The Political Legacy of the Gulag
Archipelago
Natalia Kapelko
and
Andrei Markevich1
1 Natalia Kapelko is from Yandex corporation (Moscow, Russia); Andrei Markevich
is from New Economic School (Moscow, Russia), and HSE/NES Laboratory on
Russian Economic History, Higher School of Economics (Moscow, Russia).
Address for correspondence: [email protected]; New Economic School, Office
1902, Nakhimovsky pr. 47. Moscow, Russia.
This paper is part of the research agenda at the HSE/NES Laboratory on Russian
Economic History (Moscow, Russia). We would like to thank Paul Castaneda Dower,
Simeon Djankov, Mark Harrison and Ekaterina Zhuravskaya for their comments.
2
Abstract.
We study the long-run consequences of the Gulag on political preferences in modern
Russia. We find that the location of a Gulag camp in a district is positively associated
with anti-communist voting during the last Soviet 1991 referendum and the first post-
Soviet 1996 presidential election. We interpret these results causally, arguing that
Gulag camps provided evidence on Soviet political repression, which, in turn,
affected the elections’ results. We provide additional evidence on the validity of the
“information” channel.
3
1. Introduction
Rapidly growing empirical research on the long-run consequences of history provides
many convincing examples that past experience affects development today (Nunn
2009). The question how history matters remains less clear. Institutions and culture
are the two most often cited mechanisms of persistence because they are stable and
slowly changing (North 1990; Greif 2006; Tabelini 2008). However, the literature
suggests limited explanation what makes them persistent; in what environment they
could change; what the effect of particular historical events and phenomenon on
formation of institutions, cultural norms and preferences is. Potential endogeneity of
historical episodes to previous development creates further difficulties to address
these questions empirically. We take advantage of the history of the Soviet command
system as the largest modern social experiment to study this problem (Djankov et al.
2003). We explore how the Gulag, a system of forced labor camps initiated by Stalin
in 1929, affected local political preferences in Russia during the transition from
communism to capitalism. We argue that the vivid historical memory and tangible
information about the Gulag in the districts where camps were set up increased anti-
communist voting half a century later. The past location of a Gulag camp in a district
cut communist share of voters by 2 percentage points at the 1996 presidential
election, the only election in Post-Soviet Russian history when the communist
candidate Genadij Zuganov had a chance to win. We view this 2 percentage points as
lower bound estimation of the costs since we can identify only the local effect of the
Gulag on political preferences, but there would also be a general national effect that
we cannot pick up with our data.
Repression and the forced labor system were important tools of Stalin’s
internal policy, which he widely used to penalize alleged dissidents and to get cheap
4
labor for Soviet industrialization projects. The administration of forced labor camps,
known by the Russian acronym GULAG (Glavnoe Upravlenie Lagerej), consisted of
several hundreds camps. Camps population varied between one and two and a half
million, i.e. about one percent of the Soviet population at that time. The flow of
prisoners passed through the Gulag system over the decades of Stalin’s rule was much
larger, compassing between fifteen eighteen million people. The system had limited
scale before Stalin and was dismantled by his successor Khrushchev (Gregory and
Lazarev, 2007).
Alexander Solzhenitsyn (1974) used the expression “the Gulag archipelago” to
characterize camps' geography. Indeed, Soviet penal labor system represented isolated
camp settlements distributed over the whole territory of the USSR without clear
spatial pattern (figure 1). Stalin chose the precise camps’ locations without consulting
local interests, i.e. camps construction was exogenous to local preferences. The main
factor affecting the shape of the archipelago was labor demand defined by Stalin’s
industrialization program (Khlevnuk 2004). On the one hand, industrialization
initiated mass scale construction of new plants and factories in remote areas
(Markevich and Mikhailova 2013) that led to Gulag camp construction there. On the
other hand, Gulag camps appeared in old industrialized centers – like Moscow, the
Urals etc. - because of the Gulag ability to concentrate labor resources rapidly. The
overall effect was that, as Mikhailova (2012) commented, “a Gulag map looks much
like the map of population density”.
Figure 1 somewhere here.
In this paper we explore the long-run consequences of the Gulag geography on
political preferences in modern Russia, namely onto the last Soviet 1991 referendum
on preservation of the Union and the first post-Soviet 1996 presidential election – the
5
two key elections when Russia made its choice between communist and capitalist
development paths. Due to data quality we concentrate primarily on the 1996 election
for which we have district level data, in contrast to the regional level 1991 dataset.2
We find that past presence of a Gulag camp in an area was positively
associated with anti-communist voting in the 1990s. Because of exogenous Gulag’s
geography we interpret this correlation causally. We argue that location of a Gulag
camp in a district improved local knowledge on Soviet political repression both via
both “inside” (memories of released Gulag detainees settled locally) and “outside”
(historical experience of local population witnessed the Gulag from beyond the fence)
information channels. While we cannot rule out the omitted variable problem
completely, the overall effect we find appears robust across various specifications and
datasets. In particular, in case of the 1996 election, for which we have district level
dataset, we use region fixed effects to control for locally invariant omitted variables.
Our results hold if we control for current wages, i.e. potential location of Gulag camps
in areas that had more economic prospects cannot fully explain the main finding.
With this finding we contribute to the literature on the importance of history
for the long-run development and persistency in preferences. A wide variety of papers
have documented distinct stability of social and cultural norms and beliefs over time.
Among the most recent studies, there are Voigtländer and Voth (2012), Guiso et al.
(2013) and Grosjean (2014). Voigtländer and Voth (2012) find stability of anti-
semitism patterns in Germany in the second millennia; Guiso et al. (2013) provide
evidences on persistence in civic capital in Italy over the same time span; Grosjean
(2014) demonstrates constancy of the culture of honor in the US South etc. We
2 The 1996 dataset consists of more than two thousand voting districts, one district for
each local voting commission, while the 1991 dataset covers 83 regions only.
