happening over time:
a transition from an older system in which wage and skill did not move together
strongly (communism) to one in which they arguably do (capitalism).5
5The estimates of the other parameters of the wage equation are generally in line with
other studies that use comparable modeling approaches (e.g., Gorodnichenko and Peter 2005).
The female penalty is about 40% for the Soviet period and during 2001-2015, and about 50%
during the 1990s. Tenure is imprecise for most years (a known peculiarity of the transitional
economies). The premium for working in Moscow or Saint Petersburg is negligible before 1994,
but approximately 60% for the remaining years. Foreign company premium is approximately
50% relative to the state-operated businesses throughout the period, except for the Soviet
period where the coefficients drop out since data is not available. The largest firms have the
highest premium relative to firms sized 0-10, and the premium is reducing with the firms' size
generally in monotone fashing. However, the coefficients are imprecise for smaller firms for
most years.
The next section develops a structural model and establishes that the returns
are driven primarily by the demand side.
3 Structural evidence
3.1 Theoretical model
My model is similar to many other previous models that study the evolution
of wage premiums (e.g., Autor, Katz, and Krueger 1998; Autor, Katz, and
Kearney 2008; Card and Lemieux 2001; Freeman and Katz 2007; Johnson and
Keane 2013; Katz and Autor 1999), with the work by Manacorda, Sanchez-
Paramo, and Schady (2010) being the closest counterpart. The innovation of
my model is inspired by the reduced-form evidence of the previous section. I
start by introducing the model's basics.
The model assumes that workers are risk-neutral labor income maximizers,
demand is a function of the marginal productivity of labor, and supply is exoge-
nously given. Further, wages are determined by the interaction of a downward
sloping labor demand curve and a vertical labor supply curve. The representa-
tive firm has two labor inputs with different skill levels, with capital maintained
in the background. The possibility of substituting labor inputs is fixed. Then:
ut +αtNρ
where Ytis total output; Ais skill-neutral technological change; Nis employ-
ment; udenotes workers with high school education to whom (in line with the
previous studies) I refer as `unskilled'; sdenotes worker with a bachelor degree;
tis a year, and ρ < 1 is a function of the elasticity of substitution between
skilled and unskilled labor. Denote this elasticity of substitution by σe, where
/(1ρ). The parameter αtis a measure of the relative productivity of
skilled workers relative to unskilled workers at year t. The coefficient on Nut
is normalized to one. This transformation defines the units of measurements of
One of the consensuses of the economic transition is that it favored the young
(Fleisher, Sabirianova Peter, and Wang 2005). I then allow for differences in
productivity across workers with the same level of education but of different
age. I model the employment of each skill group as a productivity-weighted
CES combination of all age groups of individuals in that skill level. That is:
Njt =
βja Nδ
jat 1
j∈ {s, u},(5)
where adenotes a generic age group, and δis a function of the elasticity of
substitution between different age groups. This elasticity of substitution, σa,
where σa=1
/(1δ), is assumed to be the same across skill groups and for any pair
of age-specific inputs, βja is a measure of the relative productivity of age-group
awith skill level j.
Now the innovation. The model allows for the age-specific supply of skilled
labor to be a CES combination of the two education groups: LB and NLB
specializations. This implies:
Nsat =γaNθ
at +Nθ
where γais a measure of the relative productivity of LB graduates, denoted ,
relative to NLB specializations, denoted , and σs=1
/(1θ)is the elasticity of
substitution between these two groups. When θ= 1, workers with LB and NLB
specializations are perfect substitutes for each other, and my model is identical
to the one proposed by Card and Lemieux (2001).
3.2 Empirical strategy
After price shock therapy, wages were set via informal plant-level bargaining
over which unions had little influence (Brainerd 1998); therefore, it is possible
to assume that wages paid in the equilibrium reflect the relative productivity
of workers. Then to fit the model to the data, equations (4), (5) and (6) can
be manipulated to derive expressions for the wages of unskilled labor and of LB
and NLB graduates of age aat time t:
wuat =Xt+ ln βua 1
nut 1
(nuat nut) (7)
wat =Xt+ ln αt+ ln βsa + ln γa1
nst 1
(nsat nst)1
(nat nsat)
wat =Xt+ ln αt+ ln βsa 1
nst 1
(nsat nst)1
(nat nsat) (9)
where Xt=ρln(At) + (1 ρ) ln(Yt), n= ln N,w= ln Wand Wdenotes
To estimate parameters, I follow the strategy proposed by Card and Lemieux
(2001), appropriately modified to account for the fact that our production func-
tion is modeled as a nested CES process with three production inputs. The
estimation is a three-step process:
At the first step I estimate σsand γa. To do that I subtract (8) from (9):
wat wat =da1
(nat nat),(10)
where dais a set of unrestricted age dummies, which accounts for log relative
productivity of LB to NLB workers (lnγa). In practice, log wage differentials
between LB and NLB workers by age and time are regressed on their relative
labor supply plus age dummies to obtain estimates of γaand σs. These estimates
are then used to compute Nsat in (6).
At the second step I estimate the elasticity of substitution between age
groups, σa, and of all age-specific productivity measures, {βja}j∈{s,u}, that are
then used to construct {Njt }j∈{s,u}in (5). After some manipulations of (7) to
(9) it is possible to obtain:
weat wuat =dt+dea 1
(nsat nuat)1
(neat nsat)e∈ {, }(11)
where dea represent unrestricted age-education effects and dtrepresent unre-
stricted time effects. In particular da = ln βsa ln βua + ln γa,da =da ln γa
and dt= ln αt(1
/σa)(nst nst)
This exercise produces an estimate for σa(as well as a new estimate of σs)
which can then be plugged back into (8) and (9) to obtain:
wuat +1
nuat =dut + ln βua (12)
wat +1
nsat +1
(nat nsat)ln γa=dst + ln βsa (13)
wat +1
nsat +1
(nat nsat) = dst + ln βsa (14)
where the left-hand side of each equation represents (log) wages corrected for
labor supply, dut =Xt(1
/σa)nut and dst =Xt+ ln αt(1
In practice, the adjusted (log) wages are regressed on skill j∈ {s, u}dummies
interacted with age dummies to produce the estimated age effects {βja }j∈{s,u}.
