Impact of Special Economic Zones on domestic market: Evidence from Russia

Abstract

Place-based policies can be an effective instrument for governments to encourage the economic development of a country. A Special Economic Zone (SEZ) is a place-based policy aimed at attracting FDI, employment growth, and supporting new economic reforms. In addition, for emerging economies an SEZ could be a potential catalyst for development; foreign investors can have a drastic impact on the productivity of domestic firms, revenues, and market shares through the implementation of new technologies and the creation of new firms. However, the effects of SEZs on the domestic market at the firm level are largely understudied. In this research, I leverage the large-scale SEZ policy implemented by the Russian government in 2005 that aims to attract foreign investors to specific parts of the country by offering tax relief. The primary objective of this research is to quantify the effects of the Russian SEZ policy on local firms. To examine the effects, I use the generalised Difference-in-Difference methodology and apply it to a panel of firms in Russia for the 2006–2019 period. The data includes time-varying SEZ treatment on firms, firm characteristics, and accounting data. The primary outcome variables of interest are revenues, profits, and total factor productivity. The research findings could contribute to the urban economic literature on place-based policies and may be helpful to policymakers in determining the effectiveness of SEZ place-based policies.

Full text

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=cpce20
Post-Communist Economies
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/cpce20
Impact of Special Economic Zones on domestic
market: Evidence from Russia
Evgeniya Dubinina
To cite this article: Evgeniya Dubinina (2022): Impact of Special Economic Zones
on domestic market: Evidence from Russia, Post-Communist Economies, DOI:
10.1080/14631377.2022.2138154
To link to this article: https://doi.org/10.1080/14631377.2022.2138154
© 2022 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group.
Published online: 08 Nov 2022.
Submit your article to this journal
View related articles
View Crossmark data

Impact of Special Economic Zones on domestic market:
Evidence from Russia
Evgeniya Dubinina
Faculty of Social Sciences, Charles University, Institute of Economic Studies, Prague, Czech Republic
ABSTRACT
Place-based policies can be an eective instrument for govern-
ments to encourage the economic development of a country.
A Special Economic Zone (SEZ) is a place-based policy aimed at
attracting FDI, employment growth, and supporting new economic
reforms. In addition, for emerging economies an SEZ could be
a potential catalyst for development; foreign investors can have
a drastic impact on the productivity of domestic rms, revenues,
and market shares through the implementation of new technolo-
gies and the creation of new rms. However, the eects of SEZs on
the domestic market at the rm level are largely understudied. In
this research, I leverage the large-scale SEZ policy implemented by
the Russian government in 2005 that aims to attract foreign inves-
tors to specic parts of the country by oering tax relief. The
primary objective of this research is to quantify the eects of the
Russian SEZ policy on local rms. To examine the eects, I use the
generalised Dierence-in-Dierence methodology and apply it to
a panel of rms in Russia for the 2006–2019 period. The data
includes time-varying SEZ treatment on rms, rm characteristics,
and accounting data. The primary outcome variables of interest are
revenues, prots, and total factor productivity. The research nd-
ings could contribute to the urban economic literature on place-
based policies and may be helpful to policymakers in determining
the eectiveness of SEZ place-based policies.
ARTICLE HISTORY
Received 1 April 2022
Accepted 12 October 2022
KEYWORDS
Special Economic Zones;
total factor productivity
changes; place-based
policies; FDI
1. Introduction
Governments usually attract FDI by implementing tax deductions for foreign investors
and stimulating policies on a federal or regional level. The SEZs were established in India,
China, Sub-Saharan African countries, Poland, and other countries. SEZ policy demands
substantial government expenditures on infrastructure, developing institutions, and pro-
viding privileges for investors. Thus, the eectiveness of SEZ policy is vital for the whole
economy (Cizkowicz et al., 2015). The SEZs can have a drastic impact on the productivity
of the domestic market, revenues, and market shares of rms through the implementa-
tion of new technologies and the creation of new rms. Foreign rms, on average, are
more productive than domestic rms (Ebenstein, 2012). Thus, their presence induces
productivity growth in the domestic market. On the one hand, domestic rms could
CONTACT Evgeniya Dubinina evgeniya.dubinina@fsv.cuni.cz Institute of Economic Studies, Faculty of Social
Sciences, Charles University, 110 01, Smetanovo nábřeží 6, Prague, Czech Republic
POST-COMMUNIST ECONOMIES
https://doi.org/10.1080/14631377.2022.2138154
© 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License
(http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any med-
ium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

benet from research and development (R&D) spillovers from foreign rms and the
growth of productivity. On the other hand, the revenues of domestic rms after policy
implementation could decrease because of higher market competition and lower market
share. Therefore, the weakest rms could be forced to leave the market. Nevertheless, the
strongest rms could survive, have growth productivity, and prot growth.
The current literature shows mixed eects from SEZs as one of the instruments for FDI
attraction on the productivity of the domestic market: Abraham et al. (2010), ITO et al.
(2010), Wang (2013), Alder et al. (2016), Lu et al. (2019), Zhaoying (2021) show a positive
spillover eect on productivity for domestic market, Lin et al. (2009) nd no eect, and HU
and Jeerson (2002) nd a negative eect. However, the sign of the eect could depend
on the quality of institutions, infrastructure, government monitoring, and evaluating
instruments of SEZ policy (Brautigam et al., 2010). Moreover, the initial number of SEZs
could impact SEZ performance. According to Zeng (2016), Sab-Saharan African countries
with 10 or 20 zones, in the beginning, received worse eects for the domestic economy in
comparison with SEZs in China that had 4 zones at the beginning.
In 2005, the Russian government started creating SEZs to attract foreign investors with
special privileges: free customs zone regime, prot tax rate deduction, and exemption
from land and transport taxes from 5 to 15 years.
1
By 2020, twenty-eight SEZs had been
created in eighteen Russian regions with dierent specialisations: industrial (15 regions),
technological (6 regions), touristic and recreational (5 regions), and logistic (1 region).
SEZs are located mostly in the central and western parts of Russia.
The eects of SEZs on the domestic market in Russia are largely understudied.
A thorough review of the literature reveals that only one paper studies the eects of
SEZ policy in Russia: Frick et al. (2018) estimate the eect of dierent factors on the
economic growth of an individual SEZ in 22 emerging countries, including Russia. The
sample consists of 346 zones for 2007 till 2012 period. The authors use the nightlights as
a proxy for economic growth outcome because of limited data availability. From the data
analysis, the authors nd that the average growth of SEZs in Russia is faster than the
national growth for the years 2007–2012. However, the SEZ area grows slower than the
surrounding area in Russia. However, to my knowledge, the eects of SEZ policy on
productivity and revenues are not studied yet on rm-level data.
In my research, I estimate the direct eects of SEZs on the revenues and factor
productivity of domestic rms with the generalised Dierence-in-Dierence (DID) tech-
nique, which could shed more light on the domestic market changes after SEZ policy
implementation. I use the panel data from 2006 to 2019, collected from the publicly
available sources, the Federal State Statistics Service of Russia, the Federal Tax Service of
Russia, and the Russian Special Economic Zones website, and from a commercial source,
Spark Interfax. The data includes time-varying SEZ treatment on rms, rms’ character-
istics, and accounting data. Dierent years of SEZ policy participation for rms give an
exogenous variation for causal inference. Specically, in the same region and industry
exist rms that are residents of SEZ (the area inside the region) and non-residents after
SEZ policy implementation. The primary outcome variables of interest are revenues and
total factor productivity. In the sample, I have 12 regions aected by SEZ policy and 177
business activities (each of the activity belong to the particular code of Russian Classier
of Types of Economic Activity) treated by SEZ policy. The research ndings could
2E. DUBININA

