Indicators based on representative surveys of firms, households and service providers Javier Herrera (IRD - DIAL) Mireille Razafindrakoto (IRD - DIAL), and François Roubaud (IRD - DIAL) Seminar on “The Empirics Of Governance“ May 1 – 2, 2008, Washington DC 1 PRESENTATION SCHEME I. Uses of household surveys II. Taking into account sources of bias III. Objective vs subjective corruption indicators IV. Conclusions & Perspectives DIAL Développement, Institutions et Analyse de Long terme 2 I. Uses of household surveys • Household survey as a « voicing » instrument; public awareness on corruption (particularly in of authoritarian political contexts). • Complex links with policy 3 Historical perspective First estimation of the extent of corruption in 1995 in Madagascar (with Household survey) Headlines in the press: « Outcry against corruption! » It then became impossible to ignore the problem 4 MEASURE 1995 1998 1995 It then became impossible to ignore the problem the Ministry of Justice took steps to introduce a DEFINITION OF POLICY Interactions QUALITY OF INSTITUTIONS / Political Regime Positive Impact system of sanctions. a draft law on the fight against corruption “ … It is important to bear in mind that, on the basis of the statistical survey conducted in May 1995 by the MADIO project, co-financed by the French Ministry for Cooperation and the European Union: ‘the issue of corruption seems to be a recurring problem which haunts the capital’s inhabitants. 96% of them consider that it is a major problem in Madagascar’. ‘Over 40% of people aged over 18 in the capital had had to pay a corrupt civil servant during the previous year’. Whatever the credibility of this survey and the interpretations that have been drawn from it, there is no doubt that corruption is a social problem in Madagascar … and it is necessary to combat the practice of corruption as energetically as possible”. BUT Draft law rejected by the Government Council in 1999 Importance of context & characteristics of institution (~ authoritarian regime and problem of governance) 2002 Institutional change Stress put on transparency 2003 Creation of an independent council (CSLCC) 2004 Independent anti-corruption office (BIANCO) 5 Headlines in the press in may 2005: « More confidence & less corruption » after the public conference presenting the first results of the survey on « governance, democracy and fight against poverty » 6 GOVERNANCE An improvement of civil servant wages or an active anti-corruption policy a sharp drop in the incidence of corruption 180 44 40 36 32 28 24 20 16 12 8 4 0 170 160 150 Active anticorruption policy Political crisis 130 120 110 100 90 1995 1996 1997 1998 Level of corruption Sources : 140 Base 100=1995 % Civil servant salaries and corruption levels in Madagascar 1995-2004 1999 2000 2001 2002 2004 Civil servants real wages (right scale) Razafindrakoto, Roubaud (2002) and 1-2-3 Survey 2002-2004, INSTAT, DIAL, authors calculations. 7 II. Taking into account sources of bias • Who answers the question matters (random sample selection within the household) • But also it matters to know more about who don’t answer (non respondent profile) • Taking into account the working of corruption (middlemen, social norms, etc.) • Robustness needs to be assessed • Informal production units needs adapted survey approach (the 123 surveys) 8 III. Taking into account sources of bias • The wording and sequence of question matters (but not only when measuring corruption, ie. labor surveys). Pilot surveys are important. • Non response to governance questions in HH surveys are lower than in standard questions (ie. Income) • Unobserved heterogeneity (« over optimistic »/ »under optimistic », use panel data -ongoing research) 9 The surveys Governance and Democracy Modules grafted onto official household survey conducted by National Statistical Institutes Survey support (stratified multi-stage sample surveys: area/HH) for the Modules (Individuals): • In Africa: 1-2-3 Survey (LFS, informal sector, consumption and poverty) - Madagascar 1995-2004, Capital + urban areas (2000, 2001), rural, entreprises surveys (time series) - West Africa 2001/2003, in 7 WAEMU capital cities (Abidjan, Bamako, Cotonou, Dakar, Lomé, Ouagadougou and Niamey) 35 594 persons interviewed • In Latin America: Standard National Household surveys (ECH, Bolivia; SIE-ENEMDU, Ecuador; ENAHO, Peru). National and regional inference for Peru and Ecuador. - Peru 2002-, national level (18 000 HH sample in 2002; continuous survey from 2003 to date). - Ecuador 2004, (20 000 HH) - Bolivia 2004, national (1 700 HH). -The survey is taking place in Colombia during the 2nd term of 2005. More than 50 000 persons interviewed 10 Strong points / Basic principles: Conducted by National Statistical Institute Systematic Presentation / Publication of the result wider public Ownership Integration