Using Household Surveys as a tool for monitoring Poverty, Governance and Democracy in Africa and the Andean Region Javier Herrera Mireille Razafindrakoto Fran

1
Using Household Surveys as a tool for
monitoring Poverty, Governance and Democracy in
Africa and the Andean Region Javier Herrera
Mireille RazafindrakotoFrançois Roubaud DIAL -
IRDWorkshop on Governance Indicators
New-Delhi, April, 20-22, 2005
2

• The global framework
• Objective
• Main partners
• The implementation process
• Importance of the dissemination policy
• The Tool
• The Mirror Survey
• Results / Type of indicators
• Desagregated data (by sex, by level of income,
  level of education, ethnic groups, etc.)
• Comparable data
• Geographical data
• Time-series
• Multi-dimensional approach

DIAL Développement Institutions et Analyses de
Long terme
3
Objective

• ? To refine measurement tools and to test the
  capacity of statistical household surveys (1-2-3
  survey) to provide relevant information on key
  aspects of democracy and governance.
• ? Implementation of representative sample surveys
  on households experiences, perceptions and
  expectations
• ? Methodology Qualitative modules on Governance
  and Democracy (opinion polls) grafted onto
  conventional household quantitative survey (the
  1-2-3 survey on employment, informal sector,
  consumption and poverty ? extended to Governance
  and Democracy issues).
• ? Multi-country approach
• In Africa 7 capital cities of Western African
  countries (Benin, Burkina-Faso, Côte dIvoire,
  Mali, Niger, Senegal, Togo),
• 7 cities of Madagascar.
• In Latin America 5 countries of the Andean
  Community (Bolivia, Columbia, Ecuador, Peru and
  Venezuela

DIAL Développement Institutions et Analyses de
Long terme
4
The global framework

• Main partners
• Andean Community General Secretariat and
• Statistical offices of the region
• Afristat and Statistical Offices of the region
• Surveys are conducted by National Statistical
  offices
• DIAL responsible for coordination and
  technical assistance
• ? Integration in the National Statistical System
• (? Institutionalisation can be considered)
• ? Statistical Offices function to provide data
  to help decision-making process
• ? In many very poor countries, Statistical office
  is the most qualified to implement surveys with
  large sample and with reliable results
• Mobilisation of existing tools (Household
  surveys already implemented) ? Marginal cost

DIAL Développement Institutions et Analyses de
Long terme
5

• Implementation process
• Design of the questionnaire is based on more
  than five years experiences in Madagascar
  (started in 1995)
• First phase in Madagascar
• variable thematic modules (adjusted every years)
• ? Test of the relevance of questions included
  in the survey
• ? Importance of the dissemination policy
• Systematic presentation / publication of the
  result ? wide audience public conference,
  dissemination in the media, etc. (thirst for
  evidence-based information, for meaningful
  figures)
• ? Reactions ? validation of the relevance of the
  questions
• ? Gives raise to specific demand from different
  stakeholders (? partnership with civil society,
  media)
• ? Second phase at a regional level
• the same questionnaire in 7 capital cities in
  West Africa (test of relevance)
• The 1-2-3 household survey was already planned in
  these countries
• The suggestion of added modules on governance and
  democracy
• met the demand in some countries

DIAL Développement Institutions et Analyses de
Long terme
6

• The tool
• Module grafted onto conventional household survey
  
• Main principles
• Light, flexible tool ? Reconductible
  ?Time-series
• Transparency, representativeness of the
  information collected (voicing)
• Quantification of the phenomenon.
• ? combination and monitoring over time of 3 types
  of information
• - subjective issues (assessment of the working
  of democracy and the State, level of support for
  policies, perception of discrimination, value
  systems),
• - objective facts on social practices and
  behaviour (access to public services, political
  involvement, incidence of corruption etc.)
• conventional socio-economic characteristics
  (gender, education, level of income, age,
  migration, employment, income, consumption etc.).
• ? multiple possibilities to break down the
  information according to different category of
  households or individuals (by level of income, by
  sex, by ethnic group, etc.).

DIAL Développement Institutions et Analyses de
Long terme
7

• The mirror survey
• To round out the collection of surveys on
  Governance and Democracy in seven West African
  capitals and in Madagascar (Antananarivo),
• ? an additional survey to get the opinions of a
  certain number of Southern and Northern experts
  (researchers, development practitioners,
  decision-makers, senior civil servants,
  politicians, etc.).
• The aim ? to compare answers from the population
  surveyed in each country with the experts point
  of view.

8

• The mirror survey
• Two sets of questions (on one of the eight
  countries the expert know the best, his
  reference)
• - what did the experts believe the respondents
  answered on average.
• - their own answer to these same questions
• (What is your personal opinion?)
• Questions 
• - Specificity of the answers of the population /
  experts, specialists?
• - Knowledge of Northern or Southern experts on
  what happens and on peoples thinking in the
  South?
• Relevance and reliability of indicators based
  only on appreciation of panel of experts?