contribute to the urban economic literature on place-based policies and can be helpful to
policymakers in Russia and other countries looking to create or change SEZ policy.
The research ndings of my paper show signicant and positive eect of SEZ policy
implementation in Russia on rm productivity and rm revenues in the domestic market.
The results are robust to including one and three leads and lags, control variables
(logarithm of capital and logarithm of labour), and xed eects for rms, regions, business
activities, and time. The parallel trend assumption holds for one lead and lag specication
of the model and three lead and lag specication of the model.
2. Review of relevant literature and background
2.1. Theoretical background of SEZs
SEZs are areas where a government chooses to have dierent rules from the rest of the
country in administrative, regulatory, and scal regimes (Farole, 2011). SEZs could operate
in various forms: export or economic processing zones, free zones, or foreign trade zones
(AkincI & Farole, 2011; Farole, 2011). The type of special economic zones depends on the
development objective, typical size, typical location, activities (sectors), markets (domes-
tic, re-export,export) (AkincI & Farole, 2011). For example, Bangladesh, Vietnam, Mauritius,
Mexico are traditional export-processing zones with size less than 100 hectares. In com-
parison, Russian SEZs cover move than 1000 hectares and do not have market restrictions
(only export or domestic).
SEZs could incentivise economic development as it is based on innovations, technol-
ogies, and creating an attractive environment for businesses (Baciuliene et al., 2021). The
SEZ policy implementation could boost regional development by attracting FDI, new
technologies, and skilled labour.
SEZs policy generates spatial concentration of rms and could yield technological and
knowledge spillovers, productivity growth from agglomeration economies associated
with moving goods, labour, and ideas (Jacobs, 1969; Krugman & Obstfeld, 2003;
Marshall, 1920; Newman & Page, 2017). The spillover eects in SEZs could be classied
into intra-industry (specialisation) or complementary, depending on the type of SEZs
(single or multi-industry), known in the literature as Marshallian and Jacobian externalities
(Jacobs, 1969; Marshall, 1920). Russian SEZs represent specialised clusters that could face
Marshallian externalities. According to (Otsuka & Sonobe, 2006), an industrial cluster
attracts traders, parts-suppliers, skilled workers, and engineers. With time inside the
industrial cluster competition pushes rms to start contracts with suppliers inside the
cluster, to trigger innovations, new marketing channels, to increase the quality of the
product, and to absorb unsuccessful enterprises (Otsuka & Sonobe, 2006). Moreover,
competitive rms could collaborate to overcome market constraints (Newman & Page,
2017; Schmitz, 1995). As a result, the production eciency improves in the industry
(Otsuka & Sonobe, 2006).
2.2. SEZ eects in current research
The creation of SEZs is a popular policy instrument for attracting FDI implemented in
countries including China, Poland,
2
India,
3
and other countries. The main objectives of
POST-COMMUNIST ECONOMIES 3

establishing SEZs are attracting FDI, unemployment decrease with creating new job
places from foreign employers, supporting new economic reforms strategy, and acting
as experimental laboratories for the application of new policies and approaches (Farole,
2011; Pastusiak et al., 2018; Zeng, 2016). Also, SEZs is a potential catalyst for develop-
ment, especially for emerging economies (Alder et al., 2016; Grant, 2020). Dorozynski
et al. (2018) show the overall growth of the economy after creating SEZs and emphasise
the major role of investors in SEZs (72% of new jobs, 50% of investment projects, 81% of
investment stock), using the example of Poland. Zeng (2010) estimates the eect of
SEZs in China on the whole economy and nds growth in new job places (30 million, the
relative growth is unknown), national GDP (22%), FDI (46%), and exports (60%). In
Bangladesh, SEZs attract $2.6 billion (relative growth is unknown) in foreign investment
and create 350,000 new jobs (Zeng, 2016). By 2015, 14 SEZs operated in Poland and
attracted 33% of total FDI inows in 4 years. Most African SEZs, except Mauritius and the
partial initial success of Kenya, Madagascar, and Lesotho, failed to attract signicant
investment, promote exports, and create sustainable employment relative to other
countries (Brautigam et al., 2010; Farole & Moberg, 2014) because of the lack of basic
infrastructure and SEZ regulation, insucient strategic planning, poor choice of location
and poor internal coordination (Brautigam et al., 2010). For the Dominican Republic,
SEZs allowed the economy to move from agricultural reliance to manufacturing, creat-
ing more than 100,000 manufacturing jobs (Akinci & Farole, 2011). Industrialization and
job creation after SEZ policy implementation were in Mauritius, the Republic of Korea,
Taiwan, China, Honduras, El Salvador, Madagascar, Bangladesh, and Vietnam. The
Klaipeda free economic zone (FEZ) in Lithuania for 15 years (2006–2020) attracted
634 million euro (that corresponds to 60% of all FDI in Klaipeda city and almost 45%
in Klaipeda county) and employed 3427 employees for 14 years (2006–2019) (Baciuliene
et al., 2021).
In the recent literature, the authors also discuss the productivity eects on the
domestic market from foreign rms after SEZ policy implementation. The productivity
eects arise from the learning behaviour of domestic rms from foreign, when rms
outside SEZs become more ecient after foreign investment (Zeng, 2016). According to
ZENG (2016), positive productive spillover eects could be in the form of new technology
transfer or innovation, growth of economic productivity, increase in economic diversica-
tion, or welfare eects for the domestic population.
However, empirical evidence of spillover eects from foreign technology on the
domestic market with SEZ policy implementation is controversial. Abraham et al. (2010)
nd positive spillovers from foreign rms in their studying of the inuence of FDI on the
performance of domestic rms and the impact of SEZs on Chinese rms. Additionally, ITO
et al. (2010) report positive spillover eects on total factor productivity (TFP) of domestic
rms within and across industries. Greenstone et al. (2010) show productivity growth after
SEZ policy implementation in the United States but only in the same county of SEZ
location. Wang (2013) compares the changes among municipalities that created SEZs,
and the impact of SEZs on the whole economy and on the productivity of domestic rms
in China using the OLS. The author nds that SEZs increase the TFP of local rms and the
earnings of local workers. Kline and Moretti (2013) nd a mixed eect from the policy:
gains from the policy are accompanied by losses in other parts of the country from the
policy. In contrast, Lin et al. (2009) do not nd any spillover eects on productivity from
4E. DUBININA