6
develop one of the lines of this research, by investigating the effect of history on
current political preferences. Analysing outcomes of elections in contemporary
Poland, Grosfeld and Zhuravskaya (2013) uncover the causal effect of the more-than-
a-century partition of the country between three empires on modern political
preferences. In another paper Grosfeld et. al. (2013) find further evidences of
persistency in political preferences in modern Eastern Europe, in particular strong
socialist anti-market culture in the former Pale of Settlement - the area where Jewish
population was confined to in the Russian empire. Similar, Acharya et. al (2014)
document the political legacy of slavery in the US. In contrast to these papers, we
look at relatively short-lasting historical shock that had long lasting consequences. In
this respect our paper is similar to the study by Acemoglu et al. (2011) who
investigate economic and political consequences of the Holocaust. Finally, our results
support the analysis by Castaneda Dower and Markevich (2014) who study the
connection between privatization imperial and post-communist Russia and find that
historical precedent impacts modern attitudes towards policy reforms.
These results improve our understanding of voting behavior and outcomes
during the transition period, with voters who remembered the experience of living
under communism. Past experience affected voting outcomes in favor of democratic
values in a new Post-Soviet reality (Rose and Mishler 1996). Specifically in the
Russian context Rose, Tikhomirov and Mishler (1997) argue that it is too naive to
focus exclusively on current events to explain Russian election outcomes. We provide
quantitative evidence for this claim and add a substantial nuance into the dominant
picture of widespread nostalgia for the communist era after the first decade of
transition hardships (Mason 1997, Sullivan 2013).
7
Finally, we contribute to the Russian economic history literature by shedding
additional light on the cost of repressions in the Gulag system. The dominant
narrative draws a picture where the dictator Stalin managed to construct and run a
system of forced labor, successfully destroying protests and alleged opposition
(Gregory and Lazarev 2007; Khelvnuk 2004). Our results suggest that his successors
eventually paid the costs.
The rest of the paper is organized as follows. The next section provides a
historical overview of Gulag archipelago history. Then, we formulate hypotheses
(section 3) and describe our data and methods (section 4). The fifth section presents
our analysis and results. Section six concludes.
2. A history of the Gulag archipelago
The Gulag origin goes back to the early Soviet years. Already in 1918, less than a
year after the Bolshevick’s revolution, Lenin suggested to imprison those who
opposed the regime in forced labor camps near big cities. There were 84 forced labor
camps in 34 provinces by the year of 1921 (Applebaum, 2003). The system of forced
labor shrank with the end of the Civil war in 1920, being limited to only the
Solovestky camp during the 1920s. It got a new life after Stalin’s Politburo resolution
issued on June 27, 1929 “On the Use of Labor of Convicted Criminals” that changed
its aim from punishment to providing cheap labor for economic projects (Ivanova
2006; Khlevnuk 2004). The Chief administration of labor camps – the Gulag
administration – was established in 1930 to run these camps. This Gulag
administration was one of several Soviet industrial units and got annual plans like the
others. The only main difference was that it exploited forced labor and because of that
was a part of the Soviet secret police ministry and did not belong to the civic
industrial ministries.
8
The Gulag archipelago saw a rapid growth during the 1930s that accompanied
the first five-years Soviet industrialization plans with their constantly increasing
demand for industrial labor. There were only 179 thousand Gulag forced laborers in
1930; the number increased more than ten times by the end of the decade to about 1.9
million in 1938. The growth continued until the year of Stalin’s death, with the peak
of 2.5 million forced laborers in early 1950s (Zemskov 1991; Applebaum 2003;
Khlevnuk 2004). By that time the Gulag accumulated about two percent of Soviet
labor force and produced arguably the same share of national income (Gregory and
Lazarev 2007). In terms of the number of labor camps, the Soviet government
organized more than three hundred separate camps over the years of Stalin’s reign
(calculated from Smirnov 1998); the peak was achieved in early 1950s when 158
labor camps operated simultaneously (Gregory and Lazarev 2003, p. 12).3
Mikhailova (2012), who specifically analyzed the Gulag geography,
concluded that “camps eventually were organized in practically all areas of the
country”. There was no clear geographical pattern in the Gulag archipelago
geography. Many camps were established with a labor-shortage motivation
(Khlevnuk 2004); but these shortages of labor were themselves caused by the rapid
industrialization plans, which initiated many new industrial projects all over the
Soviet Union. The only systematic outcome found by Mikhailova (2012) was that the
3 We employ the term «camp» to summarize the whole variety of the Gulag units,
separated from each other in terms of geography, including corrective labor Camps
(ITL), territorial department of corrective labor colonies (OITK), special camps
(Osoblagi), camp departments (lagernoe otdelenie), separated camp points (otdelnij
lagernij punkt), labor colonies, camps for political prisoners, camps for women and
children, transit camps etc. We do not count Soviet prisons as camps.
9
Gulag appeared as mainly urban phenomena because of industrial projects location.
Mikhailova (2012) estimated that “83% of all camps were located no more than 35
kilometers from a settlement with city status now, and 66% of all camps were no
more than 35 kilometers from a city of at least 10,000 inhabitants in 1939”.