in the National Statistical System Validation Democratic debate demand Reliability Relevance Bottom-up approach Investment in capacity building Light, flexible tool Reconductible time-series Sustainability Marginal cost HH survey « Voicing » empowerment, accountability Supply side Interactions Demand side 11 Non response There is (often) more willingness to answer questions on governance and democray than usual economic questions Non response rate to selected questions West Africa Mada % Coton ou Ouag adougo u Abidj an Bama ko Niame y Dakar Lomé Antan anarivo Opinion on governement functioning 0 3.6 0.9 1.1 2.7 5.2 3.5 0.1 2.2 Opinion on democracy 0 2.8 0.5 0.9 1.9 3.1 0.3 0.2 1.3 Income level (continuous) 40.2 54.3 40.9 43.3 51.6 56.3 37.2 47.3 Income level (continuous or discrete) 2.3 6.5 3.2 6.5 14.1 9.2 1.6 0.1 Total Sources : 1-2-3 Surveys, Phase 1, Governance and Democracy module, 2001/2003, National Statistical Institutes, AFRISTAT, DIAL, authors’ calculations. 12 Robustness Probabilistic surveys allows a quantitative assessement of indicators’ precision (also relevant for group and intertemporal comparaisons) In your opinion, corruption since last year? has increased 95% confidence interval has decreased 95% confidence interval is the same 95% confidence interval don't know 95% confidence interval Total rural urban Total 31,0 39,1 36,6 [29,0 - 33,0] [37,4 - 40,9] [35,3 - 38,0] 8,43 10,47 9,85 [7,5 - 9,5] [9,5 - 11,6] [9,1 - 10,7] 49,1 48,8 48,9 [47,5 - 14,0] [46,9 - 50,6] [47,4 - 50,3] 11,5 1,6 4,7 [10,3 - 12,9] [1,3 - 2,1] [4,2 - 5,2] 100 100 100 Source: our estimation based on ENAHO july 2003-june 2004, INEI, Peru. 13 Taking into account the working of corruption: middlemen; failed corruption; “accepted” corruption, amounts paid, access to services; links with inefficiency… Source: ENAHO 2004 HH survey Peru 14 Corruption incidence taking into account access to public services Incidence (percentage of households victims) has decreased Total amount paid has decreased in real terms The poor as well as the rich had benefited from this improvement Incidence and cost of corruption (2002/2003 and 2004 in Antananarivo) En 2002/2003 Total En 2004 Income per capita Total 1st quartile 4th quartile Income per capita 1st quartile 4th quartile 16,3 40,4 27,4 11,9 46,8 22,4 20,1 31,5 29,3 8,4 29,9 12,0 6,0 34,2 9,1 10,3 22,4 13,3 Mean (HH victims of corruption) 381 219 682 300 77 484 Median (HH victims) 25 3,3 20 7,8 50 3,1 30 1,2 14 1,8 50 1,2 Incidence (all households) No contact with civil service Incidence (households victims) Total annual amount (1 000 Fmg) % of income (HH victims) Source : Enquête 1-2-3, modules qualitatifs, 2003 et 2004, INSTAT, authors calculations. 15 “Don’t know” answer is not random: it concerns mainly the poor, the rural, the less educated, the socially excluded Importance of democracy Do you think that having a democratic government is important?: Yes More or less Not much No Does not know Total Rural % Urban % Not poor % Poor % Total % 34.8 38.0 13.7 4.6 8.9 100.0 55.9 30.6 8.4 3.4 1.7 100.0 56.3 29.7 8.7 3.3 2.1 100.0 41.5 36.6 11.6 4.3 6.0 100.0 49.4 32.9 10.0 3.7 3.9 100.0 Authors’ own calculations from ENAHO 2002-IV. 16 Corruption incidence for the informal sector production units can be measured Cotonou %Incidence of Informal Units that had a problem with civil servants last year Ouagadougou Abidjan Bamako Niamey Dakar Lome Total 4.7 5.0 7.0 3.5 6.2 8.5 6.2 6.2 By paying a fine 75.5 76.7 42.9 38.5 40.4 49.2 44.5 47.3 By giving a “gift” 13.9 5.7 50.2 32.6 24.7 38.1 45.1 41.3 other 10.5 17.6 6.9 28.9 34.9 12.7 10.4 11.4 % that settled the problem: Source: Alain Brilleau et al. « Le secteur informel : Performances, insertion, perspectives, enquête 1-2-3 phase 2 », STATECO n°99, 2005 p.82. 17 III. Objective vs subjective corruption indicators • Expert’s surveys vs. HH surveys • Do they measure the same phenomena? (petty vs. big corruption) • Minding the gap between : -perceptions and objective indicators -aspirations and perceived outcomes 18 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 specialists’ point of view. 19 The mirror survey (…) Two sets of questions: - what the experts think 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 people’s thinking in the South -Relevance and reliability of indicators based only on appreciation of panel of experts DIAL Développement, Institutions et Analyse de Long terme Sample size : 250 experts (30 per country in average) 20 The need for complementary approaches / indicators How far can we trust the experts’ opinion on corruption? Discrepancies between real extent of corruption and experts’ perception in Francophone Africa On the level of corruption / To what extent corruption acceptable On the relative positions (rank) of the different countries Population who believes that making bribe is acceptable Mirror Survey 40 35 30 25 20 HH Survey 15 10 5 er ag e Av Lo m é ak ar D Ta na Ba m ak o N ia m ey C ot on ou O ua ga Ab id ja n 0 General population * (% from Household (HH) survey) Expert panel (what they think could be the % of population who believes that making a bribe is acceptable) Sources: General Household Survey (35,594 persons interviewed; 4500 for each country in average); Expert panel survey or Mirror survey (246 persons surveyed; 30 experts for each country in average). * In Madagascar, results are drawn from the 2003 survey. Authors calculations. 