foreign rms. And HU and Jeerson (2002) show negative spillover eects of R&D and
technology transfer from foreign rms.
The sign of spillover eects from foreign rms could depend on the structure of the
market, on time or the conjecture of the market (Zeng, 2016). According to Zeng (2016),
strong government support, institutions, a strategic location, commercial viability, and
willingness to address environmental concerns in SEZs could make SEZ policy experience
successful. In Singapore, the Republic of Korea, and Malaysia, the requirements were
satised before policy implementation, thus, positively aect the domestic rms in
productivity (Zeng, 2016). Moreover, for successful policy implementation and domestic
production growth, a country should have the channels of technology and R&D transfers,
e.g. business incubators, conferences, innovation platforms (online or oine), and talent-
recruiting programs (Zeng, 2016).
2.3. Factors aecting SEZ eects
SEZ policy is an eective instrument for FDI attraction and productivity growth for the
economy. However, SEZ policy eects could dier for low- and middle-income econo-
mies. The low-income economies could have a poor level of institutions, infrastructure,
government monitoring, and evaluating instruments. Problematic legal, regulatory and
institutional frameworks, poor business environment, strategic planning, and adoption of
demand-driven approaches to business should be improved to increase the probability of
SEZ policy implementation success (Zeng, 2016).
The initial number of zones could impact SEZ eect on the domestic market. China
started from four zones in dierent locations and some Sab-Saharan African countries had
10 or 20 zones at the beginning (Zeng, 2016). In Russia, on the rst step of SEZ policy
implementation 5 zones were created in 2005, then from 2006 till 2010 were created 6,
and during 2011–2020 were created 17 zones. The gradual policy implementation could
give additional time for the government to monitor and evaluate the eects of the policy,
and correct the policy if needed.
Frick et al. (2018) estimate the eect of dierent factors on the economic growth of an
individual SEZ in 22 emerging countries
4
across 346 zones from 2007 till 2012, including
Russia. The authors use the nightlights as a proxy for economic growth outcome from the
Defense Meteorological Satellite Program for the years 1992–2012 because of limited data
availability. From the data analysis, the authors nd that the average growth of SEZs in
Russia is faster than the national growth for the years 2007–2012. However, the SEZ area
grows slower than the surrounding area in Russia. The results of estimation using OLSs
indicate a positive signicant eect of SEZ size, distance to largest cities from SEZs,
proximity to large markets, percent of SEZ industry in GDP of the country on SEZ growth.
The authors nd negative signicant eects of SEZ years of operation, foreign ownership
requirement on SEZ growth. And no signicant eect of subsidised utilities in SEZs,
national one-stop-shop, independence of zone regulator, and rule of law in SEZs on SEZ
growth.
Studies of SEZs usually consider the direct and spillover eects of SEZs policy on
domestic rms and domestic market of China, India, and Poland using dierent
estimation methods and data (Abraham et al., 2010; Cieslik & Ryan, 2005; Cizkowicz
et al., 2015; Ebenstein, 2012; Hyun & Ravi, 2018; Nazarczuk & Uminski, 2018; Pastusiak
POST-COMMUNIST ECONOMIES 5

et al., 2018; Y, 2019; Zhaoying, 2021). Various studies consider the overall economic
development eects of SEZ policy implementation or dierent types of SEZs
(Auruskeviciene et al., 2007; Baciuliene et al., 2021) instead of direct and spillover
eects on domestic rms and market. However, the research ndings could not be
implemented to SEZ policy eects in Russia, because the environment is dierent.
First, SEZs in Russia exist since 2005 and in China from 1978, in India from 2000. In
China, the government could adjust the performance and eects from SEZs after the
longtime experience. In India, SEZs arose from EPZs that had industrial clusters before
SEZ policy implementation. In Russia, SEZ policy was implemented without industrial
clusters, the zones were established from the scratch. Second, the research ndings for
productivity growth are mixed, depending on estimation strategy, the data, the infra-
structure, institutions of countries, and the whole environment and business climate.
Thus, Russian SEZs should be considered as separate case of SEZ eects. Also, to my
knowledge, the eects of SEZs on the domestic market in Russia have not yet been
studied. Only Frick et al. (2018) study the SEZ growth including Russia and nd that
the average growth of SEZs in Russia is faster than the national growth for the years
2007–2012, but the SEZ area grows slower than the surrounding area in Russia. In my
research, I study the eects of creating SEZs on the revenue and productivity of the
domestic market in Russia using the generalised Dierence-in-Dierence (DID) techni-
que with panel data from 2006 to 2019.
3. Data
I use the panel data from 2006 to 2019, collected from the Federal State Statistics Service
of Russia (FSSS), the Federal Tax Service of Russia (FTS), Russian Special Economic Zones
website, and Spark Interfax. From the FSSS and FTS, I scrapped rms’ individual tax
numbers (INN) to collect rms’ characteristics and accounting data from Spark Interfax.
With the Russian SEZ website,
5
I obtained all information about SEZ participants, years
of entering and exiting the SEZs. The data include time-varying SEZ treatment on rms,
rms’ characteristics (the region, the business activity), and accounting data. The pri-
mary outcome variables of interest are revenues and total factor productivity. The
treatment group consist of rms that are located in SEZs and the control group consist
of rms from the same industries as the treated rms but located outside SEZs in the
same region and industry.
After collecting all the data for 2001–2019, it turns out that the data is highly
unbalanced with many missing values. There were almost no data for the period
2001–2005 in the database; thus, this period was excluded from the studying. Also,
I exclude Moscow (compared to region) as rms are dierent from this region
compared to other regions. Moreover, Moscow has two SEZs with similar business
activities of rms that could generate synergetic spillover eects in comparison with
other regions. Additionally, I exclude regions with few rms or without rms (in these
regions SEZs established recently and not enough time has passed for rms to
appear in these SEZs) in the SEZs and rms with outcome variable data availability
of fewer than 8 years. In the sample, I have 12 regions,
6
177 business activities (each
of the activity belongs to the particular code of Russian Classier of Types of
Economic Activity).
6E. DUBININA