Political victims of Stalin’s repressions as well as criminals composed the
Gulag population. Many political prisoners were victims of random repressions. It is
hardly possible to separate innocent people convicted because of political reasons
from those who were sentenced to the Gulag because of criminal actions. First, Stalin
introduced many criminal laws, like criminal penalty for violation of the labor code,
that would hardly get anything else than an administrative sanction in a more
democratic environment (Belova and Gregory 2011). Second, we do not have detailed
data on the distribution of the Gulag population along even these biased lines. If one
counts only those who were repressed for so called counter-revolution activities (the
lower bound), this gives a share of about twenty percent (Zemskov 1991). Anecdotal
evidence suggests that this share was up to a half of the Gulag population (Lazarev
and Gregory 2007). Irrespective of the initial reasoning, life in a Gulag camp was
very difficult for both criminals and political victims; mortality rate during the Second
World War jumped up to twenty – twenty five percent of the Gulag population
(Zemskov 1991).
A shift in internal policy after Stalin’s death and the ban on further political
repressions led to gradual dismantling of the Gulag archipelago during the 1950s.
Already in 1953 – the year of Stalin’s death - the new Soviet government conducted a
wide amnesty for prisoners, which halved their number. Many industrial projects and
corresponding camps were shut down. The Gulag chief administration itself was
10
closed and reorganized in 1960. However, some of the former Gulag camps continued
their existence as a part of the Soviet penal system until the 1980s.
Former Gulag prisoners were severely limited in their choice of living even
after the end of their sentences, both legally and economically. Many of political
prisoners got a court ban to live closer than 101 km from big cities like Moscow,
Saint-Petersburg, Kiev etc. or were sent into exile in remote areas (usually close to
former camps) after their exit from the Gulag. Even those, who could choose, often
had to settle down in the same areas where they were prisoners and continue to work
at the same industrial projects as free laborers because their outside options were
heavily limited (Cohen 2010, P. 50). Local authorities encouraged them not to out-
leave. In other words, descendants of former Gulag prisoners often continued to live
in the districts of past camp location.
Local population in districts with the Gulag camps were better aware of the
Soviet system of forced labor because of both contacts with former Gulag prisoners
and their descendants as well as general experience of witnessing the Gulag from
beyond the fence. In particular, the Gulag forced labor constructed tangible
infrastructure extensively used by local population (Latypov 2011). In contrast, at
national level all Soviet propaganda denied mass political repressions and exploitation
of the forced labor in the USSR up to the late 1980s. The government put
extraordinary efforts to block alternative information channels about the past. In
particular, reading and dissemination of Solzhenitsyn “Gulag archipelago” was a
criminal offence.
3. Hypotheses
Our main hypothesis is that there is an association between the Gulag camps’ location
in the past and modern political preferences, in particular a positive link with anti-
11
communist voting during the early post-communist period. We expect this correlation
mainly because of individual perceptions of life in the Soviet Union caused by the
knowledge of history of Stalin’s political repression that shaped their modern attitudes
against the communist regime. In the districts where camps were located, these
perceptions were more negative due to better knowledge on the repression of
innocents and Gulag hardships either via personal experience and contacts with
former Stalin’s victims4 or historical experience of free population which observed
the Gulag in operation from outside the camps. Relatively low out-migration of
former Gulag prisoners concentrated contacts with them in these districts. Physical
remains of former camps like barracks, guard towers as well as industrial projects
constructed by prisoners were an important reminder of local forced labor history
(Latypov 2011). We expect that both “inside” (experience and contacts with former
prisoners) and “outside” (experience of local population) information channels on the
Gulag operated, while with our data we cannot distinguish between the two.
In addition, we expect that voting was more anti-communist in the districts
where local civil activists put additional efforts to keep historical memory on political
terror, organizing local civil rights organization that preserved memory about the
Gulag (“Memorial” foundations) and provided more information on the Soviet history
of repressions to citizens.
We expect that the negative perceptions about the Soviet past were especially
pronounced during the key elections, when the choice of the future political
development path was on the table, such as the 1991 referendum on preservation of
the Soviet Union and the 1996 presidential election in Russia. Indeed, both elections
4 According to public polls, 27 percent of Russians said that had relatives who were
penalized under Stalin (Cohen 2010, P. 150).
12
had a strong ideological component and appealed to the country’s past experience.
The 1991 referendum took place on March, 17 and was the last attempt of the Soviet
leadership to keep the unity of the Soviet Union and secure the socialist development
path. More than two thirds of voters answered positively to the referendum’s question
“Are you for the preservation of the Union of Soviet Socialist Republics as a renewed
federation of equal sovereign republics where the rights and freedom of all
individuals of any ethnicity will be fully guaranteed?” We interpret negative answers
as at least partially driven by anti-communist views and a desire to abolish the non-
democratic communist state established seventy years earlier.
Similarly, the whole 1996 presidential campaign was structured around the
debate of potential reversal of transition reforms and the choice of political regime.
As White, Rose, McAllister (1997) underlined
“The 1996 presidential election offered voters a choice between candidates
offering competing regimes. President Boris Yeltsin … chief opponent,
Communist Party leader Gennady Zyuganov, was heir to the tradition in
which the end justified the means; the party had created Soviet-style
"socialist legality" in which millions of Russians had been killed or
imprisoned in Siberia on political grounds.”
We analyze the results of the second round of the election, which took place
on July, 3 1996, when the communist opposition leader Gennady Zyuganov
challenged the incumbent candidate Boris Yeltsin (in the first round of the election on
June, 16 1996 Genady Zyuganov got 32 per cent of voters and Boris Yeltsin got 35).
Yeltsin won with 54 percent of the votes. He structured his campaign along
ideological anti-communist lines. His main slogan was «vote or lose» that appealed to
those who would be worse off if the communist candidate had won. He tried to
13
present himself as a populist leader, running intentionally as non-party candidate in
contrast to the communist boss Genadij Zuganov who was presented as an apparent
heir and potential re-constructor of the communist regime, which collapsed only five
years earlier.