At the end of this step we take these estimates and σato construct {Nj t}j∈{s,u}
in (5).
At the third and last step, I produce an estimate of the elasticity of substi-
tution between skilled and unskilled workers, σe, (7) to (9) and assuming that
the relative demand for skilled versus unskilled workers follows a linear trend
over time with 4 period-specific slopes, so that ln αt=pf0p×p+pf1pt×p
(where pis an array of dummies for 4 periods and tis trend) I obtain:
weat wuat =
(nsat nuat)
σa(nsat nst)(nuat nut )
(neat nsat)e={, },
where the left-hand side of the equation represents (log) wage differentials of
(skilled) workers possessing LB and NLB university specializations relative to
those (unskilled) workers possessing secondary school diplomas; the coefficients
f1pcapture demand-side changes favoring skilled workers. The coefficient on
the first labor supply term provides an estimate of σe. The coefficients on the
other terms, provide new estimates of σaand σs.
Once the parameters are estimated, it is possible to track the changes in
demand for LB graduates. In terms of this model, it is an increase in the
marginal productivity of LB graduates, γ. Equations (4), (5) and (6) can be
manipulated to get the ratio of the wages of university graduates with LB and
NLB specialization:
ln Wt
Wt = ln γ
ln Nt
Nt .(16)
The wage premium of LB graduates relative to NLB depends on their pro-
ductivity, their relative supply, and the technological capacity to substitute LB
with NLB. This expression can be further manipulated to get the direct measure
of demand
Dtσsln γ
1γ= ln WtNt
WtNt + (σs1) ln Wt
Wt ,(17)
The changes in the log relative demand, Dt, equals the sum of the change
in the log relative wage and a term that depends positively (negatively) on the
change in the log college wage premium when σs>1 (σs<1). If σs= 1,
then changes in the log relative demand is directly given by changes in the log
relative wage. This framework assumes that a change in the relative supply of
skilled and unskilled workers or of different age does not affect the premium to
LB and NLB graduates (Goldin and Katz 2009, Ch. 8).
3.3 Data
I use the same data used in Section 2, with years limited to 1995 – 2015, where
the loss of information on majors is minimal. Variable age is regrouped into
a categorical variable with 5 years in each category. Workers are regrouped
into three educational groups: higher school education, LB, and NLB bachelor
degree graduate. Resulting data consists of 3 levels of education (u, , ), 9 age
groups (a), 17 years (t). The supply of each education-age-year (n) cell is the
number of workers in each cell. For the wage (w), I use Model (2) to predict
wages for each cell, with setting all other residency, occupation, and gender
variables to 0.
3.4 Results
Table 4: Structural results
Step 1 Step 2 Step 3
Relative Wage by Age and Time
LB Relative to NLB Skilled Relative to Unskilled
1s-0.219*** -0.236*** -0.213***
(0.024) (0.035) (0.045)
1a-0.323*** -0.312***
(0.046) (0.069)
1995–1997 0.118***
1998–2004 0.098***
2005–2010 0.049**
2011–2015 0.019
N467 479 479
R20.83 0.97 0.93
Notes: The GLS estimates of Equation (10), Equation (11), and Equa-
tion (15).
*p < 0.05, ** p < 0.01, *** p < 0.001. Source: RLMS.
Table 4contains the generalized least squares (GLS) estimates for each step.
All regressions are weighted by the inverse of the sampling variance of the de-
pendent variable. This is done because relative wages are computed on samples
of different sizes and vary in precision. The first column reports the GLS esti-
mates of Equation (10). The reported coefficient is the negative of the inverse
of the elasticity of substitution between LB and NLB workers (note that this al-
ready identifies the key parameters that can be used to understand the demand
shifts between LB and NLB). The second column reports the GLS estimates of
Equation (11). The additionally reported coefficient in the second line is the
negative of the inverse of the elasticity of substitution between workers of dif-
ferent ages. The third column reports the GLS estimates of Equation (15). The
new coefficient in the third row is the negative of the inverse of the elasticity
of substitution between skilled and unskilled labor. The last four rows report
trends in the relative demand for skilled labor. The trend is parametrized as an
annual change per period.
The estimates in the first step clearly show that LB and NLB graduates
are not perfect substitutes. Estimates of the coefficient 1sare similar in
all three columns, providing internal consistent. Same for the coefficient 1e
reported in the middle and the right column.
The firms have the most trouble substituting skilled labor on unskilled, with
the elasticity of substitution of 2.5(= 1
/0.398; the third line in Table 4). This
value is somewhat higher relative to the United States (Katz 1999), but sim-
ilar to Latin American (Manacorda, Sanchez-Paramo, and Schady 2010) and
previous works on Russia (Sabirianova Peter 2003).
Substitution of workers of different ages is the next most troubling, with
the elasticity of 3.2(= 1
/0.312; the second line in Table 4). Again, this value is
somewhat higher relative to the United States or Canada (Card and Lemieux
2001). This generally supports the claim that transition has favored the young
(Fleisher, Sabirianova Peter, and Wang 2005). However, it clarifies that previous
studies may have to some extend confused an increase in demand for young
workers with an increase in demand of LB graduates, who are younger than
Substitution of workers with LB and NLB majors is the least troublesome,
with an elasticity of 4.7(= 1
/0.213; the first line in Table 4). This parameter
does not have a direct comparison with previous studies. Importantly, this
substitution is not one-on-one.