4. Empirical strategy
For the empirical strategy, I follow the approach of Granger (1969), Angrist and Pischke
(2009) for estimating the SEZ policy eects, when treatment starts at a dierent time for
rms. I use a generalised Dierence-in-Dierence approach including lags (time before
treatment) and leads (time after treatment) in the model:
Yit ¼αiþλtþvjþγrþX
m
τ¼0
δτDi;tτþX
q
τ¼0
δþτDi;tþτþXit0βþ 2it
Where: Yit is an outcome variable (logarithm of earnings); αi is rm xed eects; λt is
time-specic xed eects; vj is business activity xed eects; γr is region-specic xed
eect; Di;t is a dummy variable for treatment (if rm enters the SEZ it will be unity and zero
otherwise), at τ¼0 switches to treatment; δ0is the contemporaneous treatment eect;
δ1 is the medium-term eect; δþτ if treatment in the future aects Y, now-reverse
causality from the future to the past. I include xed eects to take into account non-
time varying dierences in infrastructure and institutions among regions, non-time vary-
ing dierences in internal business processes among rms, also non-time varying dier-
ences in the structure of business activities and time-varying dierences to
include
other
characteristics.
Variation in the year and industry of SEZ policy implementation yields exogenous
variation for identifying a causal eect. Specically, in the same region and industry exist
rms that are residents of SEZ (the area inside the region) and non-residents after SEZ
policy implementation. However, a rm could decide to be a resident of SEZ not in
the year of SEZ establishment but later. The residents of SEZ in Russia could be foreign
investors or domestic.
In the next section, I discuss the estimation results of the linear regression model
including leads and lags, xed eects, and the controls. In all the specications, I cluster
standard errors on the level of multiplication region and business activity because of the
studying design that treatment is also clustered on region and activity level. I include two
controls in studying: labour and capital of rms because other possible controls have
many missing values. Initially in the database, the labor variable was an interval for labor
force, e.g. 0–5 meaning that the rm has from 0 to 5 workers. For a higher number of
workers, the interval could be from 100 to 150. However, I do not have information about
the precise number of workers in the rm, that is why I smooth the dierences of the
intervals with calculating the new variable of labor:
Laborit ¼log Minit þlog Maxit
2
Where Minit is a minimum of the labour interval and Maxit is a maximum of the interval.
For capital, I also use the logarithmic form to be consistent with the linear form of the
Cobb-Douglas production function:
F K;Lð Þ ¼ AKαL1α
Where: A is technology parameter; K stays for capital; L is labour and α is the parameter
of the return to scale. The initial value of capital is presented in rubles and stays for ‘equity
POST-COMMUNIST ECONOMIES 7

capital’ in the rms’ accounting forms. In the logarithmic form Cobb-Douglas production
function has the formula:
log F K;Lð Þ ¼ log Aþαlog Kþ1αð Þlog L
To maintain the causal relation of SEZ treatment on rm earnings, I use the instrumental
variable approach. I construct the instrument from the interaction of two variables: the SEZ
industry share in the overall gross regional product in 2000 and the growth rate of the same
industry on the country level. The multiplication of the instrument components gives the
time-varying instrument variable. The data for the instrument variable was collected also from
the Federal State Statistics Service of Russia. The SEZ location should be connected with
agglomeration clusters or dominant industries in the regions as in the cases of SEZs in China
and India. Also, the SEZ location should be connected with regions where the particular
industry dominates on the country level and the growth rate of the particular industry.
However, the instrument variable should not be connected directly with the outcome variable
because rms have not got the direct impact of the industries on a country level or regional
basis on the earnings. Thus, the relevance and exclusion restriction should hold. Moreover,
instrument variable and outcome variable describe dierent periods: the instrument variable
is based on the industry share of gross regional product in 2000 and outcome variable is based
on 2006–2019 data. The time-varying part of the instrument connected with growth rate of
the industries on the country level and outcome variable is for rm-level data. As
a consequence, the exclusion restriction should hold.
5. Results
Table 1 presents the results of estimating the multi-way xed eects model (without lags
and leads). The outcome variable for all regressions, (1) - (6), is the logarithm of rm
earnings. The explanatory variable for the regressions (1) - (3) is a dummy variable for SEZ
treatment. The regressions (1) - (3) dier in the xed eects ((1) is without xed eects; (2)
is with rm, year, and region xed eects; (3) is with all xed eects). The explanatory
variable for the regressions (4) – (6) is the logarithm of labour and the logarithm of capital.
The regressions (4) - (6) also dier in the xed eects ((1) is without xed eects; (2) is with
rm, year, and region xed eects; (3) is with all xed eects). The regressions with xed
eects absorb eects that are not changing over time for rms, business activities and
regions, as well as eects that change over time (time xed eects).
The treatment coecient indicates a positive signicant eect of creating SEZs on
rm revenues in the baseline specication (1) and the specication with xed eects
((2), (3)). Moreover, the SEZ treatment coecient stays positive and signicant in the
specications with control variables ((7) – (9)). However, including the interaction
variables of SEZ and logarithm of labour and SEZ and logarithm of capital decrease
the signicance of the coecient but leave it positive. The labor coecient is signicant
and negative in all the specications where it is included ((4) – (12)). The sign of the
coecient could be connected with the form of marginal earnings for one unit of labor,
where a rm could have enough labor and an additional unit of labor will decrease the
earnings. Moreover, according to (Dced Report, 2021), growth of productivity could lead
to a decrease in employment in the short run, as new technologies reduce labor
requirement. The logarithm of capital is positive and signicant in all specications
8E. DUBININA

