Professional Documents
Culture Documents
Abstract
This paper intends to study the link between the corruption rate in different
regions of Russian Federation and various characteristics of employers working in
the government agencies in the region. I find out that the heterogeneity in age,
wage and experience gender of government officers brings along smaller corruption
rates. Additionally, estimations shows that higher percentage of female workers
can be associated with smaller reported corruption. Variation in academic degree
does not definitely make any positive or negative contribution. The paper deals
with the data provided by a corruption identification survey which allows to
compare different regions by means of corruption index, and the data of personal
characteristics of the government agencies employees.
Introduction
In the current political and economic situation the governments which are able to
maintain a stable developing economy are supposed to be a guarantee for the prosperity
of the countries. However, in most cases the way to a stabilized economy is accompanied
by negative factors which characterize to efficiency of the authorities: the qualification
of the government officers, motivation, corruption rate, etc.
Each of the factors mentioned above depends on employees working in the government institutes. In this paper I investigate how corruption rates are connected with
composition of personal features of the employees, such as: age, wage, gender, education, length of time worked in each agency. I explore a viewpoint that higher degree of
similarity between individuals in working team induce higher interpersonal trust and
as a consequence may make people prone to socially harmful actions for their own benefits. Motivated by the question whether the employee composition of the government
agencies has a large impact on the corruption in the particular region, I analyze the
available data and run the regression of the constructed corruption index on different
averaged personal characteristics.
The survey data used in this research provides the answers of the entrepreneurs
from a wide range of Russian Federations regions on a number of questions concerning
the effectiveness and the corruptness of government agencies. The survey provides the
information about the overall impression of the entrepreneurs on the business climate
in a particular region (the questions are as follows: is the court system fair, quick,
impartial, uncorrupted; were they visited or inspected by tax officials; did they have
to make informal payments; what was the total annual informal payment, etc.) The
information provided allows to construct a representative index for each region and,
therefore, run a regression using the additional data on government agencies employees.
The main finding indicates that the diversity in age, gender and wage of employees
of government agencies positively effects the business environment in regions. At the
1
same time the effect of education is not clear. In line with expectations the effect of
experience is also significant and positive.
Related Literature
The idea behind the project is that more heterogeneous institution decreases corruption. This topic is covered in both theoretical and empirical literature. One approach
is to study impact of certain personal characteristics. For example, Anand Swamy and
Azfar [2000] using several data sets show that women are less involved in bribery and
are less likely to condone bribe taking. Another attempt to look at the topic was made
in Gugushvili [2011] on the Georgian data basis. He stated that increasing levels of
education in society could positively affect the societal perception of corruption. Rijckeghem and Weder [2000] perform cross-country analysis and find significant relationship
between relative civil-service pay and corruption. Finally, Torgler and Valev [2006] have
established a strong age effect on corruption levels.
There are also numerous papers studying the relationship between corruption and
homogeneity in different levels. In the forthcoming paper [Kravtsova and Oshchepkov,
Ongoing] authors argue that trust which may be cause by similarity influences corruption. While this idea seems quite reasonable I want to test a direct connection between
heterogeneity of people using the the data of Russian Fedstat.
At the same time, Lee, Wang-Sheng, Guven, Cahit [2013] claim quite a very similar
point of view. In the paper, they got a lot of interesting findings on a micro level.
However, the one, which is most relevant in the context of my study is that trust
between people have a little if any influence on the level of corruption within a country.
Furthermore, they show that still bribes are more probable in societies with a bigger
portion of males.
There are as well papers Eric M. Uslaner [2004, 2008] saying that there is a solid
Data
The main idea behind the project was to establish link between diversity of society and
corruption level in Russia. Relying on the assumption that higher degree of similarity
between individuals leads to better interpersonal trust, I exploit cross-region data on
public officials working in state agencies in Russia. Data obtained from www.fedstat.ru
(2013) includes age, wage, gender, education and experience of government workers
by agencies and by regions. Every characteristic is divided in several brackets. For
instance, education is divided by particular degrees (BA, MA, PhD, etc.).
As a measure for diversity I use coefficient of variation, a standardized measure of
dispersion of a frequency distribution. It is defined as the ratio of the standard deviation
to the mean .