The last four lines report trends in the shift in the demand from unskilled to
word skilled labor. The annual trend for 1994-1997 is 0.118. To the best of my
knowledge, this is the highest estimate in the SBTC literature. A comparable
estimate for Mexico during 1980 and 1990 reported by Manacorda, Sanchez-
Paramo, and Schady (2010) ranges from 0.40 to 0.50. The shift for Russia is so
large because of the artificial compression of relative wages in the Soviet period.
As a result, wages had to adjust quickly to the new realities. The annual trend
gets smaller as time passes by, and for 2011-2015 no shifts in the demand can
be estimated.
Table 5reports changes in the relative annual demand for LB graduates.
The leftmost column reports implied relative demand using the elasticity of
substitution between LB and NLB graduates in just presented Table 4. The
data is consistent with a demand shock after 1990. This implies that the esti-
mates in Figure 5produced with Specification (2) are explained primarily by a
rapid increase in demand for LB graduates following the decentralization and a
Table 5: Relative demand on LB graduates
4.7 3.7 2.7 1.7
1985-1990 0.01 -0.17 -0.32 -0.44
1994-1996 7.25 6.45 5.68 5.10
1998-2003 1.10 1.06 0.59 0.16
2005-2009 1.08 0.82 0.26 0.12
2010-2015 0.72 0.22 -0.10 -0.25
Notes: Calculated with (17)
Source: RLMS 1994-2015
gradual catch-up of the supply.
The following section shows evidence that the differential is not unique to
Russia and likely happened in other transitional economies.
4 Cross-country evidence
As summarized by Violante (2016), most economic theories that explicitly for-
mulate an economic mechanism to explain SBTC adopt Nelson and Phelps
(1966)'s view of human capital. In their view, workers' education positively af-
fects the speed of the practical implementation of available technologies. Their
theory further suggests that any increase in the wage premium is transitory.
Only in the early adoption phase of new technology can those workers who
adapt more quickly reap some benefits. As time goes by, enough workers will
know how to work with the new technology to offset the wage differential (Caselli
1999; Galor and Moav 2000; Greenwood and Yorukoglu 1997).
This conceptual interpretation emphasizes the effects of learning during
episodes of radical technical change, which is in line with the productivity de-
cline that occurred in most developed economies in the 1980s. At the beginning
of the deployment of new technology, the output may temporarily decrease as
workers and firms learn how to use the new techniques (Aghion 2002; Horn-
stein and Krusell 1996). If indeed the economic transition is viewed as a form
of SBTC, then there should be a rapid creation of a class of better-paid jobs
complementary to the new technologies. This is the key empirical finding of this
paper, as summarized in Figure 5. However, there should also be a period of
learning/adaptation of new technologies. The transformational recession may
have been exactly that.
The model in Section 3.1 can be used to illustrate a potential of the output
decline during the transformational recession. In particular, the following can
be shown:
∂γ <0 if γ > Nθ
The output may drop in response to an increase in demand for LB graduates
depending on the composition of skills (Nθ
) and a degree of complemen-
tarity between LB and NLB (θ). A larger γ(it is larger for a more complex
economic system or industry, as they require more coordination) can adversely
affect economic output. Alternatively, for a given γ, a lack of LB graduates can
adversely affect economic output, as the sufficiency inequality is increasingly
harder to satisfy if Nincreases.
In sum, the model shows that a lack of LB personnel drives the (discovered in
this study) irregularity in the labor market and may drop the economic output.
A statistical tautology to this hypothesis is that the size of differential and the
output drop are positively correlated. That is, the transitional economies that
had the largest decrease in economic output also had the largest transitory wage
premium for LB graduates.
Figure 10: GDP loss and recomposition of skills during transition
Kyrgyz Republic
Czech Republic
FYR Macedonia
Slovak Republic
Y = 0.1854X + 0.2337
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
Share of graduates with law and business specialisation post transition
GDP loss during transition
Former Soviet Union republics
Central and Eastern Europe countries
Notes: The GDP loss is a percentage decrease of real GDP during the transitional recession.
Share of LB graduates is their percentage in tertiary institutions after the transitional reces-
Source: Fischer and Sahay (2000), UIS Data Dictionary for Education Statistics, National
Statistical Agencies.
A complication with testing this implication is that there are no available
micro-level datasets for most transitional economies to detect the transitory
wage premium by adopting methods similar to those used in this paper. The
Russian dataset is unique in this respect. Fortunately, the validity and gener-
ality of the hypothesis can be tested without microlevel datasets. The Russian
example shows that the wage differential changed the composition of skills avail-
able in the labor market (the very purpose of the differential). As labor markets
are competitive and applicants are rational, an unusually high proportion of en-
rollments into LB specialization implies an existence of a persistent price signal
(i.e., the desired LB wage differential) that governs the enrollment. Therefore,
the expressed hypotheses, if true, imply that the transitional economies that
had a larger decrease in economic output during the transformational recession
also had a larger fraction of students enrolled into LB specializations after the
transformational recession.
Figure 10 demonstrates that data confirms the implication. The figure plots
the fraction of LB students in the higher education system two years after
the transformational recession against the percentage loss of GDP during the
transformational recession for 26 countries. For example, Uzbekistan lost 14%
of GDP from 1989 till 1995 and in 1997 had 25% of all university students with
specialization LB; Georgia lost 75% of GDP from 1988 till 1994 and in 1996 had
39%. This approach is motivated by the Russian case: the recession stopped in
1998, and the change in the skills became evident about two years after.
Indeed the SBTC view (its peculiar version with the demand shift taking
place within the skilled labor) on the economic transition seems to reconcile
several existing cross-regional empirical inconsistencies. In China, output in-
creased during the transition period, while in CEECs, the decrease was not as
dramatic as in the FSU. This can be explained by absence of a massive demand-
side shock for skills left unmatched with an appropriate supply, as hypothesized
in this paper.