where it is included ((4) – (12)). The sign of the capital coecient is expected because
under constant return to scale, more input should lead to more output and more
earnings. Table 1 Panel C presents the multi-way xed eects model with labor factor
productivity as an outcome variable, calculated as an earnings-to-capital ratio. The
regressions (13) - (15) also dier in the xed eects. The SEZ coecient is signicant
and positive. Thus, the SEZ treatment leads to a growth of earnings per one unit of
labor. However, the result contradicts previous results in Table 1, Panel A and B for the
relation of labor productivity. The possible explanation of the result is the lack of control
variables in this specication.
The results of positive TFP growth after SEZ policy implementation are shown in the
papers by Abraham et al. (2010), Ito et al. (2010), Greenstone et al. (2010) in dierent
countries which established SEZs. Thus, my research ndings are consistent with previous
studies of SEZs in other countries. Zeng (2016) mentions that the eects of the SEZ policy
could depend on the time or conjecture of the market. A positive sign of SEZ policy
Table 1. Multi-way fixed effects model.
PANEL A (1) (2) (3) (4) (5) (6)
SEZ 2.841***
(0.273)
0.632***
(0.142)
0.632***
(0.142)
Logarithm of labor −0.301***
(0.017)
−0.028***
(0.006)
−0.028***
(0.006)
Logarithm of capital 0.661***
(0.015)
0.344***
(0.012)
0.344***
(0.012)
constant 15.473***
(0.050)
15.487***
(0.001)
15.487***
(0.001)
6.697***
(0.374)
10.167***
(0.220)
10.167***
(0.220)
Number of observations 101,122 101,100 101,100 77,949 77,758 77,758
PANEL B (7) (8) (9) (10) (11) (12)
SEZ 1.795***
(0.151)
0.462**
(0.170)
0.462**
(0.170)
1.069
(1.362)
2.737*
(1.174)
2.737*
(1.176)
Logarithm of labor −0.296***
(0.017)
−0.028***
(0.006)
−0.028***
(0.006)
−0.300***
(0.017)
−0.030***
(0.006)
−0.030***
(0.006)
Logarithm of capital 0.660***
(0.015)
0.344***
(0.012)
0.344***
(0.012)
0.660***
(0.015)
0.347***
(0.012)
0.347***
(0.012)
SEZ x Logarithm of labor 0.297***
(0.044)
0.143***
(0.031)
0.143***
(0.031)
SEZ x Logarithm of capital −0.059
(0.071)
−0.177**
(0.063)
−0.177**
(0.063)
constant 6.667***
(0.375)
10.163***
(0.220)
10.163***
(0.220)
6.686***
(0.378)
10.135***
(0.219)
10.135***
(0.219)
Number of observations 77,949 77,758 77,758 77,949 77,758 77,758
PANEL С (13) (14) (15)
SEZ 3.298***
(0.271)
0.806**
(0.250)
0.806**
(0.251)
constant 9.599***
(0.050)
9.615***
(0.002)
9.615***
(0.002)
Number of observations 101,122 101,100 101,100
Lags No No No No No No
Leads No No No No No No
Firm fixed effects No Yes Yes No Yes Yes
Year fixed effects No Yes Yes No Yes Yes
Region fixed effects No Yes Yes No Yes Yes
Business activity fixed effects No No Yes No No Yes
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings ((1) – (12)); logarithm of
earnings-to-capital ratio ((13) – (15)).
POST-COMMUNIST ECONOMIES 9

implementation in relation to the earnings of rms and productivity could indicate the
favourable conjecture and environment for SEZ policy implementation, e.g. a strategic
SEZ location, willingness to address environmental concerns in SEZs and government
support.
Table 2 presents estimation results including one lead and one lag and Figure 1
illustrates graphically these estimation results. Figure 1 has condence intervals and
coecients from Table 2 with one lag and one lead, control variables and xed eects.
The vertical bands represent ±1.96 times the standard error of each point estimate. On
horizontal line with zero starts the SEZ treatment. On the left side, before the treatment,
we see that the coecient is indistinguishable from zero that is why the pre-trends
assumption holds. The coecient for SEZ dummy variable is robust because it is still
signicant and positive in all specications ((1) – (6) of Table 2). The labour coecient is
negative and signicant in the specications where the control is included ((2) - (3), (5) –
Table 2. Multi-way fixed effects model with one lag and lead.
(1) (2) (3) (4) (5) (6)
SEZ 0.469***
(0.120)
0.452**
(0.141)
0.603***
(0.162)
0.469***
(0.121)
0.452**
(0.142)
0.603***
(0.162)
Logarithm of labor −0.053***
(0.008)
−0.032***
(0.007)
−0.053***
(0.008)
−0.032***
(0.007)
Logarithm of capital 0.347***
(0.012)
0.347***
(0.012)
Lag Yes Yes Yes Yes Yes Yes
Lead Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Region fixed effects Yes Yes Yes Yes Yes Yes
Business activity fixed effects No No No Yes Yes Yes
constant 15.451***
(0.001)
16.178***
(0.056)
10.057***
(0.220)
15.451***
(0.001)
16.178***
(0.056)
10.057***
(0.221)
Number of observations 88,347 70,039 68,365 88,347 70,039 68,365
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings.
Figure 1. Time passage relative to the year of the SEZ treatment. Notes: on the graph grey points are
the point estimates; lines below and above point estimates are 95% confidence interval.
10 E. DUBININA

(6)). As was mentioned before, the sign of the coecient could be connected with the
form of marginal earnings for one unit of labor, where a rm could have enough labor and
an additional unit of labor will decrease the earnings. And coecient for the capital is
signicant and positive in all the specications of the model where the coecient is
included ((3), (6)), thus, the coecient is robust.
Granger (1969) and Angrist and Pischke (2009) suggest including at least three periods in
lags and leads in the standard generalised DID. Thus, Table 3 presents the same specications
as in Table 2 but including three leads and lags and Figure 2 presents graphically the results of
Table 3. The results are robust to including 3 leads and lags: SEZ coecient remains signicant
and positive in all specications ((1) – (6)) of Table 3, the capital coecient remains signicant
and positive in all specications where the variable is included ((3), (6)). However, the labour
Table 3. Multi-way fixed effects model with three lags and lead.
(1) (2) (3) (4) (5) (6)
SEZ 0.775***
(0.234)
0.745***
(0.218)
0.906***
(0.252)
0.775***
(0.234)
0.745***
(0.219)
0.906***
(0.253)
Logarithm of labour −0.027***
(0.007)
−0.001
(0.009)
−0.027***
(0.007)
−0.001
(0.009)
Logarithm of capital 0.490***
(0.015)
0.490***
(0.015)
Lags Yes Yes Yes Yes Yes Yes
Leads Yes Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Region fixed effects Yes Yes Yes Yes Yes Yes
Business activity fixed effects No No No Yes Yes Yes
constant 15.251***
(0.003)
15.793***
(0.057)
7.070***
(0.282)
15.251***
(0.003)
15.793***
(0.057)
7.070***
(0.282)
Number of observations 59,826 45,087 44,662 59,826 45,087 44,662
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings.
Figure 2. Time passage relative to the year of the SEZ treatment. Notes: on the graph grey points are
the point estimates; lines below and above point estimates are 95% confidence interval.
POST-COMMUNIST ECONOMIES 11