CV =
Given the structure of a data, namely that I have a number of people in each bracket
I was able to calculate CV for each agency in each region. That means if agency is
perfectly heterogeneous and there is equal number of individuals in brackets then is
zero and index of diversity is zero. On the opposite, if the is huge homogeneity will
be large and hence index of diversity tends to be large. This approach was applied to
construct CV for age, gender, education and experience.
Since the wage data available contains only average wage for various categories of
3
workers but not their numbers I have to apply another approach. Namely, I use min to
max wage ratio in order to reflect differences in wages, however not taking into account
proportions of the workers.
On the other side of the equation there was a dilemma how to reflect BEEPS data
properly. The data set contains extensive information on how do respondents perceive
court, regulations, tax officials, licensing procedures in the region as well as informal
gifts and other payments which are obviously related to corruption level in the region.
My approach is based on a firm observation level. I run logit regression and asking
a question if there are important factor affecting probability of the reporting certain
corruption level. In order to get more relevant results I divide all agencies into 4 groups:
financial regulators, monitoring regulators, courts, labor force regulators and compare
their importance. Firms respond ordinally from 0 to 4 what they believe corruption
level in certain agency is and I used variation in wage, age, gender, education and
experience as variables of interest. The advantage of this firm-level approach is ability
to control for other firm specific factors, such as their industry, size, age legal status,
being active R&D investor, having website etc.
The Model
In the survey firms reported oridinally and this restricted choice of the potential model.
I construct binary variable indicating if firm reports corruption is severe or very severe
obstacle and apply logit analysis as an interpretable model for a categorical response.
The purpose of the analysis was to see how well that response can be predicted by
the variables of interest such as variation in age, wage, education and gender as well
as responses to other questions like industry firm operates, its legal status and other
controls. The logit model states that the probability of reporting is some linear combination of other observables. Since the coefficient cannot be consistently estimated
= + 1 Dk,r + xi + xr + i
yi,r
(1)
Instead we observe y, the binary response. The coefficients are log odds ratios and
we can interpret them as the expected change in log odds with one unit increase in
independent variable given all of the other variables in the model are held constant.
The choice of different dependent variables depending on the sector implies running
multiple regressions. My approach is to run logit for each parameter of diversity as it
could be incorrect to compare between some of them.
The second step is to study the effect of different agencies on corruption level. Hence
I estimated the model
yi,r
= + 1 Dk,r,1 + 2 Dk,r,2 + 3 Dk,r,3 + 4 Dk,r,4 + xi + xr + i
Results
The main finding is a confirmation of the strong link between diversity of the workers
and their socially harmful activities. Pooled regression shows that for all characteristics
there is positive sigh meaning that homogeneity increases corruption rates. Specifically
age is positive for pooled and also for 3 out of 4 agency groups, with the exception of
Financial Regulators, where it has negative sign. When regressing on experience we
can see the same results as for age, with the same exception.
Secondly, variation in gender which reflects fraction of male workers seems to have
positive effect on the corruption meaning that more diverse societies have lower corruption. Another point to note is significant effect for the wage. With one exception of
6
Conclusion
References
Young Lee Anand Swamy, Stephen Knack and Omar Azfar. Gender and corruption.
2000.
Eric M. Uslaner. Trust and corruption. 2004.
Eric M. Uslaner. Corruption, inequality, and the rule of law: The bulging pocket makes
the easy life. 2008.
Alexi Gugushvili. Education and attitudes towards corruption: What has changed since
1990s in georgia? 2011.
Kravtsova and Oshchepkov. Distrust of public institutions, interpersonal trust and
corruption. Ongoing.
Lee, Wang-Sheng, Guven, Cahit. Engaging in corruption: The influence of cultural
values and contagion effects at the microlevel. 2013.
Caroline Van Rijckeghem and Beatrice Weder. Bureaucratic corruption and the rate of
temptation: do wages in the civil service affect corruption, and by how much? 2000.
Robert Rothschild. Cartel stability when costs are heterogeneous. International Journal
of Industrial Organization, 17(5):717734, 1999.
Benno Torgler and Neven T. Valev. Corruption and age. 2006.