An alternative interpretation of potential association between the Gulag
location and anti-communist voting could be that the Gulag camps and industrial
projects they realized created some economic “benefits”, which, in turn, affected
voting behavior. Indeed, the Gulag appeared to realize Stalin’s industrialization
projects. The Gulag geography was positively correlated with Stalin’s
industrialization (Khlevnuk 2004) and, correspondingly, with the economic prospects
of an area. This would create an “economic” channel of the Gulag effect. Wealthier
citizens were potential winners of transitional reforms and because of that would be
strongly against their reversal (Denisova et al. 2013).
4. Data and methodology
To test our hypotheses we construct a unique dataset matching forced labor camps’
locations in the past with modern election outcomes and current social and economic
variables. Using the variation in the geography of the Gulag we estimate the impact of
the Gulag legacy on post-Soviet political preferences. With regional fixed effects and
economic present-day variables we control for potential economic channel of the
Gulag legacy, highlighting the information effect.
We combine information gathered from recent publications on the history of
the Gulag extracted from the former Soviet archives and official statistical sources
(official results of the 1991 and 1996 elections and municipality level statistics
collected by the Russian Ministry of finance) to construct our dataset. On top of that,
we collect data on present-day prisons and labor camps from the reference book “Vse
14
Turmy Rossii” [All Russian prisons] (2005) and location of “Memorial” regional
branches (one of the main aims of which is to preserve memory on the Gulag) from
the “Memorial” website.
We extract details on the geographical location of 352 forced labor camps
from Smirnov (1998) who in his turn collected them from original archival sources
located in the State Archive of Russian Federation (the fund of the Chief
Administration of labor camps at the Ministry of Internal Affairs). We then match the
Gulag locations with the 1991 Russian regions (former union republics are excluded)
and the 1996 Russian election districts.
There are 2347 election districts in the original 1996 Russian presidential
election dataset. We had to exclude 131 districts either because of duplicate voting
information (districts with identical voting outcomes but different in terms of
location) or because of lack of their geographical identification. There was at least
one Gulag camp in 210 out of 2204 1996 election districts in our sample. Only 6
districts had more than four camps and another ten had four; fifteen had three and
forty four had two; we suspect that there may be double counting in some cases in the
original source, when for example a camp changed its status or name.
Table 1 presents summary statistics of the 1996 dataset for Gulag and non-
Gulag districts, correspondingly. In 1996 in an average election district Boris Yeltsin
got 45.5 percent of voters, and the communist leader Zuganov got 48.9 percent, 4.23
percent of citizens voted against both candidates (but because of differences in
election districts size Yeltsin won with 54 percent of votes at national level). The
shares were very different in the Gulag districts – 54.2, 39.3 and 5.4, respectively. In
contrast, in an average non-Gulag district Zuganov won, getting a bit more than 50
per cent of votes (the difference between the Gulag and no-Gulag districts is
15
statistically significant at the one percent level). Figure 2 shows variation in the 1996
election outcomes and the Gulag camps’ location.
There are other differences in the characteristics of Gulag and non-Gulag
districts that might affect the election outcome, and for which it is important to
account. First, another important election variable – the turnout – differed
substantially, being higher in non-Gulag districts: 72.7 percent as opposed to 68.8 in
the Gulag districts (the difference is statistically significant at the one percent level).
Previous political science studies found that settlement status as well as welfare and
population age structure, in particular share of the elderly, affected strongly Russian
elections in the 1990s (Brudny 1997; Clem and Craumer 1997; Gehlbach 2000;
Sullivan 2013; Mason and Sidorenko-Stephenson 1997; McFaul, 1996; White 2010;
White, Rose and McAllister 1997). The Gulag districts were more urban (urban share
composed 58.2 percent in the Gulag districts vs 42.0 in the non-Gulag ones; the
difference is statistically significant at the one percent level) and wealthier (wages
were about 850 and 586 thousand rubles per month; the difference is statistically
significant at the one percent level), had smaller number of retired citizens (25.1 vs.
27.6 percent; the difference is statistically significant at the one percent level) and
larger number of citizens with high education (7.1 vs. 5.8 percent; the difference is
statistically significant at the one percent level). In terms of population density, the
Gulag districts were a bit more populated on average. Finally, because of the
historical legacy of the Soviet forced labor system and its slow transformation the
Gulag districts had more present-day prisons located on their territory. They also have
more regional branches of the Memorial civil rights organization, which aims to
preserve the memory about the Gulag.
16
Table 2 reports descriptive statistics of the 1991 referendum for Russian
regions with and without Gulag labor camps. There were six camps in the average
Gulag region. Regions with Gulag camps voted less for the preservation of the Soviet
Union in 1991 – 73.6 percent against 79.7 (the difference is statistically significant at
the one percent level). Similar to the 1996 districts with camps, 1991 regions with
camps had smaller turnout and share of pensioners on average; they were more urban
and prosperous. Finally, they had larger share of citizens with higher education and
more preset-day prisons.
To test our hypothesis on the political legacy of the Gulag archipelago we
employ the geographical variation in camps, regressing the shares of anti-communist
voting in 1991 and 1996 on the Gulag variables. We prefer to work with shares to
make election districts comparable to each other. To be precise we estimate the
following equation:
ShareAnti-communisti = α + β*Gulagi + Ω* (Controlsi) + (Regioni) + εi (1)
where i denotes the unit of observation, either region for the 1991 referendum or
election district for the 1996 election. ShareAnti-communist is the dependent
variable, either the share of voters against the preservation of the Soviet Union at the
1991 March referendum or the share of voters against the communist leader Zuganov
in the second round of the 1996 presidential election (defined as the sum of shares of
voters for the incumbent candidate Yeltsin and against all candidates). Gulag is our
main variable of interest characterizing either presence of the Gulag camps in an area
(dummy variable) or their number. Controls denote control variables we use in
various specifications. We include voter turnout and urban shares, which are standard
controls for political science literature, in all specifications. We gradually expand our
list of controls adding parameters that might affect voting outcomes and would help
17
us to fix potential economic channel of the Gulag effect, in particular we account for
present-day prisons, average wages in logarithms, shares of pensioners and citizens
with high education. For the 1996 election we control also for population density in
all specifications. Finally, to account for potential spatial patterns and regional macro
variables we use mega-regions fixed effects for the 1991 referendum and regional
fixed effects for the 1996 election that is denoted by Region. For the 1996 election we
adjust standard errors by clustering at the region level.