coecient lose signicance in the specications where all the control variables are included
((3), (6)). Thus, the robustness of the coecient is not stable and further research and data
should be done to detect the robust sign of the coecient. On the left side of the graph from
the treatment time zero in Figure 2, the coecients are indistinguishable from zero that is why
the pre-trends assumption holds.
As a robustness check, Tables A2 and Table A3. in the APPENDIX show the results of
Cerulli (2010) method of DID in a dynamic treatment setting. Table A2 includes one lag
and Table A3 includes 2 lags. These results measure the SEZ treatment eect for rms with
a dierent start time of SEZ policy implementation or without treatment.
7
For the model
with one lag (Table A2), the results for the rst treatment group (SEZ coecient) are
robust because the coecient is positive and signicant and for the second treatment
group (L1.SEZ – SEZ) are not robust; for the control group, the coecient (L1.SEZ) is not
signicant and robust. For the model with two lags (Table A3), the results for the several
treatment groups (SEZ; L2.SEZ – L1.SEZ + SEZ) are robust; for several other treatment
groups, coecients (L1.SEZ; L1.SEZ – SEZ; L2.SEZ – L1.SEZ; L2.SEZ – SEZ) are not robust; for
the control group, the coecient (L2.SEZ) is not signicant and robust.
To maintain the causal relation of SEZ treatment on rm earnings, I use the instru-
mental variable approach. The results of reduced form and rst stage estimation are
presented in Table A1 of the APPENDIX. The results indicate that the decisions for SEZ
location are not connected with dominant industries and their growth on region or
country level. However, pre-existing industrialisation increase SEZ performance (Frick
et al., 2018). Nevertheless, according to Frick et al. (2018), the SEZ location decisions
could be done on the basis of dierent factors: proximity to the ports, large markets, or
the main developed markets of the world, the access to the largest cities in the country, to
the ports. Most of the SEZs in Russia are located near the biggest cities of the country,
Moscow and Saint-Petersburg, and the main developed markets, in Europe and Asia.
SEZs in the south of the country are located near Lake Baikal and SEZs in the south-western
part are located near the Black Sea and the Caspian Sea. Industrial zones are located mainly in
the European part of the country. The logistic zone in Ulyanovsk Oblast, which was established
in 2009, located near Volga River which is the longest river in Europe. Thus, instrument
variables could be constructed based on the location with the new data.
6. Discussion
China implemented SEZ policy in 1978, African countries in the late 1990s, Poland in 1995,
and India in 2000. In comparison with other countries (China, India, Poland, African
countries, and other countries), for Russia SEZ policy is a recent reform and there is no
previous experience similar to SEZs, e.g. in India EPZs were replaced by SEZs because of
poor performance and low attraction of investors. A thorough literature review reveals
mixed results of SEZ policy eects on productivity growth: positive (Abraham et al., 2010;
Alder et al., 2016; Ito et al., 2010; Lu et al., 2019; Wang, 2013; Zhaoying, 2021) negative (Lin
et al., 2009), and no eect (Lin et al., 2009). The SEZ policy eect on productivity growth
and in general SEZ growth could depend on the quality of institutions and infrastructure,
incentive packages for investors, proximity to the ports, large markets, or the main
developed markets of the world, the access to the largest cities in the country, to the
ports, the business climate of the country, government monitoring and evaluating
12 E. DUBININA

instruments of SEZ policy, and national one-stop-shops services (Brautigam et al., 2010;
Frick et al., 2018; Zeng, 2016). Moreover, the initial number of SEZs could impact SEZ
performance. According to Zeng (2016), Sab-Saharan African countries with 10 or 20
zones, in the beginning, received worse eects for the domestic economy in comparison
with SEZs in China that had 4 zones at the beginning. Also, productivity growth after SEZ
policy implementation could depend on the dominant eect: the rst eect arises from
higher productivity of foreign rms on the domestic market, when domestic rms could
benet from R&D spillovers from foreign rms and productivity growth; the second eect
arises from higher competition after entry of foreign rms and revenues of domestic rms
decrease.
The research ndings of estimating SEZ eects on productivity growth in China, India,
Poland could not be implemented to the Russian case because of several reasons:
dierent types of SEZs (e.g. with dierent SEZ ownership); dierent outcome variables
(FDI, export, wage); the other conjecture of the market or SEZs (e.g. in India Export
Processing Zones were replaced by SEZs because of poor performance and in Poland
SEZs were established in regions with serious economic problems); SEZ experience is
longer (e.g. in China SEZs were established in 1978); dierent initiative parties of SEZ
policy implementation (e.g. in India could be initiated by the general public in comparison
with Russia, where government establish SEZs); the privileges for investors are dierent
(e.g. in India tax exemption for residents lasts 5 years and in Russia from 5 to 15 years,
depending on particular SEZ and business activity).
Further analysis of the SEZ eects on the domestic market in Russia could consider the
spatial proximity of rms to SEZs because the SEZ neighbouring regions could benet more
from technological spillovers in the same industry. Moreover, the productivity changes after
SEZ policy implementation could be studied in formal and informal sectors separately in
Russia, because the marginal rm in the informal sector is smaller than in the formal sector
(Paula & Scheinkman, 2010). Additionally, the quality of institutions, infrastructure, business
climate for the potential investors in SEZs have not been studied yet. Thus, the government
could monitor these factors to increase SEZ performance and productivity growth. However,
the research could be done with more data on Russian rms.
7. Conclusion
The Russian government started to create Special Economic Zones (SEZs) to attract
foreign investors with tax privileges in 2005. Foreign investors can have a signicant
impact on the productivity of domestic rms, revenues, and their market share through
the implementation of new technologies and the creation of new rms.
However, to my knowledge, the eects of SEZs on the domestic market in Russia are
largely understudied. In my research, I study the eects of creating SEZs on productivity
and revenue growth in Russia using the generalised Dierence-in-Dierence (DID) tech-
nique. I use the panel data from 2006 to 2019, collected from the Federal State Statistics
Service of Russia, the Federal Tax Service of Russia, the Russian Special Economic Zones
website, and Spark Interfax. The data includes time-varying SEZ treatment on rms, rms’
characteristics, and accounting data. The primary outcome variables of interest are
revenues and total factor productivity. In the sample, I have 12 regions aected by SEZ
policy and 177 business activities treated by SEZ policy. The variation in time and SEZs of
POST-COMMUNIST ECONOMIES 13

the policy implementation and rms’ decisions to enter SEZs give the possibility for
estimating the causal eects after SEZ policy implementation.
The results of the paper indicate a positive and signicant eect of SEZ policy on rm
revenues. All the coecients are robust to dierent specication including one or three
leads and lags. The labour coecient is signicant and negative that could be in the
situation when a rm could have enough labor and additional unit of labor will decrease
the earnings. The capital coecient is positive and signicant in all specications because
more input should lead to more output and more earnings. The paper shows positive eect
of SEZ policy implementation on rm revenues and capital productivity and the opposite
eect for labor. The research ndings could contribute to the urban economic literature on
place-based policies and can be helpful to policymakers in Russia and other countries.
Notes
1. Russian Special Economic Zones. http://eng.russez.ru/ (accessed December 1, 2020).
2. Polish Investment and Trade Agency. https://www.paih.gov.pl/why_poland/investment_
incentives/sez (accessed December 1, 2020).
3. Special Economic Zones in India. http://www.sezindia.nic.in/ (accessed December 1, 2020).
4. China, Philippines, Malaysia, South Korea, Thailand, Vietnam, Turkey, Russia, Ghana, Jordan,
Kenya, Lesotho, Nigeria, South Africa, Argentina, Chile, Colombia, Dominican Republic,
Honduras, Bangladesh, India, and Pakistan.
5. Russian Special Economic Zones. http://eng.russez.ru/ (accessed December 1, 2020).
6. The Republic of Buryatia, the Republic of Tatarstan, Altai Krai, the Irkutsk Region, the Lipetsk
Region, the Moscow Region, the Pskov Region, the Samara Region, the Sverdlovsk Region, the
Tomsk Region, the Ulyanovsk Region, Saint-Petersburg (comparable to region).
7. If I dene SEZ coecient as α1, SEZ coecient with one lag as α2, SEZ coecient with two lags
as α3, for the model with one lag (Table 5) the DID coecient would be α1 for rms that had
treatment in t1ð Þ but had no treatment in t2ð Þ and t; (α2α1) for rms that had
treatment in t but had no treatment in t1ð Þ and t2ð Þ; α2 for rms that had no treatment
in t;t1ð Þ, and t2ð Þ. For the model with two lags (Table 6) the DID coecient would be α1
for rms with treatment in t1ð Þ;t2ð Þand no treatment in t; α2 for rms with treatment in
t2ð Þ and no treatment in t, t1ð Þ; ðα3α2þα1Þ
for
rms with treatment in t1ð Þ and no
treatment in t, t2ð Þ; α3 for rms no treatment in t, t1ð Þ, t2ð Þ; ðα2α1Þfor rms with
treatment in t, t2ð Þ and no treatment in t1ð Þ; ðα3α2Þfor rms with treatment in t,
t1ð Þ and no treatment in t2ð Þ; ðα3α1Þfor rms with treatment in t and no treatment in
t1ð Þ, t2ð Þ.
Disclosure statement
No potential conict of interest was reported by the author(s).
Funding
The work was supported by the The Charles University Grant Agency [130122].
ORCID
Evgeniya Dubinina http://orcid.org/0000-0003-2779-9935
14 E. DUBININA