7
7.1
7.1.1
Appendix
Data Description
Corruption Level
Instead of relying on Doing Business ranking which simplifies BEEPS data I assess
corruption level in regions using more profound approach. Namely, to use raw data from
the survey to construct specific indecies: electrical and water connection, customs and
transportation related issues, competition against informal firms, construction permit,
access to finance, the court system, taxation, licensing. The structure of the data
allows to detect whether firm was involved in those procedures by making a request to
government officials and then how much did they benefit or suffer from the relations.
Another words, two types of information can be employed, objective and subjective.
First one represents facts such as number of days to obtain a license, informal gift
requested etc. Another is severity of the issue for the particular firm which is opinion
on how firm would perform if there was no such issue. Both may be effectively used
to construct one index per field such as electricity, taxation, etc. The survey data is
available for 1999, 2002, 2004, 2005, 2009 and 2012 which allows to see how indecies
change over time.
10
Min.
0
0.747
0.754
0.748
0.681
0.680
2.001
2.13
1.916
1.915
1.837
0.289
0.28
0.36
0.203
0.13
0.324
0.322
0.253
0.466
0.265
0.138
0.094
0.091
0.122
0.088
1.545
0.2
0
0
0
0
0
0
0
Max.
1
0.974
0.926
1.314
0.997
0.887
2.344
2.486
2.384
2.352
2.463
0.441
0.519
0.636
0.494
0.282
0.569
0.617
0.735
0.742
1.046
0.351
0.319
0.469
0.349
0.539
8.875
11.6
1
1
1
1
1
1
1
N
4218
2678
2678
2678
2678
2678
2678
2678
2678
2678
2678
2555
2555
2555
2555
2555
2678
2678
2678
2678
2678
2678
2678
2678
2678
2678
4218
4173
4195
4183
3746
4184
4181
4160
4207
7.1.2
Variable
Values
-9 Dont know
-7 Does not apply
0 No obstacle
j30f
1 Minor obstacle
2 Moderate obstacle
3 Major obstacle
4 Very severe obstacle
-9 Dont know
1 Never
2 Seldom
ecaq41c
3 Sometimes
4 Frequently
5 Very frequently
6 Always
-9 Dont know
-7 Does not apply
1 No impact
ecaq44c
j30b
1 Minor obstacle
2 Moderate obstacle
3 Major obstacle
4 Very severe obstacle
-9 Dont know
-7 Does not apply
0 No obstacle
h30
1 Minor obstacle
2 Moderate obstacle
3 Major obstacle
4 Very severe obstacle
11
7.1.3
Agencies homogeneity
The data on official agencies is provided by Unified Interdepartmental Statistical Information System (fedstat.ru). For each agency there is data in a form of number of
people in each group:
Age
Education
Experience
Gender
Under 30
Primary School
<1
Total
30-39
Professional Education
1-5
Female
40-49
Higher Education
5-10
50-59
Post Graduate
10-15
60-65
Candidate of Sc
15-25
Above 65
Doctor of Sc
Above 25
Data distribution
90
80
70
60
Frequency
7.1.4
50
40
30
20
10
0
0.4
0.6
0.8
1.0
1.2
1.4
Coefficient of variation
12
1.6
1.8
2.0
50
80
40
60
30
Frequency
Frequency
100
40
20
0
0.6
20
10
0.8
1.0
1.2
1.4
Coefficient of variation
1.6
1.8
0
0.0
2.0
120
100
100
80
80
Frequency
Frequency
120
60
40
20
20