The shock could have been lessened if: (1) complex industries had been
decentralized gradually; (2) there had been no complex industries; or (3) the
degree of penetration of the centralized planning was not extreme. All three con-
ditions were violated by the FSU republics. Conversely, in the Chinese economy,
the transitional decentralization was gradual, there was a lower level of indus-
trialization, and the penetration of centralized control was also relatively low.
Similarly, while CEECs were decentralized quickly and had complex capital-
intensive industries, their reliance on central planning was relatively short-lived
and selective (therefore, the LB graduate shortage was not so pronounced), as
they only joined the socialist camp after World War II. In contrast, the So-
viet republics were centralized in the 1920s (and the LB graduate shortage was
It can be argued then that the Russian transformational recession could have
been reduced if policymakers had acknowledged the dependency of the imple-
mentation of technologies on the supply of workers with complementary skills.
Conversely, unlike most of the Russian economy, the Russian education system
remained the state's property, which led to a decade of underfunding as the
government struggled to balance its finances. In 1994, the parliament passed a
series of laws that formally allowed private firms to invest in education (Belskaya
and Sabirianova Peter 2014); however, a fully functional regulatory framework
that allowed education to be privately financed by firms and applicants was not
institutionalized until the 2000s.
Thus, while the popularity of LB specializations due to high returns was
unprecedented (Public Opinion Foundation 1998,2006), the education system
was insensitive to labor market price signals for almost a decade after the begin-
ning of the reforms. Given the well-documented structural similarities between
Russian and Soviet economies (Ananyev and Guriev 2018; Mikhailova 2012),
and that the differential leveled out in 2007/2008 (exactly when the Russian
economy regained its size to that of late Soviet Russia, see Figure 1), it can be
conjectured that 6–7% of the labor force should have been supplying their LB
skills to foster the adoption of new organizational technologies in the market
economy, which would have cushioned the transformational recession. However,
at the beginning of the transitional reform, only 2–3% of the labor force had such
skills, complicating the adoption of technologies and exacerbating the recession.
5 Conclusion
In their arguably most influential paper on economic transition, Blanchard and
Kremer (1997) wrote `Once Humpty-Dumpty has fallen down, all the King's
horses and all the King's men cannot put him back together again.' The authors
referred to a breakdown of complex chains of production during the economic
transition following the dismissal of the central planner because of information
and contractual imperfections. In the current paper, I demonstrate a demand-
side driven between-groups wage differential in the Russian labor market that
induced a substantial change in the composition of employed skills. It appears
that `Humpty-Dumpty' ultimately was put together by LB graduates. On the
descriptive side, the differential provides a parsimonious explanation for the
relatively low college wage premiums from 1994 to 1996 and the decrease in
college wage premiums from 1998 to 2015. On the substantive side, the change
in skills composition ultimately replaced the central planner's functions.
To establish the differential, I estimate the wage equation. To confirm that
the demand was responsible for the differential, I fit the data into an SBTC
model that allows for imperfect substitution between LB and NLB graduates.
I then use my SBTC model to show that an increase in demand for LB gradu-
ates might drop GDP in case of scarcity of LB graduates. Finally, to confirm
the plausibility of this mechanism, I offer a positive cross-country correlation
between the GDP loss during the transition and the fraction of LB students in
the higher education system after the transition. Collectively, my finding re-
duces the economic transition into a special (and extreme) case of SBTC. This
reaffirms a recent conclusion that there is no reason to consider the economics
of transition as a separate subfield (Olofsg˚ard, Wachtel, and Becker 2018).
This paper shows the importance of studying the technology-induced differ-
entials within skilled labor. These differentials are known to present in modern
economies today (Cattani and Pedrini 2021; Elias and Purcell 2013) due to a
new wave of adaptation of high return technologies. These differentials deserve
further scrutiny on empirical and conceptual levels, as they are known to cause
unexpected and (as the Russian case shows) severe drops in economic perfor-
mance and income equality. These two effects in isolation are temporal, but if
they are incorrectly interpreted and the policy response is suboptimal, adverse
consequences are possible.
The results of this paper are also relevant in the context of international ef-
forts to deter Russia's military actions against Ukraine. The effects of the sanc-
tions mirror the effect of transition. After the transition, the Russian economy
became tightly integrated into new global production chains. The sanctions and
the exodus of foreign companies are being explicitly targeted to disrupt these
chains, forcing the Russian economy into a new large-scale structural transfor-
mation (Bank of Russia 2022). The renewed economic development in Russia –
as during the transition period– will only be possible after the new production
chains are reestablished. The finding of this paper clearly shows that the adap-
tation of the labor market has a first-order significance for this process. The
Russian government should deregulate the education sector to allow new price
signals to reorient the skill structure for the new economy's needs.
Instead, the Russian government harshly regulates the education system
and uses it as a tool of propaganda and coercion. Students who protest the
war are expelled, while academics who do not sign the letters supporting war
are discriminated against. Sanctioning the government educational sector will
improve the effectiveness of the existing sanctions against the current regime by
preventing its structural transformation and preservation. It will also reduce the
Russian government's capacity to suppress dissent and encourage the transfer
of educational resources toward private institutions, reducing state control and
enabling regime change while also curbing future autocratic tendencies.
The designers of the Russian economic transition knew that failure to rebuild
education and criminal justice systems risks future autocratization, but they
only had an opportunity to reform trade and prices (Gaidar 2007). The future
readers of this paper will likely see the economic transition of the 1990s and the
reform of the late 2020s (which will inevitably follow after Putin's disastrous
rule) as a single process of establishing the modern post-imperial Russia.
Abadie, A., S. Athey, G. Imbens, and J. Wooldridge (2017). When Should You
Adjust Standard Errors for Clustering? url:W(cit. on p. 6).
