The Sources and Uses of Survey Data on Armenia

Overview BEEPS LiTS Population Census Household Surveys Extra Slides

The Sources and Uses of Survey Data on
Armenia

Shushanik Hakobyan

University of Virginia

September 18, 2009


Outline

1 Overview of Micro Data

2 Business Environment and Enterprise Performance Survey
(BEEPS)

3 Life in Transition (LiTS)

4 2001 Population Census

5 Household Surveys


Types of Micro Data

Unit of observation
A ﬁrm
An individual

Data structure
Cross-section data: Observations on a set of variables in a
given period, t, for individual units i=1,2,. . . ,n.
Independent cross-sections => Pooled data
Panel (or longitudinal) data


Micro Data on Armenia

Firm-level data
Business Environment and Enterprise Performance Survey
(BEEPS)

Individual-level data
Life in Transition Survey (LiTS)
Population Census
Household Surveys (CRRC, NSS)


Sample Questions

Sample Questions
How tax evasion varies with frequency of bribes to tax ofﬁcials?
How likely are the individuals in one income group to move up,
down, or remain in the same income group over time?
How remittances inﬂuence spending on education?
What is the pattern of educational assortative mating in
Armenia? Do women marry up, down, or men with the same
educational attainment?
What is the return to education in Armenia (do wages rise with
educational attainment)?
How to measure income distribution using household surveys?


Firm-Level Data

Business Environment and Enterprise Performance Survey
(BEEPS)
Examines the quality of the business environment
Conducted by the EBRD and World Bank
Surveys firms in 26 transition countries plus Turkey
Rounds/number of surveyed firms: 1999/4,104; 2002/6,667;
2005/9,655
2005 Manufacturing Overlay: 1,713 enterprises in 7 countries
Panel Component: 1,500 firms surveyed in both 2002 and 2005
Comparator Surveys in 2004: 3,300 enterprises in 7
non-transition economies
Format: STATA and CSV
http://www.ebrd.com/country/sector/econo/surveys/beeps.htm
Also available at http://www.enterprisesurveys.org/ (Registration
required)


Application of BEEPS

How tax evasion varies with frequency of bribes to tax ofﬁcials?

Fraction of Sales Concealed by Frequency of Bribes to Tax Ofﬁcials

Variable Mean Std. Dev. N
1 Never 0.12 0.21 2,445
2 Seldom 0.21 0.25 763
3 Sometimes 0.21 0.24 729
4 Frequently 0.28 0.27 455
5 Usually 0.31 0.27 236
6 Always 0.31 0.28 181
7 Not Reported 0.18 0.28 288
Total 0.18 0.25 5,097
Missing – – 643

Source: Joulfaian, David, Bribes and Business Tax Evasion


Determinants of Tax Evasion
Tobit estimates of Determinants of Tax Evasiona

a Estimates control for industry, size and country.

* Signiﬁcant at the 5 percent level, ** at 10 percent level.


Individual-Level Data

Life in Transition (LiTS)
Provides insights into how transition has affected the lives of
people (comprehensive assessment of relationships among life
satisfaction and living standards, poverty and inequality, trust in
state institutions, satisfaction with public services, attitudes to a
market economy and democracy)
Conducted by the EBRD and World Bank
Covers 27 countries in transition plus Turkey and Mongolia
Surveys 1,000 households in each country (29,000 in total)
Retrospective questions about employment in the period
between 1989 and 2006
Format: STATA and CSV
http://www.ebrd.com/country/sector/econo/surveys/lits.htm


LiTS: Income Mobility

How likely are the individuals in one income group to move up,
down, or remain in the same income group over time?

Subjective ranking: income mobility between 1989 and 2006
2006 Quintiles
1989 Quintiles 1 2 3 4 5 All Individuals
poorest 20% 26.7% 53.0% 16.9% 1.3% 2.1% 100% 267,362
2nd quintile 8.8% 27.7% 58.1% 5.4% 0.0% 100% 320,140
3rd quintile 10.5% 34.0% 44.0% 10.3% 1.2% 100% 506,791
4th quintile 11.5% 42.3% 41.0% 5.2% 0.0% 100% 720,929
richest 20% 18.5% 30.0% 36.6% 10.6% 4.3% 100% 229,474
All 13.6% 38.0% 40.8% 6.6% 1.1% 100% 2,044,696

Computed from the 2006 LiTS household survey (see data page of www.aea.am to download)


LiTS: Remittances and Education Spending
How remittances inﬂuence spending on education?

Summary Statistics
No Remittances With Remittances
Variables of interest Mean Std Dev Mean Std Dev
Education Expenditures 184 363 389 555
Health Expenditures 98 227 148 302
Male 71% 46% 55% 50%
Age of Head 50 14 50 17
School age children 1.79 0.83 1.70 0.86
University education 17.8% 38.3% 25.3% 43.7%
Post-graduate education 0.4% 6.0% 0.0% 0.0%
Receive social beneﬁts 20.2% 40.2% 9.9% 30.0%
Gifts from locals 2.0% 13.9% 6.6% 25.0%
Observations 461 91


LiTS: Remittances and Education Spending
Dependent variable: Education expenditures
Specification: OLS

Variables Coefficient s.e. t-statistic
Remittances 243.21* 57.06 4.26
Remittances most important -155.52** 81.20 -1.92
Male head -13.65 36.78 -0.37
Age – head -1.09 6.90 -0.16
Age squared 0.00 0.06 -0.05
Head education – university 186.43* 42.37 4.40
Head education – post-graduate 1396.56* 270.195 5.17
Head unemployed -47.45 122.00 -0.39
Receive social benefits -157.23* 42.70 -3.68
Receive transfers from charities -77.96 171.51 -0.45
Receive gifts from locals 40.85 100.82 0.41
Constant 250.14 168.42 1.49
Observations 552
* Significant at the 5 percent level, ** at 10 percent level.