0.5
1.0
1.5
2.0
Coefficient of variation
2.5
3.0
3.5
0.2
0.4
0.6
0.8
Coefficient of variation
1.0
1.2
60
40
0
0.0
0
0.0
0.2
0.4
0.6
Coefficient of variation
0.8
1.0
7.1.5
Federal Agencies
1
Federal Exchequer
10
11
12
13
14
15
16
27
14
(1)
j30f
(2)
j30f
(3)
j30f
0.964
(0.733)
1.095
(0.738)
0.0570
(0.900)
ageFin
(4)
j30f
(5)
j30f
(6)
j30f
-0.627
(0.897)
1.724***
(0.368)
-1.007**
(0.512)
2.544***
(0.575)
+
-3.044**
(1.247)
0.791*
(0.440)
0.247
(0.633)
2.642***
(0.647)
+
-2.887***
(0.924)
-3.401***
(0.962)
-0.0734
(0.0877)
-0.00319
(0.0814)
0.202**
(0.0972)
0.455***
(0.143)
0.688***
(0.0836)
0.325***
(0.124)
0.394***
(0.122)
0.109***
(0.0416)
0.0513
(0.0502)
0.00349
(0.0914)
-0.0901
(0.147)
-2.754**
(1.125)
3,967
3,965
3,370
Industry effect
-0.715
(0.890)
1.681***
(0.363)
-1.029**
(0.507)
2.341***
(0.569)
-
Legal Status
-2.145***
(0.671)
-0.0619
(0.0871)
-0.00695
(0.0810)
0.196**
(0.0968)
0.448***
(0.142)
0.699***
(0.0831)
0.303**
(0.123)
0.418***
(0.121)
0.0447
(0.0323)
0.0616
(0.0497)
0.0201
(0.0909)
-0.0933
(0.147)
-2.139***
(0.793)
ageCourt
ageMonit
ageLabor
Has website
Visited by tax officials
Requested oper. license
Licensed from foreign company
Informal competitors
R&D investor
Started construction
gdp
firmAge
Medium Firm
Large Firm = 3
Constant
Observations
-1.811***
(0.624)
3,967
3,965
3,370
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
15
(1)
j30f
(2)
j30f
(3)
j30f
-1.249***
(0.430)
-1.288***
(0.433)
-0.721
(0.480)
DegreeFin
(4)
j30f
(5)
j30f
(6)
j30f
1.047**
(0.519)
-2.041***
(0.300)
-0.0710
(0.321)
0.176
(0.222)
+
1.869***
(0.591)
-1.413***
(0.343)
-0.440
(0.359)
-0.0880
(0.248)
+
0.675
(1.262)
0.680
(1.288)
-0.0872
(0.0880)
-0.00260
(0.0815)
0.199**
(0.0974)
0.443***
(0.143)
0.714***
(0.0838)
0.298**
(0.124)
0.422***
(0.122)
0.0602*
(0.0321)
0.0625
(0.0502)
0.0106
(0.0915)
-0.117
(0.148)
-2.116
(1.502)
3,967
3,965
3,370
Industry effect
1.113**
(0.516)
-2.042***
(0.297)
-0.0312
(0.318)
0.153
(0.221)
-
Legal Status
1.567
(0.958)
-0.0758
(0.0876)
-0.00747
(0.0810)
0.192**
(0.0969)
0.449***
(0.142)
0.706***
(0.0832)
0.299**
(0.123)
0.413***
(0.121)
0.0419
(0.0290)
0.0613
(0.0497)
0.0241
(0.0909)
-0.0884
(0.147)
-0.519
(1.097)
DegreeCourt
DegreeMonit
DegreeLabor
Has website
Visited by tax officials
Requested oper. license
Licensed from foreign company
Informal competitors
R&D investor
Started construction
gdp
firmAge
Medium Firm
Large Firm
Constant
1.698*
(0.926)
Observations
3,967
3,965
3,370
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
16
(1)
j30f
(2)
j30f
(3)
j30f
2.686***
(0.689)
2.568***
(0.696)
1.649
(1.037)
expFin
(4)
j30f
Industry effect
-3.184***
(0.541)
1.466***
(0.326)
0.781**
(0.323)
0.368**
(0.171)
-
Legal Status
expCourt
expMonit
expLabor
-0.0644
(0.0871)
Visited by tax officials
-0.00301
(0.0810)
Requested oper. license
0.196**
(0.0969)
Licensed from foreign company
0.447***
(0.142)
Informal competitors
0.699***
(0.0830)
R&D investor
0.292**
(0.123)
Started construction
0.416***