To provide further evidence that knowledge on Stalin’s labor force camps
drives the Gulag results for the 1996 presidential election, we adjust equation (1) to
explore how the proximity to potential sources of information affects the results.
5. Empirical results
Table 3 presents our results for the 1996 presidential election. We employ the Gulag
dummy as our variable of interest because of little variation in number of camps in the
Gulag districts (135 out of 210 Gulag districts had one camp). This also allows us to
avoid the problem of the potential double counting of Gulag camps.
The results of our basic specification are in column 1. The coefficient on the
Gulag dummy is positive and highly significant. The location of a Gulag camp in the
district is positively associated with the anti-communist share of votes at the 1996
presidential election. Ceteris paribus the share of anti-communist votes is 2.5
percentage points larger in the Gulag districts than in the districts without camps. The
coefficients on the control variables demonstrate that anti-communist voting was
more revealed in the urban densely populated districts. In the basic specification we
already control for regional fixed effects that allows us to account for non-observable
region invariant variables. In particular, with region fixed effects we account for
Stalin’s industrialization strategy to a huge extent, given that the Soviet government
18
did not specify national plans at district level, compiling them at regional level as
maximum (Gregory and Stuart 1999). To account for other variables that might affect
election outcome and to fix potential economic channel of the Gulag effect we
employ additional controls (columns 2-6).
In column 2 and 3 we add dummies for present-day prisons (sizo) and labor
camps (ITL), which location is path dependent and because of that is correlated with
the Gulag geography. Because of an extensive debate on the potential falsification of
election outcomes in favor of incumbent candidates in post-Soviet Russia (Myagkov
and Ordeshook 2008), one could argue that incumbent candidates could get more
votes in the election districts with present-day prisons, as it was easier to falsify the
outcome since prisons are closed institutions under tight control of administration.
Indeed, the coefficients on both dummies are positive and significant at the one
percent level, which suggests a positive association between post-Soviet prisons and
camps and voting against the communist leader, yet the coefficient on our main
variable of interest remains positive and significant at the one percent level.
In column 4 we report the results of the regression analysis with average
district wage variable. With wages we control for current wealth in a district that
might be affected by the Soviet industrialization where Gulag labor forced was widely
used, i.e. we narrow down potential economic channel of the Gulag effect. Even if
there is no economic channel of the Gulag effect, current wages is still an important
variable that shapes voting behavior. Indeed, the coefficient on wages is positive and
highly significant, but the coefficient on the Gulag dummy remains positive and
significant as well, while diminishing in magnitude by one third. I.e. we cannot reject
an economic channel hypothesis, but this channel cannot rule out our main finding.
19
We interpret this as further evidence supporting the information channel of the Gulag
effect.
Column 5 and 6 present the results of the specifications with shares of
pensioners and citizens with high education. First, the existed literature (Sullivan
2013; White 2010; White, Rose and McAllister 1997) argues that these variables are
important voting determinants in modern Russia; second, the Soviet industrialization
affected both parameters, and we could pin down potential economic channel. Our
results corroborate previous findings that the share of pensioners is negatively
associated with anti-communist voting, and the share of citizens with high education
is associated positively. Missing values of these controls cut our sample substantially,
but the main correlation between the Gulag presence and anti-communist share
survives.
Table 4 provides further evidence that the correlation between 1996 voting
outcome and the camps geography reveals the political legacy of the Gulag. In
column 1 we adjust our main specification controlling for election district area taken
in logs and adding an interaction term between this variable and the Gulag dummy.
The coefficient on the interaction term is negative and significant, i.e. the legacy of
the past is more pronounced in small districts. Indeed, in sufficiently large districts
citizens could be unaware of past events if they took place in a distant part of the
district. Because district size could be correlated with urbanization level as well as
local prosperity, this finding could be interpreted in support of the economic channel
of the Gulag effect. To explore information channel effect further we employ linear
distance from the center of an election district to the nearest Gulag camp (column 2).
The coefficient on this variable is negative and highly significant. One standard
deviation increase in a distance to a Gulag camp (0.96) decreased the distance to
20
Gulag effect by 0.02 percentage points or about eight per cent. In column 3 we allow
the Gulag effect to vary in the number of camps, adding dummies for districts with
more than one camp and more than two camps, respectively. Positive coefficients on
these dummies suggest that the correlation between Gulag camps and modern
political preferences increases with the number of camps in a district, while these
coefficients are imprecisely estimated. We interpret these results as evidence of the
information channel effect. Citizens were more likely aware about Stalin’s
repressions, if they had more tangible evidence on that, i.e. the closer a Gulag camp
was located, the higher density of camps was, and this knowledge on Soviet history
affected their voting behavior. Results presented in columns 4 and 5 support such
interpretation where we add Memorial regional branch dummy and its interaction
with the Gulag dummy. Coefficients on the Memorial dummy are positive and
significant. One of the aims of the Memorial civil rights organization is to preserve
memory of Stalin’s repressions; its presence in a district is positively associated with
anti-communist voting behavior. While one could argue that the Memorial geography
is endogenous to political preference, this is less likely for the interaction term. The
coefficient on the interaction is positive while imprecisely estimated; Gulag districts
with Memorial branches voted less for communists potentially because of Memorial
efforts to preserve the historical memory of repressions.