Population Census
Integrated Public Use Microdata Series International
(IPUMS-International)
Free access to census microdata for 44 countries
2001 Armenian Population Census: 10% sample
Universe: 81,929 households, 326,560 individuals
Format: ﬁxed-format ASCII, supported by STATA, SAS and
SPSS
https://international.ipums.org/international (Registration
required)


Educational Assortative Mating

Deﬁnition: Positive (negative) assortative mating on a characteristic
means that individuals tend to match with partners who are similar
(dissimilar) with respect to that characteristic.
Do the Armenian women marry up, down, or men with the same
educational attainment? How these choices change across different age
cohorts and how differences in age, among others, inﬂuence the choices
made?
Data: 2001 Armenian Population Census
Restricted to women born between 1921 and 1970: 46,653 couples
Observe age, educational attainment, province or region of
residence
Measures of educational attainment
1 up to 3 years of primary education
2 lower secondary or 4-8 years
3 higher secondary
4 professional
5 university graduate


Descriptive Statistics

Source: Hakobyan, Shushanik and David Joulfaian, Educational Assortative Mating in Armenia


Educational attainment of partners by female birth cohorts


Ordered Probit Estimates of Choices

Estimates control for region ﬁxed effects.



Household Surveys
NSS Household Surveys
not publicly available
surveys approximately 4000 households
CRRC Data Initiative


2003 NSS Household Survey

What is the return to education in Armenia?

Summary Statistics: 2003 Monthly Salary by Education Level (in Drams)

Education Number of Obs Avg monthly wage Std Dev
Less than Secondary 86 26167 20580
Secondary 585 29992 24669
Secondary Special 781 25781 24119
Less than High 24 34534 28577
High Education 810 34758 92064
Total 2286 30146 59302

Restricted to ages 25 to 60 and those who report monthly compensation only


OLS: Return to Education

w i = β 0 + β 1 Age + β 2 Age 2 + β 3 Second + β 4 Secspec +
i i i i

+ β 5 Lesshigh + β 6 High + β 7 Female + ε
i i i i

All Women only Men only
Variable Coef. t-stat Coef. t-stat Coef. t-stat
Age 2338** 1.82 485 0.49 3502 1.47
Age squared -27** -1.77 -6 -0.94 -41 -1.44
Secondary 5494 0.82 5148** 1.69 5016 0.44
Secondary Special 4525 0.68 1981 0.97 4730 0.42
Less than High 12826 0.95 7122 0.67 16080 0.63
High Education 13135* 1.98 5387 1.37 18758** 1.66
Female -22051* -8.96
Constant -14828 -0.56 5172** 1.81 -40109 -0.81


Income Measures

Definitions of income based on Luxembourg Income Studies
(LIS - http://www.lisproject.org/)
Market Income (MI): gross salaries/wages, self-employment
income, property income, pensions
Gross Income (GI): MI plus old-age and survivors benefits,
sickness benefits, disability benefits, child benefits,
unemployment compensation benefits, maternity benefits,
military benefits, social insurance, social assistance, near-cash
benefits (housing), alimony/child support, private transfers,
scholarship, other income
Net Disposable Income (DPI): GI minus income taxes,
mandatory employee contributions, mandatory contributions
for self-employment


Income Distribution

Highly skewed distribution

Market Income Distribution
40
30
Percent
2010
0

0 100000 200000 300000
Household Market Income


Lorenz Curve and Gini Coefﬁcient

Lorenz curve of market income
Stata command: glcurve varname, lorenz

Gini coefﬁcient of market income = 0.52
Stata command: ineqdeco varname


Computing Income Deciles
Sections used: Diary of income, Education, Monetary and Commodity Flows
between Households
Time unit: month
Currency consistency: convert dollars and rubles into drams
Unit of observation: Individual
Stata command
xtile newvar = varname, n(10)
tab newvar, summ(varname)
Market Income per capita Gross Income per capita

Deciles Mean Std Dev Mean Std Dev

1 0 0 203.57 340.69
2 730.19 185.24 1819.80 478.10
3 1558.42 384.50 3458.48 461.78
4 3194.01 533.58 5236.65 513.76
5 4977.41 453.64 6799.48 466.63
6 6343.55 379.53 8483.87 581.32
7 8101.41 661.70 10562.37 640.52
8 10898.31 964.11 13365.54 982.29
9 15109.87 1634.80 17719.36 1596.65
10 29160.64 12751.66 33421.15 16059.51


Thank you!
For questions, please email
sh9db@virginia.edu.


Tobit: Marginal Effects

where λ is inverse Mills ratio.

The Sources and Uses of Survey Data on Armenia

More Related Content

Viewers also liked (20)

Similar to The Sources and Uses of Survey Data on Armenia (20)

More from CRRC-Armenia (10)

The Sources and Uses of Survey Data on Armenia