Acemoglu, D. (Dec. 1999). “Changes in Unemployment and Wage Inequality:
An Alternative Theory and Some Evidence”. In: American Economic Review
89.5, pp. 1259–1278. url:W(cit. on p. 1).
Acemoglu, D. and D. H. Autor (2011). “Chapter 12 - Skills, Tasks and Technolo-
gies: Implications for Employment and Earnings”. In: ed. by D. Card and O.
Ashenfelter. Vol. 4. Handbook of Labor Economics. Elsevier, pp. 1043–1171.
url:W(cit. on pp. 1,6).
Aghion, P. (2002). “Schumpeterian Growth Theory and the Dynamics of Income
Inequality”. In: Econometrica 70.3, pp. 855–882. url:W(cit. on p. 23).
Alexeev, S. (2022). “Wage equation misrepresents gay wage discrimination: over-
looked evidence from Russia”. In: International Journal of Manpower.url:
W(cit. on p. 8).
Ananyev, M. and S. Guriev (2018). “Effect of Income on Trust Evidence from
the 2009 Economic Crisis in Russia”. In: The Economic Journal.url:W
(cit. on p. 26).
Andr´en, D., J. S. Earle, and D. S˘ap˘atoru (2005). “The Wage Effects of Schooling
Under Socialism and in Transition: Evidence from Romania, 1950–2000”. In:
Journal of Comparative Economics 33.2. Returns to Schooling in Transition:
The Chinese, European, and Russian Experiences, pp. 300–323. url:W(cit.
on p. 1).
Angrist, J. D. and J.-S. Pischke (June 2010). “The Credibility Revolution in
Empirical Economics: How Better Research Design Is Taking the Con out of
Econometrics”. In: Journal of Economic Perspectives 24.2, pp. 3–30. url:
W(cit. on p. 6).
Arrow, K. J. and G. Debreu (1954). “Existence of an Equilibrium for a Com-
petitive Economy”. In: Econometrica 22.3, pp. 265–290. url:W(cit. on
p. 3).
Atkeson, A. and P. J. Kehoe (1995). “Industry Evolution and Transition: Mea-
suring Investment in Organization Capital”. In: Federal Reserve Bank of
Minneapolis, Research Department.url:W(cit. on p. 3).
(2005). “Modeling and Measuring Organization Capital”. In: Journal of Po-
litical Economy 113.5, pp. 1026–1053. url:W(cit. on p. 3).
Autor, D. H., L. F. Katz, and A. B. Krueger (1998). “Computing Inequality:
Have Computers Changed the Labor Market?” In: The Quarterly Journal
of Economics 113.4, pp. 1169–1213. url:W(cit. on p. 17).
Autor, D. H., L. F. Katz, and M. S. Kearney (2008). “Trends in US wage inequal-
ity: Revising the revisionists”. In: The Review of economics and statistics
90.2, pp. 300–323 (cit. on p. 17).
Bank of Russia (Mar. 2022). The Bank of Russia keeps the key rate at 20% p.a.
Tech. rep. url:W(cit. on p. 27).
Belskaya, O. and K. Sabirianova Peter (2014). “How Does the Expansion of
Higher Education Change the Returns to College Quality? Insights from
Sixty Years of Russian History”. In: 9th IZA/World Bank Conference on
Employment and Development.url:W(cit. on pp. 3,26).
Belskaya, V., K. Sabirianova Peter, and C. M. Posso (2020). “Heterogeneity in
the Effect of College Expansion Policy on Wages: Evidence from the Russian
Labor Market”. In: Journal of Human Capital.url:W(cit. on pp. 3,5,9).
Blanchard, O. and M. Kremer (1997). “Disorganization”. In: The Quarterly
Journal of Economics 112.4, pp. 1091–1126. url:W(cit. on pp. 3,26).
Brainerd, E. (1998). “Winners and Losers in Russia's Economic Transition”. In:
The American Economic Review 88.5, pp. 1094–1116. url:W(cit. on pp. 2,
— (2000). “Women in Transition: Changes in Gender Wage Differentials in
Eastern Europe and the Former Soviet Union”. In: ILR Review 54.1, pp. 138–
162. url:W(cit. on p. 1).
Bresnahan, T. F., E. Brynjolfsson, and L. M. Hitt (Feb. 2002). “Information
Technology, Workplace Organization, and the Demand for Skilled Labor:
Firm-Level Evidence”. In: The Quarterly Journal of Economics 117.1, pp. 339–
376. url:W(cit. on p. 1).
Card, D. and T. Lemieux (2001). “Can Falling Supply Explain the Rising Return
to College for Younger Men? A Cohort-Based Analysis*”. In: The Quarterly
Journal of Economics 116.2, pp. 705–746. url:W(cit. on pp. 17,18,22).
Carnoy, M., P. K. Loyalka, G. Androushchak, and A. Proudnikova (2012). “The
Economic Returns to Higher Education in the BRIC Countries and their Im-
plications for Higher Education Expansion”. In: Higher School of Economics
Research Paper No. WP BRP 02/EDU/2012.url:W(cit. on p. 8).
Caroli, E. and J. Van Reenen (Nov. 2001). “Skill-Biased Organizational Change?
Evidence from A Panel of British and French Establishments”. In: The Quar-
terly Journal of Economics 116.4, pp. 1449–1492. url:W(cit. on p. 1).
Caselli, F. (1999). “Technological Revolutions”. In: American economic review
89.1, pp. 78–102. url:W(cit. on p. 23).
Cattani, L. and G. Pedrini (2021). “Opening the black-box of graduates' hor-
izontal skills: diverging labour market outcomes in Italy”. In: Studies in
Higher Education 46.11, pp. 2387–2404. url:W(cit. on p. 27).