21 DEMOCRACY Measuring the gap between aspirations and effectiveness as regards democratic principles Perception of the main democratic principles by the population Are they fundamental? Are they respected in the country? 100 90 80 70 60 % 50 40 30 20 10 0 Cotonou Ouaga Abidjan Bamako Niamey Dakar Lomé Tana The six principles of democracy are: Fundamental Respected Sources : Enquêtes 1-2-3, module Démocratie, 2001/2003, Instituts Nationaux de la Statistique, AFRISTAT, DIAL authors calculations. 22 DEMOCRACY Measuring the gap between aspirations and effectiveness as regards democratic principles Lomé Religious freedom Free, transparent elections 100 Abidjan Political freedom (right to choose party) 50 Religious freedom 0 Freedom of expression, of the press Equality before the law Fundamental Respected Free, trans parent elections Fundamental Respected 80 60 40 P olitical freedo m (right to cho os e party) Religious freedom 20 0 Freedo m of expres s ion, of the pres s Freedo m to travel Equality before the law Free, transparent elections 100 50 0 Equality before the law Bamako Free, transparent elections 100 50 Political freedom (right to choose party) Antananarivo Political freedom (right to choose party) Religious freedom Free, transparent elections 100 50 Freedom to travel Respected Religious freedom 50 0 Freedom of expression, of the press 0 Fundamental Free, transparent elections 100 0 FundamentalEquality before the law Respected Freedom of expression, of the press Freedom to travel Fundamental Respected Political freedom (right Religious to choose freedom party) Equality before the law Fundamental Respected Niamey Freedom of expression, Freedom to of the press travel Freedom to travel Cotonou 10 0 Religious freedo m 50 0 Freedom to travel Dakar Free, transparent elections 100 Po freedo to c p Free expre the Freedom to travel Fundamental Equality Respected before the law Ouagadougou Free, transparent elections 100 Political freedom (right to choose party) Religious freedom Freedom of expression, of the press Freedom to travel 50 0 P fre (r choo Free expr of th Fundamental Equality before the law Respected Equality before the law Sources : Enquêtes 1-2-3, module Démocratie, 2001/2003, Instituts Nationaux de la Statistique, AFRISTAT, DIAL authors calculations. 23 The need for complementary approaches / indicators No correlation between Objective measure (incidence of corruption) Subjective perception of the working of civil service % of negative opinion on the working of the civil service 90% Cor=-0,48 80% Lomé 70% 60% Cotonou Ensemble Niamey Bamako 50% 40% Ouaga Abidjan Dakar 30% Tana 20% 10% 0% 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% Incidence of corruption 24 The global lessons • Advantages of household sample surveys Transparency of measurement procedures Representativeness Quantification, Comparability of indicators over time. • In-depth policy-oriented analyses More appropriate than international indicators and aggregates. •(perception, Both objective (behaviour, actual experiences) and subjective information satisfaction) Monitoring and relating the two fundamental dimensions of these phenomena. • Socio-economic disaggregation These two dimensions can be combined with traditional variables related to the socio-economic characteristics of individuals and households (income, occupation, sex, age, ethnic group, etc.). Possibility to disaggregate information between different population categories (gender, poverty, ethnic groups, discriminated people, etc. • International comparability 25 IV. Conclusions & Perspectives • Measure different dimensions of governance • A detailed understanding of different forms of corruption is necessary to analyse its determinants and consequences • Different sources of potential bias should be considered in designing & analysing surveys • Both, objective and subjective, governance indicators need to be collected to better understand aspiration/outcome gaps and why corruption may become or not a political issue. 26 END For more works on Governance, Democracy and Poverty see our web site http://www.dial.prd.fr Our article: “Governance, Democracy and Poverty Reduction: Lessons drawn from household Surveys”, International Statistical Review (2007), vol. 75, issue 1, pp.70-95 is available upon request 27