Table 5 reports our results for the 1991 referendum. Similar to the 1996
election, we first run our basic specification with minimum controls and then expand
their list gradually (while this cuts our sample from 83 to 69 regions). In contrast to
the 1996 exercise we employ number of Gulag as our variable of interest (not a Gulag
dummy). The 1991 regions are much larger in terms of territory than the 1996
election districts because of that there is substantial variation in number of camps in
21
the Gulag regions; there are also only a limited number of regions without camps. The
coefficient on the Gulag variable is negative and significant in all specifications, i.e.
Soviet citizens were less for preservation of the Union in regions where there were
more Gulag camps during Stalin’s reign. A one standard deviation increase in the
number of camps in a region (5.54) is associated with 0.02 percentage points decrease
in the share of votes for the preservation of the Soviet union. Given the limited
sample, we view this result as supportive evidence on the political legacy of the
Gulag which we found for the 1996 election.
Finally, for the 1996 election we do a number of robustness exercises to check
how stable our results are. First, we repeat our basic specification in logs instead of
shares controlling for population size. Second, instead of regional dummies we use
controls for longitude and latitude. Third, we replace the anti-communist share of
votes for the communist share, i.e. not counting abstain votes. Our main results hold.
6. Conclusions
We find a positive correlation between the presence of the Gulag in a district and
present-day political preferences. The identification of the effect is based on within
regions geographical variation of the Gulag camps. While we cannot rule out the
omitted variable problem completely, this finding is very stable in various
specifications. We interpret this correlation causally, as the political legacy of the
Gulag, arguing that in the former Gulag districts citizens were better aware of the
Soviet past and Stalin’s repressions (information channel), and this affected their
voting when the choice on the future development paths was on the table in the 1990s.
With regional fixed effects and present-day economic variables we fix potential
economic channel of the Gulag effect, while cannot reject its existence.
22
At the 1996 presidential election the communist opposition leader Genadij
Zuganov got 2.5 per cent less votes in the Gulag districts than in non-Gulag districts.
Counterfactually it means, that if there was no Gulag under Stalin, the anti-communist
share at the national level would be about 0.5 percentage points less than it actually
was. That would not change the 1996 election outcome and Boris Yeltsin’s second
presidential term, but one should view this estimation as a low bound of the political
legacy of the Gulag because information on Stalin’s repressions was available for
citizens in non-Gulag districts as well.
23
References.
Acemoglu, D., T. Hassan and J. Robinson (2011). ‘Social Structure and Development:
A Legacy of the Holocaust in Russia.’ The Quarterly Journal of Economics, 126:
895–946.
Acharya, Avidit, Matthew Blackwell, and Maya Sen (2014). The Political Legacy of
American Slavery. Mimeo.
Applebaum, Ann (2003). Gulag. A History. New York: Doubleday.
Andrew (1996). The price of nationalism: Evidence from the Soviet Union. Public
Choice, 87 (1-2): 1-18.
Blyth, Stephen (1995). The Dead of the Gulag: An Experiment in Statistical
Investigation, Journal of the Royal Statistical Society, Series C (Applied
Statistics), 44 (3).
Brudny Yitzhak M. (1997). In pursuit of the Russian presidency: Why and how
Yeltsin won the 1996 presidential election. Communist and Post-communist
Studies, 30(3): 255–275.
Castañeda Dower, Paul and Andrei Markevich (2014). A History of Resistance to
Privatization in Russia, Journal of Comparative Economics, Forthcoming.
Clem Ralph S. and Peter R. Craumer (1997). Urban-Rural Voting Differences in
Russian Elections, 1995-1996: A Rayon-Level Analysis. Post-Soviet Geography
and Economics, 38(7): 379-395.
Djankov, S., E. Glaeser, R. La Porta, F. Lopez-de-Silanes and A. Shleifer (2003).
“The New Comparative Economics.” Journal of Comparative Economics, 31(4),
595-619.
24
Gehlbach, Scott (1997). Shifting Electoral Geography in Russia's 1991 and 1996
Presidential Elections, Post-Soviet Geography and Economics, 2000, 41(5): 379-
387.
Gregory, Paul R. and Valery Lazarev (2007). The Economics if Forced Labor: the
Soviet Gulag. Stanford, CA: Hoover Institution Press.
Gregory, Paul R. and Robert C. Stuart (1999). Comparative Economic Systems.
Houghton Mifflin Company
Grosfeld, I., A. Rodnyansky and E. Zhuravskaya (2013). «Persistent Anti-Markert
Culture: A Legacy of the Pale of Settlement after the Holocaust,' American
Economic Journal: Economic policy 5(3): 189–226.
Grosfeld, I. and E. Zhuravskaya (2013). ''Persistent effects of empires: Evidence from
the partitions of Poland', Working paper, February 2013.
Grosjean, P. (2011). 'The Weight of History on European Cultural Integration: A
Gravity Approach,' American Economic Review, 101(3): 504-08.
Grosjean, P. (2014). «A History of Violence: The Culture of Honor and Homicide in
the US South", Journal of the European Economic Association, Forthcoming.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales (2013), ‘‘Long Term Persistence,’’
IEF WP 23/13.
Harris, James R. (1997). The Growth of the Gulag: Forced Labor in the Urals Region,
1929-31, Russian Review, 56(2).
Ivanova, G. M. (2006). Istoria Gulaga 1918-1958 [A History of Gulag 1918-1958]
Moscow: Nauka.