Chase, R. S. (1998). “Markets for Communist Human Capital: Returns to Ed-
ucation and Experience in the Czech Republic and Slovakia”. In: Industrial
and Labor Relations Review 51.3, pp. 401–423. url:W(cit. on p. 1).
Cheidvasser, S. and H. Ben´ıtez-Silva (2007). “The Educated Russian's Curse:
Returns to Education in the Russian Federation during the 1990s”. In:
Labour 21.1, pp. 1–41. url:W(cit. on pp. 2,8).
Chernozhukov, V. (2005). “Extremal Quantile Regression”. In: The Annals of
Statistics 33.2, pp. 806–839. url:W(cit. on p. 5).
Chin, A., C. Juhn, and P. Thompson (2006). “Technical Change and the De-
mand for Skills during the Second Industrial Revolution: Evidence from the
Merchant Marine, 1891–1912”. In: The Review of Economics and Statistics
88.3, pp. 572–578. url:W(cit. on p. 1).
Clark, A. (2003). “Returns to Human Capital Investment in a Transition Econ-
omy: The Case of Russia, 1994-1998”. In: International Journal of Manpower
24.1, pp. 11–30. url:W(cit. on p. 2).
Combes, P.-P. and L. Gobillon (2015). “Chapter 5 - The Empirics of Agglom-
eration Economies”. In: Handbook of Regional and Urban Economics. Ed.
by G. Duranton, J. V. Henderson, and W. C. Strange. Vol. 5. Handbook of
Regional and Urban Economics. Elsevier, pp. 247–348. url:W(cit. on p. 5).
Dessein, W. and T. Santos (2006). “Adaptive Organizations”. In: Journal of
Political Economy 114.5, pp. 956–995. url:W(cit. on p. 1).
Elias, P. and K. Purcell (2013). “Classifying graduate occupations for the knowl-
edge society”. In: Institute for employment research. University of Warwick.
Future track working paper 5. url:W(cit. on p. 27).
Ericson, R. E. (1983). “A Difficulty with the “command” Allocation Mecha-
nism”. In: Journal of Economic Theory 31.1, pp. 1–26. url:W(cit. on
p. 3).
— (1991). “The Classical Soviet-Type Economy: Nature of the System and
Implications for Reform”. In: The Journal of Economic Perspectives 5.4,
pp. 11–27. url:W(cit. on p. 3).
(Mar. 2006). “Command Versus `Shadow': The Conflicted Soul of the Soviet
Economy”. In: Comparative Economic Studies 48.1, pp. 50–76. url:W(cit.
on p. 3).
— (2017). “Command Economy”. In: The New Palgrave Dictionary of Eco-
nomics. London: Palgrave Macmillan UK, pp. 1–16. url:W(cit. on p. 3).
Fischer, S. and R. Sahay (Apr. 2000). The Transition Economies After Ten
Years. Working Paper 7664. National Bureau of Economic Research. url:
W(cit. on p. 24).
Fleisher, B., K. Sabirianova Peter, and X. Wang (2005). “Returns to Skills and
the Speed of Reforms: Evidence from Central and Eastern Europe, China,
and Russia”. In: Journal of Comparative Economics 33.2, pp. 351–370. url:
W(cit. on pp. 1,17,22).
Freeman, R. and L. Katz (2007). Differences and Changes in Wage Structures.
National Bureau of Economic Research Comparative Labor Markets Series.
University of Chicago Press. url:W(cit. on p. 17).
Gaidar, E. (2007). Collapse of an Empire: Lessons for Modern Russia. Brookings
Institution Press. url:W(cit. on p. 27).
Galor, O. and O. Moav (2000). “Ability-Biased Technological Transition, Wage
Inequality, and Economic Growth*”. In: The Quarterly Journal of Eco-
nomics 115.2, pp. 469–497. url:W(cit. on p. 23).
Garicano, L. and E. Rossi-Hansberg (2006). “Organization and Inequality in
a Knowledge Economy*”. In: The Quarterly Journal of Economics 121.4,
pp. 1383–1435. url:W(cit. on p. 1).
Gimpelson, V. and R. Kapeliushnikov (2011). “Labor Market Adjustment: is
Russia Different?” In: IZA Discussion Paper.url:W(cit. on p. 2).
Goldin, C. and L. Katz (2009). The Race between Education and Technology.
Harvard University Press. url:W(cit. on p. 20).
Gorodnichenko, Y. and K. S. Peter (2005). “Returns to Schooling in Russia
and Ukraine: A Semiparametric Approach to Cross-country Comparative
Analysis”. In: Journal of Comparative Economics 33.2. Returns to Schooling
in Transition: The Chinese, European, and Russian Experiences, pp. 324–
350. url:W(cit. on pp. 1,8,16).
Gorodnichenko, Y., K. S. Peter, and D. Stolyarov (2010). “Inequality and Volatil-
ity Moderation in Russia: Evidence From Micro-level Panel Data on Con-
sumption and Income”. In: Review of Economic Dynamics 13.1, pp. 209–
237. url:W(cit. on p. 3).
Greenwood, J. and M. Yorukoglu (1997). “1974”. In: Carnegie-Rochester Con-
ference Series on Public Policy 46, pp. 49–95. url:W(cit. on p. 23).
Gregory, P. R. and I. L. Collier (1988). “Unemployment in the Soviet Union:
Evidence from the Soviet Interview Project”. In: The American Economic
Review 78.4, pp. 613–632. url:W(cit. on p. 7).
Gregory, P. R. and J. E. Kohlhase (1988). “The Earnings of Soviet Workers:
Evidence from the Soviet Interview Project”. In: The Review of Economics
and Statistics 70.1, pp. 23–35. url:W(cit. on pp. 1,8,9).