9

• Some results to illustrate the usefulness of this
  type of surveys
• Results of the mirror survey
• ? In general, it seems that experts are more
  pessimistic on how Africa works than the
  african citizens.
• Answers (values) of the population in the South
  are less specific than experts think (democratic
  values etc.)
• ? confirmation that it is interesting to know
  peoples opinion on reform ? ownership

10
Preliminary results of the mirror survey
11
How far can we trust the experts opinion on
corruption?An experiment based on surveys in
francophone Africa
Percentage Benin Burkina Faso Ivory Coast Mada-gascar Mali Niger Senegal Togo Average

Incidence of corruption
General population 8,7 15,2 16,5 16,3 10,1 8,2 10,8 9,6 13,1
Expert panel (what they believe public would reply) 54,1 35,2 60,7 57,1 52,0 56,1 51,1 62,5 54,0

Believes that making a bribe is acceptable behaviour
General population 3,6 8,2 5,2 10,5 5,0 3,1 2,2 3,8 5,2
Expert panel (what they believe public would reply) 31,3 28,7 29,2 32,9 33,3 33,8 35,5 21,8 31,5

Corruption is a major problem
General population 94,2 87,4 91,0 96,9 88,4 91,6 87,9 82,8 90,3
Expert panel (what they believe public would reply) 84,8 67,4 72,3 76,4 67,2 62,3 69,4 84,0 72,9
Expert panel (personal opinion) 96,3 65,0 94,1 88,9 78,1 72,7 80,5 92,3 85,3

Sources General Household Survey (35,594 persons
interviewed 4500 for each country in average)
Expert panel survey (246 persons surveyed 30
experts for each country in average).
12
How far can we trust the experts opinion on
corruption?

• experts overestimate the extent to which the
  general population experiences corruption.
• Overall, experts hold a far more negative view of
  reality than the general public.
• Not only overestimation there are significant
  differences between the two surveys concerning
  the relative positions of the eight countries.
• There is in fact no correlation between the rates
  of corruption measured by the household surveys
  and by the mirror survey (-0,19, n.s.)
• On the other hand, the expert opinion results
  drawn from the mirror survey are similar to
  corruption indicators found in international
  databases (-0.52 with KKZ control of corr.)

13
How far can we trust the experts opinion on
corruption?

• Experts dont necessarily have a good
  understanding of the real extent of corruption in
  the 8 countries surveyed.
• The question what is really being measured in
  these perception indicators (drawn from experts
  panel)?
• The results do not invalidate the relevance of
  such indicators since they do capture a common
  perception linked to corruption phenomena, even
  if they dont correspond to the true experiences
  of corruption.
• They should be combined with a new set of
  indicators based on objective measure and not
  only on perception in order to understand the
  full complexity of corruption.

14
Results of the household surveysIncidence and
determinants of the petty corruption in West
Africa
Sources 1-2-3 Surveys, PARSTAT, WAEMU 2001-2003
Note  logit model (selection bias correction
procedure). (resp. --)  significant
coefficient (positive resp. negative) at 5 .
(resp. -)  idem at 10 . n.s.  non significant
at 5 ).some robust findings ? profile of
groups that are victims of corruption. -The
wealthiest and heads of household appear to be
more especially targets of corrupt officials-
Youth (the risk decreases with age) and being a
foreigner (this is the case in Abidjan) increase
vulnerability, other things being equal. -
Overall and contrary to the preconceived idea--gt
ethnic group and religion do not affect the rate
of corruption (except very few cases) - Civil
servants appear to be less likely to be victims
of corruption (civil servant solidarity).
15
Incidence et determinants of the petty corruption
in West Africa
Incidence and cost of corruption in Niamey
Sources 1-2-3 Surveys, PARSTAT, WAEMU 2001-2003
? Poorest quartile (in terms of income) is less
victim of corruption ? The annual total amount
paid by households (victims) ? 16 of the
income of the poorest quartile of the
population ? 1 of the income of the wealthiest
quartile
16
Public wages and corruption in Madagascar
1995-2001 Measure of the incidence of
corruption (objective extent of this phenomenon
the proportion of inhabitants of the capital who
have fallen victim to corruption during the past
year (small-scale corruption)) The measurement
of this type of indicator is quite exceptional in
developing countries. ? quantifies the problem
spurs on the authorities to take steps to deal
with it. Beyond ? monitoring over time (since
2000 these indicators have been included every
year in official statistical surveys) improves
understanding of the phenomenon. Figure
illustrates the link between improving civil
servant wages and a sharp drop in the incidence
of corruption between 1995 and 2001.
Source Source Razafindrakoto and Roubaud (2001)
based on employment surveys 1995-2001,
MADIO/INSTAT.
17
For more informationDIAL websitewww.dial.prd.
fr
18
ANNEXES
19
Preliminary results of the mirror survey
For Senegal, 3 modalities (country/ethnic
group/religion, sect) instead of 2 like in the
other countries (country/ethnic group)Sources
1-2-3 Survey 2001/02, PARSTAT. Representative
sample of adults aged 18 and over.Mirror
survey  106 individuals (32 on Abidjan, 17
Cotonou, 15 Dakar, 11 Ouagadougou, 10
Bamako)A priori, the experts origine
(Northern/Southern) is not a discriminatory
factor explaining answers for the mirror
surveyDifferences / mistakes concerning
experts appreciation confirmed when results are
compared country by country.