References
Abraham, F., Konings, J., & Slootmaekers, V. (2010). FDI spillovers in the Chinese manufacturing
sector. Economics of Transition and Institutional Change, 18(1), 143–182. https://doi.org/10.1111/j.
1468-0351.2009.00370.x
Akinci, G., & Farole, T. (2011). Special Economic Zones: Progress, emerging challenges, and future
directions. In Directions in Development; trade. World Bank, 1–19.
Alder, S., Shao, L., & Zilibotti, F. (2016). Economic reforms and industrial policy in a panel of Chinese
cities. Journal of Economic Growth, 21(4), 305–349. https://doi.org/10.1007/s10887-016-9131-x
Angrist, J., & Pischke, J. (2009). Mostly harmless econometrics: An empiricist’s companion. Mostly
Harmless Econometrics: An Empiricist’s Companion.
Auruškevičienė, V., Kuvykaitė, R., Šalčiuvuenė, L., & Žilys, L. (2007). Identication of key success
factors in free economic zone development in Lithuania. Economic and Management, 12,
277–284.
Baciuliene, V., Navickas, V., & Petroke, I. (2021). Impact of free economic zones on regional economic
development: The case of klaipeda free economic zone in Lithuania. International Journal of
Entrepreneurial Knowledge, 9(1), 97–111. https://doi.org/10.37335/ijek.v9i1.120
Brautigam, D., Farole, T., & Tang, X. (2010). China’s investment in African special economic zones:
Prospects, challenges, and opportunities. World Bank - Economic Premise, 5, 1–6. https://open
knowledge.worldbank.org/handle/10986/10202
Cerulli, G. (2010). Modelling and measuring the eect of public subsidies on business R&D: A critical
review of the econometric literature. The Economic Record, 86(274), 421–449. https://doi.org/10.
1111/j.1475-4932.2009.00615.x
Cieslik, A., & Ryan, M. (2005). Location determinants of Japanese multinationals in Poland: Do special
economic zones really matter for investment decisions? Journal of Economic Integration, 20(3),
475–496. https://doi.org/10.11130/jei.2005.20.3.475
Cizkowicz, P., Cizkowicz-Pekala, M., & Rzonca, A. (2015). The eects of special economic zones on
employment and investment: Spatial panel modelling perspective. NBP Working Papers 208.
DCED Report. (2021). Increased productivity creates economic growth – DCED. https://www.enter
prise-development.org/what-works-and-why/evidence-framework/increased-productivity-
creates-economic-growth/ (accessed June 19, 2021)
Dorozynski, T., Swierkocki, J., & Urbaniak, W. (2018). Determinants of investment attractiveness of
polish special economic zones. Entrepreneurial Business and Economics Review, 6, 161–180.
https://doi.org/10.15678/EBER.2018.060409
Ebenstein, A. (2012). Winners and losers of multinational rm entry into developing countries:
Evidence from the Special Economic Zones of the People’s Republic of China. Asian
Development Review, 29(1). https://doi.org/10.2139/ssrn.1962940
Farole, T. (2011). Special Economic Zones in Africa: Comparing performance and learning from
global experience. In Directions in development; trade. World Bank, 1–297 .
Farole, T., & Moberg, L. (2014). It worked in China, so why not in Africa?: The political economy
challenge of special economic zones. WIDER Working Paper Series. World Institute for
Development Economic Research (UNU-WIDER).
Frick, S., Rodriguez-pose, A., & Wong, M. (2018). Toward economically dynamic special economic
zones in emerging Countries. Economic Geography, 95(1), 1–35. https://doi.org/10.1080/
00130095.2018.1467732
Granger, C. (1969). Investigating causal relation by econometric and cross-sectional method.
Econometrica, 37(3), 424–438. https://doi.org/10.2307/1912791
Grant, M. (2020). Why special economic zones? Using trade policy to discriminate across importers.
The American Economic Review, 110(5), 1540–1571. https://doi.org/10.1257/aer.20180384
Greenstone, M., Hornbeck, M., & Moretti, E. (2010). Identifying agglomeration spillovers: Evidence
from winners and losers of large plant openings. The Journal of Political Economy, 118(3), 536–598.
https://doi.org/10.1086/653714
Hu, A., & Jeerson, G. (2002). FDI impact and spillover: Evidence from China’s electronic and textile
industries. The World Economy, 25(8), 1063–1076. https://doi.org/10.1111/1467-9701.00481
POST-COMMUNIST ECONOMIES 15