Gregory, P. R., M. Mokhtari, and W. Schrettl (1999). “Do the Russians Really
Save That Much?—Alternate Estimates from the Russian Longitudinal Mon-
itoring Survey”. In: The Review of Economics and Statistics 81.4, pp. 694–
703. url:W(cit. on p. 7).
Grossman, G. (1962). “The Structure and Organization of the Soviet Economy”.
In: Slavic Review 21.2, pp. 203–222. url:W(cit. on p. 3).
Grossman, G. (1963). “Notes for a Theory of the Command Economy”. In:
Soviet Studies 15.2, pp. 101–123. url:W(cit. on p. 3).
(1966). Gold and the Sword: Money in the Soviet Command Economy. Reprint
(University of California, Berkeley. Institute of International Studies). Cen-
ter for Slavic and East European Studies. url:W(cit. on p. 3).
Hornstein, A. and P. Krusell (1996). “Can Technology Improvements Cause
Productivity Slowdowns?” In: NBER Macroeconomics Annual 11, pp. 209–
259. url:W(cit. on p. 23).
Ickes, B. W. (2016). “Output Fall – Transformational Recession”. In: The New
Palgrave Dictionary of Economics. London: Palgrave Macmillan UK, pp. 1–
13. url:W(cit. on p. 3).
Johnson, M. and M. P. Keane (2013). “A Dynamic Equilibrium Model of the US
Wage Structure, 1968–1996”. In: Journal of Labor Economics 31.1, pp. 1–49.
url:W(cit. on p. 17).
Jovanovic, B. and P. L. Rousseau (2005). “Chapter 18 - General Purpose Tech-
nologies”. In: ed. by P. Aghion and S. N. Durlauf. Vol. 1. Handbook of
Economic Growth. Elsevier, pp. 1181–1224. url:W(cit. on p. 1).
Katz, K. (1999). “Were There No Returns to Education in the USSR? Estimates
from Soviet-period Household Data”. In: Labour Economics 6.3, pp. 417–434.
url:W(cit. on pp. 2,14,22).
Katz, L. F. and D. H. Autor (1999). “Chapter 26 - Changes in the Wage Struc-
ture and Earnings Inequality”. In: ed. by O. C. Ashenfelter and D. Card.
Vol. 3. Handbook of Labor Economics. Elsevier, pp. 1463–1555. url:W(cit.
on pp. 1,6,17).
Kirkeboen, L. J., E. Leuven, and M. Mogstad (May 2016). “Field of Study,
Earnings, and Self-Selection”. In: The Quarterly Journal of Economics 131.3,
pp. 1057–1111. url:W(cit. on p. 6).
Koenker, R., V. Chernozhukov, X. He, and L. Peng (2018). Handbook of Quantile
Regression. A Chapman & Hall book. CRC Press, Taylor & Francis Group.
url:W(cit. on p. 5).
Kornai, J. (1979). “Resource-Constrained versus Demand-Constrained Systems”.
In: Econometrica 47.4, pp. 801–819. url:W(cit. on p. 3).
Kozyreva, P., M. Kosolapov, and B. M. Popkin (2016). “Data Resource Profile.
RLMS Phase 2: Monitoring the Economic and Health Situation in Russia
1994 – 2013”. In: International Journal of Epidemiology 45.2, pp. 395–401.
url:W(cit. on p. 6).
Krueger, A. B. and J.-S. Pischke (1995). “A Comparative Analysis of East and
West German Labor Markets: Before and After Unification”. In: Differences
and Changes in Wage Structures. NBER Chapters. National Bureau of Eco-
nomic Research, Inc, pp. 405–446. url:W(cit. on p. 1).
Kyui, N. (2016). “Expansion of Higher Education, Employment and Wages:
Evidence from the Russian Transition”. In: Labour Economics 39, pp. 68–
87. url:W(cit. on pp. 3,5,8).
Manacorda, M., C. Sanchez-Paramo, and N. Schady (2010). “Changes in Re-
turns to Education in Latin America: The Role of Demand and Supply of
Skills”. In: ILR Review 63.2, pp. 307–326. url:W(cit. on pp. 17,22).
Mikhailova, T. (Sept. 2012). Gulag, WWII and the Long-run Patterns of Soviet
City Growth. eng. Kiel und Hamburg. url:W(cit. on p. 26).
Milgrom, P. and J. Roberts (1990). “The Economics of Modern Manufactur-
ing: Technology, Strategy, and Organization”. In: The American Economic
Review 80.3, pp. 511–528. url:W(cit. on p. 1).
unich, D., J. Svejnar, and K. Terrell (2005). “Returns to Human Capital
Under The Communist Wage Grid and During the Transition to a Market
Economy”. In: The Review of Economics and Statistics 87.1, pp. 100–123.
url:W(cit. on p. 8).
Nelson, R. R. and E. S. Phelps (1966). “Investment in Humans, Technologi-
cal Diffusion, and Economic Growth”. In: The American Economic Review
56.1/2, pp. 69–75. url:W(cit. on p. 23).
Newell, A. and B. Reilly (1996). “The Gender Wage Gap in Russia: Some Em-
pirical Evidence”. In: Labour Economics 3.3, pp. 337–356. url:W(cit. on
p. 2).
(1999). “Rates of Return to Educational Qualifications in the Transitional
Economies”. In: Education Economics 7.1, pp. 67–84. url:W(cit. on p. 2).
Novokmet, F., T. Piketty, and G. Zucman (2018). “From Soviets to Oligarchs:
Inequality and Property in Russia 1905-2016”. In: The Journal of Economic
Inequality 16.2, pp. 189–223. url:W(cit. on p. 1).
Ojala, J., J. Pehkonen, and J. Eloranta (2016). “Deskilling and decline in skill
premium during the age of sail: Swedish and Finnish seamen, 1751–1913”.
In: Explorations in Economic History 61, pp. 85–94. url:W(cit. on p. 1).