Khlevnyuk, O. V., ed. (2004). Istoria Stalinskogo Gulaga [A History of Stalin's
Gulag], Vol. 3 Economika Gulaga [The Economy of Gulag], Moscow: Rosspen.
Cohen, Stephen (2010). The Victims Returns. Survivors of the Gulag after Stalin.
25
Exeter, NH: Publishing works.
Latypov, Robert. Ed. (2011). Gulag v rossijskoj pamyati. Sbornik statej uchastnikov
rossijsko-germanskoj issledovatelskoj gruppi v Permskom krae. [The Gulag in
Russian memory. A volume of articles by Russian-German research group in Perm
region]. Perm.
Markevich Andrei and Tatiana N. Mikhailova (2013) “Economic geography of
Russia” in Michael Alexeev and Shlomo Weber (Eds.) The Russian Economy.
Oxford University Press.
Mason, David S. and Svetlana Sidorenko-Stephenson (1997). Public Opinion and the
1996 Elections in Russia: Nostalgic and Statist, Yet Pro-Market and Pro-Yeltsin,
Slavic Review, 56(4): 698-717.
McFaul, Michael (1996). Russia's 1996 Presidential Elections, Post-Soviet Affairs,
12(4): 318-350.
Mikhailova, Tatiana (2012): Gulag, WWII and the Long-run Patterns of Soviet City
Growth. Mimeo.
Misha Myagkov and Peter C. Ordeshook (2008). Russian Elections: An Oxymoron of
Democracy. VTP WORKING PAPER #63.
Nunn, Nathan (2009) “The Importance of History for Economic Development.”
Annual Review of Economics, 1(1): 65-92.
Rose, Richard and William Mishler (1996). Testing the Churchill Hypothesis: Popular
Support for Democracy and Its Alternatives, Journal of Public Policy, 16(1).
Rosefielde, Steven (1981). An Assessment of the Sources and Uses of Gulag Forced
Labor 1929-56, Soviet Studies, 33(1).
Smironov M.B. (1998) Sistema ispravitelnno-trudovikh lagerej v SSSR 1923-1960.
Spravochnik [System of labor camps in the USSR, 1923-1960. Reference book].
26
Moscow: Zvenia.
Solzhenitsyn, Alexander (1974). The Gulag Archipelago, 1918-1956: An Experiment
in Literary Investigation. Westview Press .
Sullivan, Charles J. (2013). Missing the Soviet Motherland Nostalgia for the USSR in
Russia Today. Mimeo.
Tabellini, Guido (2010), “Culture and institutions: economic development in the
regions of Europe,” Journal of the European Economic Association, 8(4): 1542-
4774.
White, Stephen (2010). Soviet nostalgia and Russian politics, Journal of Eurasian
Studies, 1(1): 1-9.
White, Stephen, Richard Rose and Ian McAllister (1997). How Russia Votes,
Chatham House Publishers, Chatham, NJ.
Voigtländer, Nico, and Hans-Joachim Voth (2012), “Persecution Perpetuated: The
Medieval Origins of Anti-Semitic Violence in Nazi Germany”, Quarterly Journal
of Economic, 127(3): 1339-1392.
Zemskov V.N. (1991). Gulag (istoriko-sotsiologicheskij aspect) [GULAF (historical
and sociological aspects], Sotsiologicheskie issledovaniya [Sociological studies], 6:
10-27.
27
Figure 1. The Gulag archipelago geography.
Source: Smirnov (1998), available also online at www.memo.ru .
28
Figure 2. The Gulag archipelago geography and anti-communist voting at 1996
Russian presidential elections.
29
Table 1. Summary statistics of 1996 Russian presidential elections for election
districts with and without Gulag camps.
A. Gulag camps districts.
Variable Obs Mean Std. Dev. Min Max Share of votes for Zuganov 210 0.393 0.142 0.128 0.758 Share of votes for Yeltsin 210 0.542 0.139 0.190 0.827
Share of votes “against all” 210 0.054 0.017 0.014 0.135 Turnout share 210 0.688 0.074 0.494 0.971 Urban share 210 0.582 0.342 0 1
Population density (per sq km) 210 0.316 1.092 0.000 8.882
Present-day prisons (Sizo) dummy 210 0.143 0.351 0 1
Present-day labor camps dummy 210 0.381 0.487 0 1
Wage (1996 rubles) 163 846992 489889 282600 2590930 Share of pensioners 142 0.251 0.053 0.070 0.406
Share of citizens with high education 85 0.071 0.039 0.022 0.199
Memorial branch dummy 210 0.138 0.346 0 1 B. No Gulag camps districts.
Variable Obs Mean Std. Dev. Min Max Share of votes for Zuganov 1994 0.500 0.153 0.056 0.811 Share of votes for Yeltsin 1994 0.445 0.147 0.158 0.936
Share of votes “against all” 1994 0.041 0.018 0.000 0.106 Turnout share 1994 0.727 0.080 0.506 1.000 Urban share 1994 0.420 0.368 0 1
Population density (per sq km) 1994 0.288 0.995 0.000 15.603
Present-day prisons (Sizo) dummy 1994 0.070 0.255 0 1
Present-day labor camps dummy 1994 0.158 0.365 0 1
Wage (1996 rubles) 1736 585843 389116 127000 3671000 Share of pensioners 1555 0.276 0.067 0.048 0.550
Share of citizens with high education 801 0.058 0.033 0.015 0.464
Memorial branch dummy 1994 0.016 0.126 0 1
30
Table 2. Summary statistics of 1991 referendum on preservation of the Soviet Union.
Regions with and without Gulag camps.