Hyun, Y., & Ravi, S. (2018). Place-based development: evidence from Special Economic Zones in
India. Boston University - Department of Economics - The Institute for Economic Development
Working Papers Series.
Ito, B., Yashiro, N., XU, Z., Chen, X., & Wakasugi, R. (2010). How do Chinese Industries benet from FDI
spillovers? RIETI Discussion Paper Series 10-E-026.
Jacobs, J. (1969). The economy of cities. Random House.
Jin, Y (2019). Special Economic Zones: A welfare analysis of labor market spillover eect Job Market
Paper.
Kline, P., & Moretti, E. (2013). Local economic development, agglomeration economies, and the big
push: 100 years of evidence from the Tennessee Valley authority. Working Paper. Working Paper
Series. National Bureau of Economic Research.
Krugman, P., & Obstfeld, M. (2003). International economics; theory and policy. Pearson Education.
Lin, P., Liu, Z., & Zhang, Y. (2009). Do Chinese domestic rms benet from FDI inow? Evidence of
horizontal and vertical spillovers. China Economic Review, 20, 677–691. https://doi.org/10.1016/j.
chieco.2009.05.010
Lu, T., Wang, J., & Zhu, L. (2019). Place-based policies, creation, and agglomeration economies:
Evidence from China’s economic zone program. American Economic Journal: Economic Policy, 11
(3), 325–360. https://doi.org/10.1257/pol.20160272
Marshall, A. (1920). Principles of Economics. Macmillan.
Nazarczuk, J., & Uminski, S. (2018). The impact of Special Economic Zones on export behaviour:
Evidence from polish rm-level data. E a M. Ekonomie a Management, 21(3), 4–22. https://doi.org/
10.15240/tul/001/2018-3-001
Newman, C., & Page, J. (2017). Industrial clusters: The case for special economic zones in Africa.
WIDER Working Paper, No. 2017/15. UNU-WIDER.
Otsuka, K., & Sonobe, T. (2006). Cluster-based industrial development: An East Asian model. Palgrave
MacMilan.
Pastusiak, R., Bolek, M., Jasiniak, M., & Keller, J. (2018). Eectiveness of special economic zones of
Poland. Zbornik Radova Ekonomskog Fakulteta U Rijeci: Časopis Za Ekonomsku Teoriju I Praksu/
Proceedings of Rijeka Faculty of Economics: Journal of Economics and Business, 36(1), 263–285.
https://doi.org/10.18045/zbefri.2018.1.263
Paula, Á., & Scheinkman, J. (2010). Value-added taxes, chain eects, and informality. American
Economic Journal: Macroeconomics: Macroeconomics, 2(4), 195–221. https://doi.org/10.1257/
mac.2.4.195
Schmitz, H. (1995). Collective eciency: Growth path for small-scale industry. The Journal of
Development Studies, 31(4), 529–566. https://doi.org/10.1080/00220389508422377
Special economic zones in India. http://www.sezindia.nic.in/ (accessed December 1, 2020).
Wang, J. (2013). The economic impact of Special Economic Zones: Evidence from Chinese
municipalities. Journal of Development Economics, 101, 133–147. https://doi.org/10.1016/j.jde
veco.2012.10.009
Zeng, D. (2010). Building engines for growth and competitiveness in China: Experience with Special
Economic Zones and industrial clusters. World Bank Publications, 1–260. https://openknowledge.
worldbank.org/handle/10986/2501
Zeng, D. (2016). Special Economic Zones: Lessons from the global experiance. PEDL Synthesis Paper
Series.
Zhaoying, L. (2021). Quantifying the eects of Special Economic Zones using spatial econometric
models. Discussion Papers in Economics and Business 21-01, Osaka University, Graduate School
of Economics.
16 E. DUBININA

APPENDIX
Table A1. Linear regression with instrument variable.
REDUCED FORM (1) (2) (3) (4) (5) (6)
Instrument −0.002***
(0.000)
−0.002***
(0.000)
−0.003***
(0.000)
−0.002***
(0.000)
−0.002***
(0.000)
−0.003***
(0.000)
Logarithm of labour −0.045***
(0.007)
−0.026***
(0.006)
−0.045***
(0.007)
−0.026***
(0.006)
Logarithm of capital 0.346***
(0.012)
0.356***
(0.012)
constant 16.568***
(0.134)
17.350***
(0.138)
11.473***
(0.249)
16.568***
(0.134)
17.350***
(0.138)
11.473***
(0.249)
Number of observations 101,100 79,994 77,758 101,100 79,994 77,758
FIRST STAGE (7) (8) (10) (11) (12) (13)
Instrument 0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
Logarithm of labour −0.001***
(0.000)
−0.000*
(0.000)
−0.001***
(0.000)
−0.000*
(0.000)
Logarithm of capital 0.000
(0.000)
0.000
(0.000)
constant −0.002
(0.004)
0.008
(0.005)
0.004
(0.008)
−0.002
(0.004)
0.008
(0.005)
0.004
(0.008)
Number of observations 145,591 114,053 86,742 145,591 114,053 86,742
Lags No No No No No No
Leads No No No No No No
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Region fixed effects Yes Yes Yes Yes Yes Yes
Business activity fixed effects No No No Yes Yes Yes
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings.
Table A2. Multi-way fixed effects model with one lag.
(1) (2) (3) (4) (5) (6)
SEZ 0.574***
(0.137)
0.510***
(0.142)
0.553***
(0.164)
0.574***
(0.137)
0.510***
(0.142)
0.553***
(0.164)
L1. SEZ 0.133
(0.138)
0.077
(0.141)
−0.095
(0.0156)
0.133
(0.138)
0.077
(0.141)
−0.095
(0.156)
Logarithm of labour −0.056***
(0.007)
−0.036***
(0.007)
−0.056***
(0.007)
−0.036***
(0.007)
Logarithm of capital 0.328***
(0.012)
0.328***
(0.012)
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Region fixed effects Yes Yes Yes Yes Yes Yes
Business activity fixed effects No No No Yes Yes Yes
constant 15.547***
(0.001)
16.303***
(0.054)
10.551***
(0.224)
15.547***
(0.001)
16.303***
(0.054)
10.551***
(0.225)
Number of observations 96,221 76,848 74,609 96,221 76,848 74,609
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings.
POST-COMMUNIST ECONOMIES 17

Table A3. Multi-way fixed effects model with two lags.
(1) (2) (3) (4) (5) (6)
SEZ 0.539***
(0.130)
0.493***
(0.143)
0.512**
(0.158)
0.539***
(0.130)
0.493***
(0.144)
0.512**
(0.158)
L1. SEZ 0.125
(0.152)
0.101
(0.149)
0.063
(0.159)
0.125
(0.152)
0.101
(0.149)
0.063
(0.159)
L2. SEZ 0.170
(0.140)
0.105
(0.119)
−0.030
(0.132)
0.170
(0.141)
0.105
(0.120)
−0.030
(0.132)
Logarithm of labour −0.052***
(0.007)
−0.035***
(0.007)
−0.052***
(0.007)
−0.035***
(0.007)
Logarithm of capital 0.297***
(0.013)
0.297***
(0.013)
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Region fixed effects Yes Yes Yes Yes Yes Yes
Business activity fixed effects No No No Yes Yes Yes
constant 15.602***
(0.001)
16.334***
(0.054)
11.137***
(0.233)
15.602***
(0.001)
16.334***
(0.054)
11.137***
(0.233)
Number of observations 90,747 72,021 69,774 90,747 72,021 69,774
Notes: Standard errors in parentheses: * p < 0.05, ** p < 0.01, *** p < 0.001. Standard errors are clustered on the level of
multiplication region and business activity. Dependent variable: logarithm of earnings.
18 E. DUBININA