Olofsg˚ard, A., P. Wachtel, and C. M. Becker (2018). “The economics of transi-
tion literature”. In: Economics of Transition and Institutional Change 26.4,
pp. 827–840. url:W(cit. on p. 27).
Parente, P. M. and J. M. S. Silva (2016). “Quantile Regression with Clustered
Data”. In: Journal of Econometric Methods 5.1, pp. 1–15. url:W(cit. on
pp. 15,16).
Pehkonen, J., T. Neuvonen, and J. Ojala (2019). “Technological change and
wage premiums amongst high-skilled labour”. In: Applied Economics Letters
26.19, pp. 1563–1566. url:W(cit. on p. 1).
Public Opinion Foundation (Apr. 1998). The Most Respected Profession Today
(in Russian language). (Accessed on 06/18/2017). url:W(cit. on p. 26).
Public Opinion Foundation (Nov. 2006). Higher Education: the Relevance and
Popularity of Humanitarian and Social Specialties (in Russian language).
(Accessed on 06/18/2017). url:W(cit. on p. 26).
Roland, G. and T. Verdier (1999). “Transition and the Output Fall”. In: Eco-
nomics of Transition 7.1, pp. 1–28. url:W(cit. on p. 3).
Rousseau, P. L. (2016). “General Purpose Technologies”. In: The New Palgrave
Dictionary of Economics. London: Palgrave Macmillan UK, pp. 1–6. url:
W(cit. on p. 1).
Rutkowski, J. (1996). “High Skills Pay Off: the Changing Wage Structure During
Economic Transition in Poland”. In: Economics of Transition 4.1, pp. 89–
112. url:W(cit. on p. 1).
Sabirianova Peter, K. (2003). “Skill-biased Transition: The Role of Markets,
Institutions, and Technological Change”. In: IZA Discussion Paper No. 893;
William Davidson Institute Working Paper No. 616.url:W(cit. on pp. 1,
Stanovnik, T. and M. Verbiˇc (2005). “Wage and Income Inequality in Slovenia,
1993–2002”. In: Post-Communist Economies 17.3, pp. 381–397. url:W(cit.
on p. 1).
Violante, G. L. (2016). “Skill-Biased Technical Change”. In: The New Palgrave
Dictionary of Economics. London: Palgrave Macmillan UK, pp. 1–6. url:
W(cit. on pp. 1,23).
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Only data from developed countries were used to estimate the sexual orientation difference in wages. This paper is the first, which aims to identify the wage discrimination of gay men in Russia – a country where institutional discrimination and ignorance against gay men are known to present. Gays are identified as men who reported having sex with other men in several waves of the national household survey. A wage equation is used to estimate the gay wage penalty. Extending the wage equation to implement a difference-in-difference design, the paper also evaluates the effect of the gay-propaganda law of 2013 on gay wages. No wage discrimination is identified. The law also has no adverse effect on gay wages. Cross-country comparison and theoretical generalizations are premature, and better identification strategies are needed to understand sexual orientation differences. Policymakers should be aware that in both discriminatory and equitable environments, there may be hidden inequality even if researchers do not detect it. The findings are implausible and add to existing evidence that gay discrimination measured with wage equation suffers from endogeneity and should be interpreted with caution. Particular caution should be exercised in cross-sectional and time-series comparisons, as a tendency to report the orientation honestly and unobserved confounders vary by location and time.
Full-text available
This paper assesses labour market outcomes associated with the acquisition of different graduate competences in the Italian labour markets particularly focusing on ‘soft skills’ developed during HE programs. By using a classification of graduate jobs that distinguishes between different horizontal skills, namley managerial and communication skills, the paper conducts an empirical analysis of the benefits of being employed in such occupations having completed a degree program. These benefits are assessed across time, referring to the period following the crisis and to most recent available data. Results show that only ‘communication’ skills provide higher benefits than discipline-related skills. Moreover, the higher education system seems not to provide higher education graduates with distinctive ‘managerial’ competences compared to non-graduate workers suggesting that these types of social skills can be developed through alternative learning paths, such as work-based experience or other extra-curricular activities.
Full-text available
This study examines the impact of the steam engine, which produced wide and long-lasting economic growth from the 19th century to the early 20th century, on the wages of high-skilled seamen in the Swedish merchant maritime shipping industry. The analysis focuses on the years 1869–1914, which was a transition period during which traditional sailing ships were replaced by steam-powered vessels. The study shows that all high-skilled occupations received wage gains under steam technology. The evidence on wage polarization amongst the high-skill occupations remains subtle, although there is certain evidence that wage premiums vary by occupation.
Full-text available
This paper combines national accounts, survey, wealth and fiscal data (including recently released tax data on high-income taxpayers) in order to provide consistent series on the accumulation and distribution of income and wealth in Russia from the Soviet period until the present day. We find that official survey-based measures vastly under-estimate the rise of inequality since 1990. According to our benchmark estimates, top income shares are now similar to (or higher than) the levels observed in the United States. We also find that inequality has increased substantially more in Russia than in China and other ex-communist countries in Eastern Europe. We relate this finding to the specific transition strategy followed in Russia. According to our benchmark estimates, the wealth held offshore by rich Russians is about three times larger than official net foreign reserves, and is comparable in magnitude to total household financial assets held in Russia.
This paper draws on a natural experiment to identify the relationship between income and trust. We use a unique panel dataset on Russia where GDP experienced an 8 per cent drop in 2009. The effect of the crisis had been uneven among Russian regions because of their differences in industrial structure inherited from the Soviet period. After instrumenting average regional income by Soviet industrial structure, we find that 10 per cent decrease in income is associated with 5 percentage point decrease in social trust. We also find that post‐crisis economic recovery did not fully restore pre‐crisis trust level. This article is protected by copyright. All rights reserved.