A. Region with Gulag camps.
Variable Obs Mean Std. Dev. Min Max Share of votes for preservation
of the Union 57 0.736 0.074 0.493 0.914 Number of Gulag camps 57 6.158 5.738 1 26
Turnout share 57 0.760 0.046 0.664 0.852 Urban share 49 0.711 0.095 0.468 0.920
Number of present-day prisons and labor camps 57 29.789 15.363 6 80
Wage (in 1990 rubles) 49 317.331 78.710 237 639 Share of pensioners 49 0.175 0.038 0.050 0.231
Share of citizens with high education 49 0.074 0.014 0.051 0.119
B. Region without Gulag camps.
Variable Obs Mean Std. Dev. Min Max Share of votes for preservation
of the Union 26 0.797 0.071 0.626 0.919 Number of Gulag camps 26 0 0 0 0
Turnout share 26 0.794 0.076 0.588 0.881 Urban share 20 0.618 0.122 0.383 0.815
Number of present-day prisons and labor camps 26 20.769 14.214 3 51
Wage (in 1989 rubles) 20 305.746 108.972 188 589 Share of pensioners 20 0.182 0.060 0.067 0.246
Share of citizens with high education 20 0.072 0.014 0.054 0.113
31
Table 3. The 1996 Russian presidential election and the Gulag legacy.
(1) (2) (3) (4) (5) (6) VARIABLES Share of anti-communist votes Gulag dummy 0.025*** 0.023*** 0.020*** 0.016** 0.011* 0.019*
[0.007] [0.007] [0.007] [0.007] [0.007] [0.010]
Turnout share -0.005 0.005 0.017 0.085 0.124 0.162
[0.103] [0.103] [0.103] [0.081] [0.086] [0.123]
Urban share 0.099*** 0.094*** 0.092*** 0.030*** 0.025** -0.002
[0.009] [0.009] [0.010] [0.011] [0.012] [0.017]
Population density 0.012*** 0.009** 0.009** 0.006* 0.004 0.005
[0.004] [0.004] [0.004] [0.003] [0.005] [0.004]
Present-day prisons (Sizo) dummy
0.032*** 0.027*** 0.029*** 0.030*** 0.003
[0.008] [0.009] [0.007] [0.008] [0.008]
Present-day labor camps dummy
0.016*** 0.014*** 0.012** 0.012
[0.005] [0.005] [0.005] [0.008]
Log wage
0.156*** 0.149*** 0.154***
[0.013] [0.013] [0.016]
Share of pensioners
-0.260*** -0.268***
[0.062] [0.096]
Share of citizens with high education
0.662***
[0.226]
Region fixed effects YES YES YES YES YES YES Constant 0.298*** 0.289*** 0.274*** -1.749*** -1.611*** -1.660***
[0.069] [0.069] [0.069] [0.162] [0.170] [0.254]
Observations 2,204 2,204 2,204 1,899 1,672 750 R-squared 0.738 0.740 0.742 0.797 0.796 0.750
Robust standard errors in brackets. *** p
32
Table 4. Proximity to the Gulag and the 1996 presidential election.
VARIABLES (1) (2) (3) (4) (5) Share of anti-communist votes
Gulag dummy 0.288***
0.022** 0.021*** 0.020***
[0.089]
[0.008] [0.007] [0.008]
Log distance to nearest Gulag camp
-0.020***
[0.004]
Turnout share -0.008 -0.001 -0.005 -0.001 -0.001
[0.103] [0.101] [0.103] [0.103] [0.103]
Urban share 0.097*** 0.093*** 0.099*** 0.099*** 0.099***
[0.010] [0.009] [0.010] [0.009] [0.009]
Population density 0.010** 0.009** 0.012*** 0.010*** 0.010**
[0.004] [0.004] [0.004] [0.004] [0.004]
Log district area -0.012***
[0.004]
Gulag dummy* Log district area
-0.001 [0.003] 2 and more Gulag
camps dummy 0.005
[0.014] 3 and more Gulag
camps dummy 0.011
[0.017] Memorial branch
dummy 0.035*** 0.033**
[0.012] [0.016]
Gulag dummy* Memorial branch
dummy
0.005
[0.020]
Region fixed effects YES YES YES YES YES Constant 0.307*** 0.498*** 0.274*** 0.273*** 0.274***
[0.079] [0.076] [0.069] [0.069] [0.069]
Observations 2,204 2,204 2,204 2,204 2,204 R-squared 0.743 0,748 0,742 0.742 0.742
Robust standard errors in brackets. *** p
33
Table 5. The 1991 referendum on the preservation of the Soviet Union and the Gulag
legacy.
(1) (2) (3) (4) (5) (6)
VARIABLES Share of votes for preservation of the Union Number of Gulag
camps -0.004** -0.003** -0.003* -0.003* -0.003** -0.003* [0.002] [0.001] [0.001] [0.001] [0.001] [0.001]
Turnout share 0.731*** 0.457* 0.455* 0.439* 0.525** 0.573**
[0.141] [0.233] [0.231] [0.232] [0.233] [0.253]
Urban share -0.285** -0.288** -0.288** -0.220* -0.153
[0.112] [0.115] [0.116] [0.119] [0.130]
Number of present-day prisons and labor camps
-0.000 -0.000 -0.000 -0.000
[0.001] [0.001] [0.001] [0.001]
Log wage
-0.009 -0.101 -0.111*
[0.042] [0.061] [0.064]
Share of pensioners
-0.617** -0.758**
[0.263] [0.300]
Share of citizens with high education
-1.120**
[0.522]
Mega-regions fixed effects
YES YES YES YES YES YES
Constant 0.212* 0.63** 0.62** 0.69* 1.21*** 1.29***
[0.113] [0.247] [0.243] [0.353] [0.430] [0.450]
Observations 83 69 69 69 69 69 R-squared 0.520 0.581 0.583 0.583 0.621 0.644
Robust standard errors in brackets. *** p