(0.121)
gdp
0.00394
(0.0389)
firmAge
0.0624
(0.0497)
Medium Firm
0.0299
(0.0910)
Large Firm
-0.0880
(0.147)
Constant
-2.066*** -2.251*** -2.624***
(0.279)
(0.362)
(0.467)
(5)
j30f
-3.361*** -3.214***
(0.549)
(0.609)
1.417***
0.888**
(0.330)
(0.438)
0.783**
0.616*
(0.325)
(0.363)
0.395**
0.328*
(0.173)
(0.194)
+
+
-
-0.700*
(0.385)
-0.840*
(0.446)
-0.0959
(0.0879)
0.0131
(0.0815)
0.207**
(0.0975)
0.453***
(0.143)
0.687***
(0.0835)
0.285**
(0.124)
0.413***
(0.122)
0.0183
(0.0347)
0.0392
(0.0506)
0.0349
(0.0916)
-0.0846
(0.148)
-1.295**
(0.525)
3,967
3,965
3,370
Has website
Observations
3,967
3,965
3,370
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
17
(6)
j30f
(1)
j30f
(2)
j30f
(3)
j30f
1.211
(0.833)
1.373
(0.842)
0.526
(0.938)
genderFin
(4)
j30f
(5)
j30f
(6)
j30f
1.368*
(0.720)
-0.118
(0.582)
-0.407
(0.737)
1.018**
(0.475)
+
1.295
(0.805)
-0.124
(0.650)
-1.576*
(0.868)
0.911*
(0.528)
+
Industry effect
1.535**
(0.713)
-0.0768
(0.576)
-0.675
(0.730)
0.865*
(0.469)
-
Legal Status
genderCourt
genderMonit
genderLabor
Has website
-0.0611
(0.0871)
Visited by tax officials
-0.00759
(0.0810)
Requested oper. license
0.197**
(0.0968)
Licensed from foreign company
0.447***
(0.142)
Informal competitors
0.699***
(0.0830)
R&D investor
0.305**
(0.123)
Started construction
0.412***
(0.122)
gdp
0.0426
(0.0293)
firmAge
0.0602
(0.0498)
Medium Firm
0.0178
(0.0909)
Large Firm
-0.0935
(0.147)
Constant
-1.237*** -1.491*** -2.190*** -1.246***
(0.173)
(0.284)
(0.369)
(0.179)
Observations
3,967
3,965
3,370
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
18
3,967
-0.0666
(0.0877)
-0.00329
(0.0813)
0.223**
(0.0974)
0.437***
(0.143)
0.698***
(0.0834)
0.318**
(0.124)
0.435***
(0.122)
0.0448
(0.0321)
0.0582
(0.0500)
0.00902
(0.0912)
-0.106
(0.147)
-1.513*** -2.102***
(0.289)
(0.368)
3,965
3,370
(1)
j30f
(2)
j30f
(3)
j30f
4.172***
(0.789)
4.364***
(0.798)
3.749***
(0.887)
PositionFin
PositionCourt
PositionMonit
PositionLabor
Industry effect
Legal Status
-0.0965
(0.0898)
Visited by tax officials
-0.0150
(0.0844)
Requested oper. license
0.138
(0.102)
Licensed from foreign company
0.412***
(0.150)
Informal competitors
0.689***
(0.0862)
R&D investor
0.327**
(0.131)
Started construction
0.423***
(0.126)
gdp
0.0281
(0.0491)
firmAge
0.0363
(0.0525)
Medium Firm
0.0490
(0.0946)
Large Firm = 3
-0.0450
(0.155)
Constant
-2.446*** -2.711*** -3.204***
(0.273)
(0.360)
(0.472)
(4)
j30f
(5)
j30f
1.285**
1.417**
(0.626)
(0.633)
-3.933*** -4.019***
(0.745)
(0.755)
0.634
0.775
(0.592)
(0.599)
3.868*** 3.953***
(1.372)
(1.384)
+
-
3,735
3,733
3,155
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
19
1.869***
(0.720)
-3.753***
(0.839)
0.402
(0.655)
6.159***
(1.577)
+
-0.413
(0.727)
-0.652
(0.773)
-0.114
(0.0904)
-0.00813
(0.0849)
0.166
(0.102)
0.409***
(0.152)
0.687***
(0.0869)
0.303**
(0.132)
0.445***
(0.127)
0.0527
(0.0521)
0.0270
(0.0529)
0.00723
(0.0954)
-0.0784
(0.155)
-2.059**
(0.935)
3,735
3,733
3,155
Has website
Observations
(6)
j30f